Pierre never tires: a Vapi voice agent engineered to qualify training leads, handle CPF/Pôle Emploi financing objections, and book appointments — at any scale, without a break.

## Illico Voice AI — Pierre, a Vapi voice agent engineered to qualify training leads, handle CPF and Pôle Emploi objections, and book appointments at any scale

This is the most impressive feat of the Illico project: the design of the mind and voice of Pierre, an artificial voice agent built on the **Vapi Voice AI infrastructure**. Pierre is not a toy chatbot given a voice; he is a carefully engineered commercial agent, shaped through dozens of iterations of persona design, prompt engineering, and real-world call testing, until he became indistinguishable from a seasoned human call center agent — and, in many dimensions, genuinely better.

### Realism and persona engineering

A voice agent is only as good as the illusion of humanity it sustains across the full length of a call. This one is built to that standard.

- **Human vocal mimicry**: vocal fluidity is parameterized to microseconds, tuned across the full Speech-to-Text, LLM, and Text-to-Speech pipeline. Turn-taking, backchannel sounds, breathing pauses, and reactive acknowledgments are all tuned so that the prospect never hears a "robot pause" that would break the illusion. Technically, Pierre is often indistinguishable from a seasoned human agent on the phone.
- **Persona consistency**: Pierre has a name, a background, a tone, and a set of conversational habits that are consistent across every call. He is not a generic AI voice — he is a specific, recognizable character designed for this commercial role.

### Intensive qualification matrix

The intelligence behind Pierre is subjected to drastic **system directives and prompt engineering** that transform a general-purpose LLM into a highly specialized training advisor.

- **Objection handling**: Pierre is explicitly trained to methodically sweep the prospect's objections — lack of time, doubt about ROI, family situation, financing worries — without ever compromising the fluid thread of the discussion. Each objection has a planned response tree that sounds improvised but is engineered.
- **Complex financing analysis**: Pierre performs algorithmic real-time analysis of the prospect's financing solutions — **CPF** (Compte Personnel de Formation), **Pôle Emploi** funding eligibility, private budgets, employer co-funding — and adapts the conversation accordingly. This alone replaces a substantial amount of human expertise that is usually the bottleneck of a training sales team.
- **Active qualification**: Pierre doesn't just answer questions; he actively qualifies the prospect against the agency's ideal customer profile, identifying whether the lead should be booked for a closing call with a human advisor, nurtured, or respectfully declined.
- **Action passing**: once Pierre has identified a qualified prospect and aligned on a time, he orchestrates the actual appointment booking through the Make.com tool-calling layer (see the parallel project), turning intent into a locked CRM slot inside the same call.

### Why Pierre is commercially different

Most voice AI deployments fail commercially because they fail in one of three ways: they sound robotic, they handle objections poorly, or they cannot actually close the loop into a CRM. Pierre solves all three in one coherent product. He is an unparalleled salesperson in his specific domain: tireless, ultra-reactive, consistent, and capable of handling a volume of **thousands of simultaneous calls** without any of the quality drift a human team would experience at scale.

### Engineering discipline behind the scenes

- **Iterative prompt refinement** tied to the analytics dashboard (see parallel project): every flagged failure becomes a candidate prompt adjustment.
- **Guardrails**: Pierre is explicitly prevented from making promises outside his remit, from discussing competitors, or from improvising on pricing.
- **Escalation paths**: when a call goes outside Pierre's territory, he gracefully escalates to a human — without breaking persona and without losing the prospect's trust.

### The delivered outcome

- A voice agent that handles training lead qualification, CPF and Pôle Emploi analysis, objection handling, and appointment booking — end to end, on every call.
- Quality that is indistinguishable from a seasoned human agent, at a fraction of the cost.
- Effectively infinite scalability: thousands of simultaneous calls with zero degradation.

### Technology stack

- **Vapi** as the Voice AI infrastructure.
- **Speech-to-Text** and **Text-to-Speech** components tuned for human-grade conversational fluidity.
- **Linguistic prompt engineering** as a core discipline for persona, objection handling, and qualification logic.
- **Artificial agent architecture** integrated with CRM tool-calling for real-world commercial action.

---

**Ready to automate your business?** [Discover our AI Automation service →](/expert-automatisation-ia)

Pierre, Vapi voice agent: training lead qualification, CPF/PE financing analysis, autonomous appointment booking. Indistinguishable from human. Thousands of calls.

Illico: Ultra-Realistic Vapi Voice Agent

Voice AI (Illico) - Vapi Agent Engineering

When the Illico AI says 'I'll book that for you' — a Make.com webhook checks GoHighLevel availability in real time and locks the appointment in the CRM, automatically.

## Illico Voice AI — the Make.com and GoHighLevel orchestration that turns a conversation into a confirmed CRM appointment in a single beat

This is the invisible but absolutely essential plumbing that allows an AI Voice Agent to have a real impact on a corporate booking calendar. Without this orchestration layer, the AI agent can talk beautifully about appointments — but cannot actually book one. With it, the AI agent closes appointments mid-conversation, in real time, with zero human latency, straight into the company's CRM.

### The power of tool calling, done right

The defining capability of a modern voice agent is its ability to take action mid-conversation, not just describe possible actions. This project implements that capability through a tight **tool calling** architecture wiring Vapi, Make.com, and GoHighLevel together.

- **Intention interception**: while conversing on the phone with the prospect, the AI agent determines in real time whether it can close the discussion with an appointment. The moment the intent is clear, the AI computationally triggers a tool call, which is instantly intercepted by a **Make.com** webhook layer. No human dispatching, no waiting queue, no manual relay.
- **Asynchronous calendar verification**: within a fraction of a second, Make.com queries the **GoHighLevel** CRM calendar to validate the genuine availability of the human advisor at the requested time. This is real availability, checked against the live CRM, not a cached approximation.
- **Automated fallback slots**: if the requested slot is already taken, the loop immediately returns control to the AI with a proposed list of adjacent available slots. The AI proposes them to the prospect vocally, conversationally, without stumbling or breaking the flow of the call.
- **Closing in the CRM**: once the prospect accepts a slot, Make.com concretely locks the qualified appointment in the GoHighLevel agenda, attaches all the context gathered during the call (qualification tags, financing method, notes, transcript link), and triggers any downstream automation the business needs — confirmation SMS, calendar invites, Slack pings to the human advisor.

### Why this architecture is the real unlock

Most voice AI demos stop at "the AI sounded nice on the phone." That is the easy half of the problem. The hard half is turning a conversation into an actual, committed, CRM-grade appointment — with real-time availability checking, fallback handling, error recovery, and full context preservation. This orchestration handles every piece of that hard half, with production-grade reliability.

### Production-grade reliability

- **Error recovery**: every failure mode is explicitly handled — CRM API unavailable, slot vanished between check and booking, prospect changes their mind mid-confirmation. Nothing is left to chance.
- **Idempotency**: no double-bookings, ever. Every appointment operation is keyed on a unique call ID so that retries never create duplicates.
- **Auditability**: every tool call, every CRM operation, every state transition is logged with enough detail to reconstruct the full commercial journey of any prospect after the fact.
- **Bi-directional synchronization**: when an advisor updates the CRM, the system reflects that in the availability logic immediately, so the AI never proposes a slot that an advisor has just blocked.

### The delivered outcome

- Appointments booked mid-call, in real time, straight into GoHighLevel.
- Zero human latency between prospect intent and CRM confirmation.
- Full context preservation — every appointment carries the complete call context.
- A reliable, audited architecture that scales to thousands of calls per day.

### Technology stack

- **Make.com** as the orchestration and webhook layer.
- **GoHighLevel (CRM)** as the calendar and prospect system of record.
- **API architecture** built around tool calling and bidirectional webhooks.
- **Bidirectional webhooks** for real-time availability checks and confirmations.

---

**Ready to automate your business?** [Discover our AI Automation service →](/expert-automatisation-ia)

Make.com orchestrates Vapi and GoHighLevel: real-time calendar check, auto appointment booking, fallback slots. Zero human latency in the funnel.

Illico: Vapi + Make.com + GoHighLevel CRM

Voice AI (Illico) - Make & GHL Orchestration Scenarios

A live Next.js dashboard tracking every call handled by the Illico AI voice agent — live conversion rates, appointment stats, and error logs for continuous prompt refinement.

## Illico Voice AI — the live analytics dashboard that gives management surgical visibility into the performance of their AI telephone salesperson

This project is the analytical eye that monitors the AI telephone salesperson of a prestigious training placement agency. Deploying an AI voice agent into production is one thing; knowing what it actually does at scale, in real time, across thousands of calls, is another. This dashboard closes that gap. It transforms an otherwise opaque AI system into a fully measurable commercial asset, with conversion rates, error tracking, and call-level insights at management's fingertips.

### Real-time monitoring and insights

The dashboard is an ultra-reactive **Next.js** application hosted on **Vercel**, chosen specifically to guarantee total data fluidity: no perceptible latency between a call ending and its insights appearing in the dashboard. For a management team used to waiting 24 hours for a spreadsheet export, this live visibility is transformative.

The dashboard allows human decision-makers to instantly visualize the return on investment of their AI voice switchboard:

- **Live conversion rate**: real-time calculation of concrete physical appointments generated compared to the volume of calls handled. Conversion is tracked by funnel stage, by time of day, by prospect type, and by campaign, so management can see not only the global rate but the composition behind it.
- **Error tracking and anomaly detection**: proactive monitoring of misunderstandings, ambiguous objections, conversational dead ends, and outright bugs surfaced during calls. Every anomaly is surfaced with its full conversational context, so prompt engineers can refine the AI's directives with precision rather than guesswork.
- **Per-call timelines**: each call is represented as a structured timeline — greeting, qualification, objection handling, financing discussion, appointment proposal, confirmation. Management can click through any call to see exactly where it succeeded or stalled.
- **Comparative reporting**: dashboards compare AI performance against historical human agent performance on the same funnel stages, giving management a clear, quantitative view of the ROI of switching to AI.

### Why live analytics matter for voice AI

Voice AI is a new category for most businesses. It generates both enthusiasm and legitimate skepticism. Without live analytics, management is reduced to anecdotes — "Mrs. Dupont said the robot understood her perfectly" — which is no basis for a scaling decision. With proper analytics, the question becomes quantitative: what is the actual conversion rate, what are the drop-off points, how does performance compare to human agents, and where can prompts be refined to close the gap? This dashboard answers those questions continuously, not once per quarter.

### Prompt refinement as a continuous loop

The dashboard is designed to feed a **continuous refinement loop**. Every flagged error becomes a candidate for a prompt update; every successful call becomes a reference for what good looks like. Over time, the AI improves not by magic but by disciplined observation, because the dashboard makes that observation frictionless.

### The delivered outcome

- A live, real-time view of every call handled by the AI voice agent.
- Conversion, error, and timeline analytics sophisticated enough to run a commercial operation on.
- A continuous feedback loop that turns call data into prompt improvements.
- Complete management confidence in an otherwise opaque AI deployment.

### Technology stack

- **Next.js** for the reactive dashboard application.
- **Vercel** for instant deployment and global edge delivery.
- **Analytical dashboarding** with rich data visualization.
- **Real-time data pipelines** to keep the dashboard continuously in sync with the voice agent's activity.

---

**Ready to build your premium website?** [Discover our Premium Web Development service →](/expert-site-web-premium)

Next.js Vercel dashboard monitoring the Illico AI voice agent: conversion rates, error tracking, per-call insights in real time.

Illico: Analytics Dashboard AI Voice Agent

Voice AI (Illico) - Web Analytic Dashboard

MaxSEO's execution engine: an n8n workflow that generates detection-resistant AI content with randomized table structures and auto-publishes directly to CMS via API.

## MaxSEO AI — the n8n execution engine that turns strategy into AI-detection-resistant SEO content, published straight to CMS

If the MaxSEO Dashboard is the command center, this **n8n workflow** is its formidable execution engine. It is the very essence of "AI-Driven as code": strategic intent enters at one end, fully formatted, published SEO articles come out at the other end, and nothing in between requires a human operator. This workflow is what turns MaxSEO from a nice UI into a genuine industrial SEO machine.

### Prompt engineering and detection-resistance

The hardest challenge in AI text generation today is not producing text — any LLM does that trivially. The hard challenge is escaping Google's anti-spam systems, specifically the **Helpful Content Update** and its successors, which aggressively demote content that looks machine-generated at scale. MaxSEO's workflow is engineered end-to-end around this constraint.

- **Complex semantic footprint**: the n8n scenarios integrate explicit instructions that force the LLM to randomly alternate among **6 distinct comparison table structures**, rotate between multiple introduction patterns, and vary sentence-length distributions across articles. The structural footprint is never the same twice.
- **Signature blurring**: this algorithmic variation makes it possible to deliberately blur the AI's stylistic signature and make the produced text read as purely organic — no two articles share the same skeleton, the same transitional phrases, or the same paragraph rhythm.
- **Anchored factual content**: every article is grounded in the variable library from the dashboard (locality data, pricing bands, seasonal modifiers) so the text is genuinely informative rather than generic filler. Helpful Content cares about helpfulness; this workflow delivers it.
- **On-page SEO discipline**: H1/H2/H3 hierarchy, internal linking, meta title and description, schema hints, alt text for images — all of it is produced to spec in a single pass rather than patched up later.

### Direct publishing via API — no keyboard involved

Once the article is generated, with dynamic injection of local tags, correctly nested headings, internal links, and optimized metadata, n8n does not pause to ask for human authorization. It **pushes the formatted article directly onto the target site's CMS** through the CMS's own API. WordPress, custom headless CMS, proprietary platforms — the connector layer handles them all.

This direct publishing is where most competitors break down. They generate nice articles, then ask a human to paste them into WordPress. MaxSEO closes the loop. From strategic intent in the dashboard to a live, indexed, SEO-optimized article on a production site, the only human time is the strategic decision itself.

### Why this is genuinely different

Most "AI SEO tools" on the market are thin wrappers around GPT with a publish button. MaxSEO's workflow is a production system that takes AI detection seriously, takes CMS publishing seriously, and treats every article as a small deployment rather than a throwaway generation. That discipline is the reason the platform works at scale where other approaches collapse under Google's quality filters.

### The delivered outcome

- SEO articles generated and published without human intervention, at scale.
- Built-in structural variation that resists AI-detection-style ranking penalties.
- Direct API publishing to any CMS with a usable endpoint.
- Full logging of every generation and every publication, for auditability and iteration.

### Technology stack

- **n8n** as the orchestration and workflow engine.
- **LLM (Artificial Intelligence)** for content generation.
- **Prompt engineering** as a first-class discipline, not an afterthought.
- **CMS API integration** for automatic publishing across WordPress and custom CMS platforms.

---

**Ready to automate your business?** [Discover our AI Automation service →](/expert-automatisation-ia)

MaxSEO n8n engine: article generation with AI-detection-proof semantic variation, direct CMS API publishing. From brief to live without intervention.

MaxSEO n8n — AI-Detection-Proof SEO Articles

MaxSEO AI - n8n Workflow Engine

MaxSEO's command center: a Next.js dashboard where SEO experts define AI instructions, formatting templates, and inject hundreds of geolocated variables — total prompt engineering control.

## MaxSEO AI — the command center for AI-powered, geolocated SEO content at industrial scale

MaxSEO is the visible and elegant face of an AI-powered SEO content generation platform. Rather than letting SEO teams beg a chatbot for articles that may or may not match their strategy, MaxSEO gives them a proper command center — a SaaS dashboard where every parameter of AI content generation is configurable, auditable, and reproducible. The goal is simple: transform AI text generation from a creative toy into an industrial-grade marketing asset.

### The command center

We developed a high-performance web management dashboard built on **Next.js** and **React**, designed specifically for SEO experts who need to orchestrate content campaigns at scale. The dashboard is not a wrapper around a chat window; it is a purpose-built cockpit where every decision — from persona to format to geographic coverage — is a first-class configurable object.

### Prompt engineering, but as a real product

The user does not simply ask the AI for an article. They **dictate strict algorithmic rules** that the generation engine must obey, every time, for every article, without drift. This is the defining difference between a serious SEO operation and a hobbyist one.

- **Strict system instructions**: the user defines the psychological, linguistic, and tonal parameters of the AI — persona, reading level, sentence length, vocabulary constraints, forbidden phrases, and much more. The AI becomes a disciplined writer with house rules, not a freewheeling improviser.
- **Variable formatting templates**: the user configures structuring templates — price comparison tables, pros/cons lists, FAQ blocks, CTA blocks, internal linking patterns. Every article assembled by the engine respects the selected template structure, which is essential for SEO consistency across a large site.
- **Geolocated strategy**: the interface allows the **dynamic injection of hundreds of locality variables** — cities, regions, postal codes, neighborhoods, landmarks — for powerful territorial networking. This is the core mechanism that makes local SEO at scale economically possible: a single strategic template yields hundreds of well-differentiated, geographically specific articles, each genuinely useful to readers in its target locality.
- **Data variable library**: beyond geography, the platform supports custom variable libraries (price ranges, typical service durations, seasonal modifiers, etc.), so every generated article is factually rich rather than hollow filler.

### Why this matters commercially

Local SEO is a massive commercial opportunity that most small and mid-sized operators cannot capture because they cannot produce the volume of geographically differentiated content that Google rewards. MaxSEO closes that gap. With the dashboard, a single SEO strategist can set up, review, and deploy a local content campaign across an entire region — with genuinely useful articles, not thin programmatic filler — in a fraction of the time a copywriting team would require.

### Designed for reproducibility and accountability

- **Campaign versioning**: every configuration is versioned, so a team can iterate on prompts and templates without losing the ability to roll back.
- **Preview and approval flows**: articles can be reviewed and approved before they are pushed to production sites, giving teams full editorial control when they want it.
- **Performance linkage**: because every article carries metadata tying it back to its generation configuration, performance data from Google Search Console can be attributed to specific prompt and template decisions — closing the feedback loop.

### Technology stack

- **Next.js** and **React** for the SaaS dashboard.
- **UI/UX design** tailored to SEO power users rather than general audiences.
- **Prompt engineering** as a first-class configurable product surface.
- **Geolocated variable injection** as the engine of scalable local SEO.

---

**Ready to build your premium website?** [Discover our Premium Web Development service →](/expert-site-web-premium)

MaxSEO AI Next.js dashboard: control AI-powered SEO content generation — prompt engineering, geolocated variables, configurable formatting. Local SEO at scale.

MaxSEO AI Dashboard — Geolocated SEO Content

MaxSEO AI - Web Pilot Platform

An n8n SEO orchestra that rewrites content, fires GSC re-indexing requests, and coordinates a full PBN network — entirely automated, without touching a single site manually.

## Intelligent programmatic SEO orchestration — piloting a multi-site PBN from a single n8n control center, without ever touching a site manually

Driving an offensive SEO strategy across an entire network of websites (a Private Blog Network, or PBN) is, for most operators, a brute-force manual chore: logging into each site, rewriting content, tweaking internal links, pushing re-indexation requests one by one. This project replaces all of that with a **complex n8n automation architecture** that treats the network as a single programmable system. The result is a fleet of hundreds of sites operating in lockstep from an automated control center — with the responsiveness, discipline, and safety margins that manual operation simply cannot deliver.

### The modular orchestrator

Rather than intervening site by site, the solution is built as a set of composable algorithmic micro-services, each with a single responsibility:

- **The SEO Modifier**: deeply reworks on-page semantic markup and rewrites the target pages' content in total autonomy. This is where the actual SEO change happens — new headings, updated copy, refreshed internal linking, adjusted metadata. The Modifier is idempotent, auditable, and can be dry-run before being applied live.
- **The GSC Proxy Mapper**: a secure interface layer that connects the orchestrator to the **Google Search Console API**, allowing the system to request immediate re-indexing of hundreds of modifications as they happen. Without this piece, the SEO changes would be invisible to Google for days or weeks; with it, they are picked up within hours.
- **The Master Orchestrator**: the true conductor of the architecture. This top-level scenario distributes tasks across the network, smooths request volume to avoid tripping Google's anti-bot heuristics, enforces safety windows between operations on the same site, and ensures global synchronization so that the network acts as one coordinated whole rather than a chaotic swarm.

### Why each piece matters

A PBN lives or dies on three factors: **content quality**, **technical signals**, and **indexation speed**. The Modifier guarantees the first two at scale. The GSC Proxy Mapper guarantees the third. And the Master Orchestrator guarantees that none of this triggers anti-spam countermeasures — which, for a PBN, is quite literally existential. Lose your indexation and you lose every site at once.

### Safety and discipline built into the architecture

- **Request pacing**: the orchestrator never lets too many requests hit Google (or any one target site) in a short window. All traffic is smoothed, jittered, and rate-limited.
- **Staggered rollouts**: content changes propagate across the network in controlled waves rather than all at once, removing the telltale synchronized footprint that would expose a PBN.
- **Audit trail**: every modification, every re-indexation request, every action is logged with enough detail to reconstruct the full state of the network at any point in time.
- **Kill switch**: if anything looks off — ranking drops, unexpected GSC responses, unusual anti-bot signals — the entire fleet can be paused from a single control point.

### The strategic impact

This is home automation applied to SEO. Managing a fleet of hundreds of sites from an automated control center is no longer a small-agency chore handled by three interns; it is an industrialized operation with the discipline of a software deployment pipeline. The leverage is enormous: a single SEO engineer can now drive a network that used to require a team, and can do so with better safety margins than the team ever achieved.

### The delivered outcome

- A single n8n control plane driving SEO operations across the entire network.
- Automatic content rewriting, internal linking updates, and metadata refreshes.
- Direct integration with Google Search Console for near-instant re-indexation.
- Built-in safety, pacing, and audit capabilities.

### Technology stack

- **n8n** as the orchestration engine.
- **Google Search Console API** for re-indexation requests.
- **Programmatic SEO** principles for modular, repeatable operations.
- **Workflow development** patterns for safety, pacing, and synchronization.

---

**Ready to automate your business?** [Discover our AI Automation service →](/expert-automatisation-ia)

Programmatic n8n SEO orchestration for PBN network: automatic semantic rewriting, Google Search Console API re-indexing, multi-site coordination.

Programmatic SEO n8n — PBN Multi-Site GSC

Intelligent Programmatic SEO Orchestration

10,000+ trading card listings created on Shopify automatically — a Python script that ingests the TCGdex API, builds exhaustive product sheets, and creates collections algorithmically.

## Ingesting 10,000+ trading card products into Shopify with a single asynchronous Python script — built to respect API rate limits and run unattended

How do you propel a monumental catalog of more than **10,000 product references** onto a Shopify store without spending months on manual product entry? With a purpose-built automation script that treats Shopify as a programmable surface rather than a human data entry interface. This project demonstrates exactly that — and in doing so eliminates what would otherwise be thousands of hours of copy-pasting, formatting, and manual collection assignment.

### Massive data ingestion

We designed an autonomous script in **Python** that interacts directly with the public TCGdex API, the reference database for trading card games. The script ingests the entire catalog — card by card — including card names, types, rarities, sets, health points, attack mechanics, energy costs, illustrator credits, language variants, and associated visuals. Nothing is scraped from HTML; everything is retrieved from a clean, documented API, which makes the pipeline robust, maintainable, and easy to re-run whenever TCGdex publishes new expansions.

### Algorithmic e-commerce creation

The raw data is then processed, normalized, and sent via the **Shopify REST API** to automatically generate a coherent catalog inside the store:

- **Exhaustive product sheets**: each card becomes a fully structured Shopify product with title, description, rich HTML body, tags, metafields for rarity and set, variant attributes for language and condition, and its official illustration as the product image. Nothing is left blank, nothing is templated into sameness — every card has the attributes that actually matter to a collector.
- **Intelligent arrangement**: products are automatically assigned into dynamically created Shopify collections based on set, rarity, type, and game mechanic. Collections are created on the fly when needed, so the catalog remains organized the way a collector would naturally browse it rather than dumped into a single flat list.
- **SEO-ready content**: product titles, handles, meta descriptions, and alt text are all generated with SEO in mind, so the store ranks for specific card searches without a separate manual optimization pass.

### Precision mechanics

- **Rate limit bypassing done right**: the Shopify REST API is aggressive about rate limits and will ban noisy callers. The script uses an **asynchronous** architecture with intelligent backoff, pacing every request to stay within Shopify's leaky-bucket quota. It runs fast when it can, slows down when it must, and never triggers a ban.
- **Monitoring and logs**: every action is tracked in real time with structured logs. Any failure — missing image, invalid field, transient API error — is recorded with its exact context, retried where appropriate, and surfaced for review where not. The operator always knows what succeeded and what needs attention.
- **Idempotency**: re-running the script never creates duplicates. Products are matched on a stable external ID, so reruns simply update existing records and create only what is genuinely new.

### The strategic impact

A catalog of 10,000 products is, for most stores, a multi-month project involving data entry teams, quality checks, and dedicated project management. With this pipeline, the same catalog ships in hours, with better consistency than a human team could realistically maintain. The same pattern applies to any e-commerce business with a structured external data source — collectibles, auto parts, book catalogs, wine references, and countless other verticals where the product truth lives in a database rather than in a content team's head.

### The delivered outcome

- A Shopify store with 10,000+ structured, well-categorized trading card products.
- A reusable Python automation layer that can be re-executed whenever TCGdex adds new sets.
- A clear separation between catalog data (in TCGdex) and commerce presentation (in Shopify) — each system doing exactly what it does best.

### Technology stack

- **Python** for the automation script.
- **Shopify REST API** for catalog creation and updates.
- **TCGdex API** as the upstream data source.
- **Asynchronous programming** for rate-limit-aware throughput.
- **Structured logging and monitoring** for operator visibility.

---

**Ready to automate your business?** [Discover our AI Automation service →](/expert-automatisation-ia)

10,000+ TCGdex cards imported to Shopify automatically: exhaustive product sheets, dynamic collections, rate limit bypassing. Async Python.

Import 10,000 TCGdex Cards to Shopify Auto

TCGdex E-commerce Automation

A surgical B2B lead machine for German-speaking Switzerland: multi-source scraping, hidden email extraction via regex, Hunter.io validation, and zero-duplicate output.

## An automated B2B lead generation machine for the German-speaking Swiss craftsmanship sector — surgical targeting, verified deliverability, zero duplicates

This project is an industrial-grade lead generation pipeline designed to systematically find, enrich, and verify B2B prospects in the craftsmanship sector (plumbers, electricians, roofers, carpenters, heating specialists) across German-speaking Switzerland. Where a manual researcher might assemble a few hundred prospects in a week with questionable deliverability, this pipeline produces thousands of **verified, deduplicated, deliverable contacts** with minimal human intervention. For any outbound sales team targeting this market, it is the difference between guessing and operating at scale.

### Intelligent collection and enrichment

- **Multi-source orchestration**: raw business data is collected in bulk from both Google Places APIs and the dominant Swiss directory (Local.ch). Using multiple sources catches businesses that would be missed by a single-source approach — many Swiss artisans are listed in one directory but not the other.
- **Deep enrichment**: once a business is identified, the pipeline autonomously navigates its website to extract context — services offered, geographical coverage, team size hints, legal name variations — enriching the raw record into a useful commercial profile.
- **Email hunting via regex**: hidden email addresses, often buried in contact pages or PDF documents, are extracted via carefully tuned regular expressions. This is critical for the Swiss market where many artisans do not expose their email in structured directory fields but only on their own website.
- **Real-time verification with Hunter.io**: every extracted email is passed through the Hunter.io API to certify its deliverability before it enters the final list. Invalid, role-based, and catch-all addresses are filtered or flagged. The result is a drastically lower bounce rate when the Sales team actually starts sending.

### Reliability and clean data

- **Python + Pandas**: the pipeline is developed in Python with the analytical power of Pandas for data manipulation, deduplication, and quality checks.
- **Algorithmic deduplication**: a layered deduplication system ensures that no artisan ever appears twice in the output — matching on legal name, trading name, phone number, address, and domain. No plumber will ever receive the same outreach email twice through a variant spelling.
- **Anti-crash architecture with checkpoints**: scraping jobs of this size routinely encounter network errors, API rate limits, and occasional server outages. The checkpoint system ensures the pipeline resumes exactly where it left off, with zero duplicate work and zero lost progress.
- **Quality reporting**: each run produces an audit report describing how many businesses were discovered, how many enriched successfully, how many emails were verified, and the exact reasons for any rejection — so the Sales team always knows what they are buying.

### Why this pipeline matters commercially

A Swiss-German-speaking craftsmanship outreach campaign lives or dies on list quality. The difference between a 40% bounce rate and a 3% bounce rate is the difference between burning your domain's sender reputation and running a sustainable outbound motion. This pipeline delivers that 3% bounce rate by design, because verification is done up front, not discovered after the fact in a mailbox full of NDRs.

### The delivered outcome

- Thousands of verified, enriched, deduplicated Swiss artisan prospects ready to feed an outbound sales motion.
- A reusable, audit-trailed framework that can be retargeted to other Swiss industries with minimal changes.
- A clear ROI for the Sales team: every hour of their time is spent on genuine prospects, not on chasing bad data.

### Technology stack

- **Python** for the pipeline logic.
- **Pandas** for data processing and deduplication.
- **Google Places API** for primary business discovery.
- **Local.ch** as a secondary Swiss directory source.
- **Hunter.io API** for email verification and deliverability checks.
- **Regular expressions (Regex)** for email extraction from unstructured web pages.

---

**Need targeted B2B data or a catalog migration?** [Discover our Web Scraping & Data service →](/expert-scraping-data)

Python scraping pipeline for Swiss artisan B2B leads: Google Places + Local.ch, regex email extraction, Hunter.io validation, Pandas deduplication.

Swiss B2B Leads Artisans — Python Pipeline

Data Scraping Pipeline "Adrien_78"

ODŌRA reinvents niche perfume discovery with AI recommendations, 3D bottle visualization, and AR — a unified web app, mobile app, and brand partner backoffice.

## ODŌRA — reinventing the discovery and shopping experience for niche perfumery, powered by AI, 3D, and AR

ODŌRA (odora.app) is an innovative platform that revolutionizes how people discover, evaluate, and buy niche perfumes. Traditional perfume e-commerce is fundamentally broken: buyers cannot smell the product before purchase, product pages are visually flat, and the discovery process is dominated by loud commercial brands rather than the rare, crafted houses that define modern perfumery. ODŌRA tackles each of those problems with a different piece of technology and unifies them into a single luxury-grade experience across web, mobile, and a backoffice for partner brands.

### AI-powered olfactory recommendation

At the heart of the platform, a dedicated **artificial intelligence** recommendation algorithm learns from every user's preferences, browsing history, purchase history, and explicit feedback to suggest the perfect perfume for the moment. What is normally a deeply subjective, often intimidating choice becomes a data-supported, confidence-building decision. The algorithm handles the full complexity of olfactory families, notes, accords, and personal affinities — producing recommendations that feel personal rather than generic.

### Absolute immersion with 3D modeling and augmented reality

The user experience takes a leap forward with the integration of **3D visualization** and **augmented reality**. Visitors can view each bottle from every angle, directly in their browser, with photorealistic modeling. On mobile, AR lets users virtually place the bottle on their own vanity or desk to see how it looks at home, at real scale. This is the kind of immersion that luxury perfume boutiques offer in person — translated, for the first time, into a browser and a phone screen.

### A unified multi-channel ecosystem

The ODŌRA project is not a single website. It is a complete ecosystem of three tightly integrated products:

- An elegant, responsive **web application** where customers discover, browse, and purchase perfumes, supported by the AI recommendation layer and the 3D experience.
- A native, high-performance **mobile application** that extends the same experience — plus AR capabilities — to iOS and Android devices.
- An **administrator backoffice** that lets the ODŌRA team manage the catalog, onboard new niche perfume houses, negotiate direct partnerships, and track the commercial performance of every brand, every bottle, every campaign.

### Why this matters for niche perfumery

Niche perfumery has always struggled online because its products are sensory in a way that the web cannot easily express. Flat product photos and lengthy paragraphs of "top notes, heart notes, base notes" do not capture what actually matters: how it looks, how it feels, whether it fits your identity. ODŌRA is the first serious attempt we have built to translate that sensory richness into a digital medium, using AI to guide discovery, 3D to convey craftsmanship, and AR to bring the bottle into the user's real space.

### What this project demonstrates

ODŌRA showcases the ability to combine emerging technologies — AI recommendation, real-time 3D, augmented reality — into a single coherent luxury experience, without drifting into gimmick territory. Every piece of tech has a specific job that serves the user. Nothing is there to impress a jury; everything is there because it makes niche perfumery easier, more confident, and more enjoyable to discover online.

### Technology stack

- **Artificial intelligence** recommendation engine for olfactory matching.
- **3D modeling** with real-time browser rendering.
- **Augmented reality** on mobile devices.
- **Next.js** for the web application.
- Native **mobile development** for high-performance iOS and Android experiences.
- **User experience design** focused on luxury polish and sensory fidelity.

---

**Ready to build your premium website?** [Discover our Premium Web Development service →](/expert-site-web-premium)

ODŌRA: niche perfume platform with AI recommendations, 3D visualization and augmented reality. Full ecosystem: web, mobile + brand partner backoffice.

ODŌRA: Niche Perfumes + AI + 3D + AR

ODŌRA

A 360° digital transformation for a savings association: showcase site, private member portal with financial dashboard, and a full admin backoffice — built on Supabase + Next.js 14.

## A complete 360° digital transformation for a solidarity savings association — modernizing tontines with enterprise-grade architecture

This project is the full digital transformation of a financial solidarity association whose mission is to help its members save and finance meaningful projects through shared tontines. For years, the association had been managing everything — membership, savings cycles, beneficiary rotations, training content — through spreadsheets, informal group chats, and a patchwork of manual workflows. The goal of this rebuild was to take that entire operation and turn it into a modern, secure, scalable digital ecosystem fit for the association's ambitions.

### A 360° digital ecosystem

Three distinct but tightly integrated products were built, each serving a specific audience.

- **Dynamic showcase site**: a modern public presentation of the association, its values, its impact, and a curated showcase of successfully funded projects. This is the front door — where prospective members learn about the association, read testimonials, and understand how the tontine model works.
- **Collaborative member space**: a private portal accessible only after authentication, where members consult their personal financial dashboard (contributions paid, upcoming tontine turns, total saved, next beneficiary cycle), browse the community directory, and follow exclusive financial education modules. The portal turns membership from a passive relationship into an active, continuous engagement.
- **Command backoffice**: an exhaustive administration interface for the association's managers, letting them pilot every flow, manage roles, onboard and offboard members, supervise savings cycles, and generate financial reports without ever touching a database directly.

### High-flying cloud architecture

The platform was built with modern web standards to ensure scalability, reliability, and long-term maintainability.

- **Next.js 14 (App Router)** and **React** for instant navigation, server-side rendering where it counts, and a clean, modern component model.
- **Supabase** as the serverless backend infrastructure, handling authentication, storage, and real-time features, on top of a robust **PostgreSQL** database for relational integrity of financial data.
- **Prisma ORM** for type-safe data access and structured schema migrations.
- **TypeScript** end to end for a codebase that stays maintainable as the association's team evolves.
- Sophisticated **role-based access control (RBAC)** ensuring the watertightness of financial data — members see only their own data, managers see only what their role requires, and privileged operations are logged and auditable.

### Why enterprise architecture for a solidarity association

There is a persistent myth that associations and non-profits should "make do" with informal tools because their budgets are smaller. The opposite is true. Associations that handle real money on behalf of their members owe them **more** rigor, not less — the members are not paying customers who can easily switch providers, they are stakeholders whose savings matter. Bringing enterprise-grade architecture to this association is a way of respecting its members and its mission.

### The delivered outcome

- A modern, coherent digital presence that matches the association's ambitions and values.
- A private, secure member space that turns membership into an everyday, tangible experience.
- A powerful backoffice that frees managers from spreadsheet chaos and enables them to focus on community and strategy.
- A scalable architecture ready for new member growth and new product lines (additional savings tracks, new training modules, new member regions).

### Technology stack

- **Next.js 14** with App Router and **React**.
- **Tailwind CSS** for the design system.
- **Supabase** for authentication, storage, and real-time backend features.
- **PostgreSQL** for the relational data layer.
- **Prisma ORM** for type-safe data access.
- **TypeScript** across the whole codebase.

---

**Ready to build your premium website?** [Discover our Premium Web Development service →](/expert-site-web-premium)

Digital solidarity tontine: showcase site, private member portal with financial dashboard, RBAC backoffice. Next.js 14 + Supabase + PostgreSQL.

Solidarity Tontine: Next.js 14 + Supabase RBAC

Solidarity Savings Tontine

A bank-grade photo album SaaS where creators design layouts and clients drag-drop their memories — auto-generated QR codes, zero-trust auth, and native n8n webhook support.

## A professional photo album SaaS built for creators and protected with bank-grade security

This professional platform delivers a turnkey solution for creating and managing memorable photo albums at scale. It is designed for two distinct user groups that have to coexist seamlessly: creators who design layouts and sell their work, and end customers who want the simplest possible experience to fill those layouts with their own memories. The product bridges a creator economy use case with SaaS-grade infrastructure, including security discipline normally reserved for banking apps.

### Intuitive user experience

- **Custom design**: creators build layout templates as structured PDFs, defining where photos go, how many are expected, and what the aesthetic will be.
- **Ultra-simplified customer flow**: end customers then drag and drop their best memories into the template using a web interface engineered to be painlessly easy — even for users who are not comfortable with technology.
- **Instant generation**: once the album is finalized, the platform automatically generates premium QR codes that redirect to a digital version of the album, perfect for gifting, sharing with family, or integrating into physical prints.

### Bank-grade security (Zero Trust)

User trust is paramount when clients entrust you with private family memories. The infrastructure is armored at every layer with a Zero Trust posture.

- **Strong authentication** using NextAuth combined with bcrypt hashing at 12 rounds — well above the minimum bar for modern security.
- **Strict access filter**: every client lives in a fully isolated environment. It is architecturally impossible for a user to access another user's albums, even with a crafted URL — isolation is enforced on every query, not just on the client.
- **API protection**: severe rate limiting (100 requests per 15 minutes per client) prevents brute force and abuse, and strict CORS barriers block any unauthorized external origin from reaching the API.
- **Audit trail**: every access and every action is logged, so any anomaly can be investigated.

### Ready for automation

The platform is fully "n8n ready". It exposes webhooks and secure API keys that make it trivial to plug into any complex corporate workflow — trigger email sequences when an album is finalized, push QR codes into print-on-demand services, sync orders into accounting tools, whatever the business requires. The application is designed to be part of a larger system rather than a standalone island.

### Why this combination matters

A photo album SaaS that is beautiful but insecure is a lawsuit waiting to happen. A photo album SaaS that is secure but clunky loses every customer in the first minute. Delivering both — fluid UX for creators and customers, impenetrable backend infrastructure — is the hard part. This project is the proof that the two are not mutually exclusive; they reinforce each other when the architecture is done right.

### Technology stack

- **Next.js** for the full-stack application.
- **NextAuth** for authentication, with bcrypt hashing for credential storage.
- **REST API** endpoints hardened with rate limiting and strict CORS rules.
- **PDF generation** for layout templates and final album exports.
- **Postman** for the API testing suite.
- **n8n** integration through webhooks for automation-ready workflows.

---

**Ready to build your premium website?** [Discover our Premium Web Development service →](/expert-site-web-premium)

Photo album SaaS: PDF templates, premium QR Codes, Zero Trust security (NextAuth + bcrypt), client isolation, rate limiting, n8n-ready.

Pro Photo Album SaaS — Zero Trust Security

Professional Photo Album Creation Platform

N-Bed drops a fully isolated, n8n-powered chatbot onto any website in minutes — Shadow DOM, 100% configurable, CDN-ready, works on WordPress, React, Vue, or plain PHP.

## N-Bed — the drop-in chatbot widget that connects any website to an n8n workflow in minutes

Adding a conversational assistant to an existing website is one of those projects that sounds trivial and turns out to be surprisingly painful. CSS conflicts break the layout. Dependency clashes break the site's JavaScript. The CMS does not play well with modern React. The result is a half-finished chatbot that everyone hates. **N-Bed** was engineered specifically to end that story. It is a fully configurable, fully isolated chatbot widget that plugs into any website — WordPress, Symfony, Next.js, Vue, plain PHP, static HTML — and connects directly to an n8n workflow as the AI backend.

The widget is production-ready, fully documented, and accessible at [https://n-bed.bovo-digital.tech/](https://n-bed.bovo-digital.tech/).

### The main features

- **Complete CSS isolation via Shadow DOM**: the widget renders inside its own Shadow DOM, so your site's CSS can never leak in and the widget's CSS can never leak out. No more "why did my header break after I installed the chat?".
- **Standalone bundle**: React is included in the widget's own sandbox, so there is no requirement on the host site's stack and no conflict with other React versions running elsewhere.
- **100% configurable**: every piece of the user-facing experience — texts, colors, icons, emojis, animations, positioning — is controllable from the outside.
- **Compatible with every stack**: works as a CDN drop-in on any website, regardless of how it was built.
- **Fully responsive**: desktop, tablet, and mobile, all tuned and tested.
- **Customizable animations**: bounce, zigzag, or none — choose what fits the site's tone.

### Three integration modes

N-Bed supports three different integration modes so it fits into any team's workflow.

**Mode 1 — data-* attributes (Simple)**
Ideal for quick integrations and CMS-heavy stacks (WordPress, Drupal, Joomla). Just drop a script tag and some HTML attributes, and the widget configures itself. No JavaScript expertise required.

**Mode 2 — JavaScript object (Advanced)**
Full programmatic configuration via a JavaScript object exposed on window. This mode unlocks every option and is ideal for developer-led integrations or dynamic configurations driven by the host application.

**Mode 3 — Mix (Recommended)**
The hybrid approach: data-* attributes for stable base values, JavaScript object for dynamic or advanced overrides. This is the sweet spot between simplicity and full control.

### Advanced configuration surface

The widget exposes customization across many dimensions.

- **Texts**: title, welcome messages, input placeholders, button labels.
- **Colors**: full palette control — primary, secondary, background, text — for brand consistency.
- **Positioning**: four possible positions (bottom-right, bottom-left, top-right, top-left) to fit any layout.
- **Animations**: launcher animation, open and close animations, all fine-tunable.
- **Behaviors**: auto-open on load, conversation persistence across page navigations, automatic scroll to latest message.
- **Features**: reactions on messages, Markdown rendering, file attachments, timestamps.
- **Custom CSS**: injection of arbitrary CSS rules scoped to the widget's Shadow DOM, for teams that want complete visual control.

### API and n8n integration

The widget talks to n8n via webhook, with support for every common streaming pattern so the conversational UX feels instant.

- Simple JSON responses for baseline use cases.
- **SSE (Server-Sent Events)** streaming, for typewriter-style responses.
- **NDJSON** streaming, for structured progressive payloads.
- **Plain Text** streaming, for maximum compatibility with any backend.
- File attachment handling via base64, so users can send documents and images straight into the workflow.

### Complete documentation

The project ships with a full documentation site covering every aspect of installation and customization.

- Installation guide for both CDN and local hosting scenarios.
- Usage examples for different stacks (WordPress, Next.js, Symfony, vanilla HTML).
- Complete API reference for every configuration option.
- Advanced customization recipes for common visual and behavioral tweaks.
- Troubleshooting checklist and support contact paths.

### Why a widget like N-Bed matters

For any agency or internal team building conversational AI products, a reusable, stack-agnostic widget is a force multiplier. Instead of re-implementing the front-end layer for every new client or project, the team ships N-Bed and focuses on the actual intelligence — the n8n workflow, the prompts, the integrations. The front end becomes a solved problem.

### Technology stack

- **React** and **TypeScript** for the widget internals.
- **Shadow DOM** for full style isolation.
- **n8n** as the webhook-based AI backend.
- **CDN-first distribution** for frictionless installation.
- **JavaScript** public API surface for programmatic control.

---

**Ready to automate your business?** [Discover our AI Automation service →](/expert-automatisation-ia)

N-Bed: isolated React chatbot widget (Shadow DOM), 100% configurable, connected to n8n via webhook. Compatible WordPress, PHP, Vue, Symfony. Plug & play.

N-Bed: n8n Chatbot Widget — Works Everywhere

N-Bed - Integrable Chatbot Widget for n8n

Aura Academy: a full-stack Next.js 15 e-learning platform with gamification, EditorJS content, NextAuth, Schema.org SEO, and dynamic sitemap — deployed on Vercel, built to scale.

## Aura Academy — a complete, production-grade e-learning platform built on Next.js 15, TypeScript, and rigorous SEO

E-learning platforms have become a commodity — except when they are not. The difference between a generic LMS and a platform that actually grows is in the details: how content is created, how learners are motivated, how the platform ranks in search engines, how the architecture scales under real load. **Aura Academy** is a modern, performant e-learning platform built end-to-end for exactly that bar of quality. This project covered the entire stack: technical architecture, product features, gamification, analytics, SEO, and deployment.

The solution is fully live and accessible at [https://aura-academy.bovo-digital.tech/](https://aura-academy.bovo-digital.tech/).

### The modern technical stack

- **Next.js 15 and React 19**: a modern framework with App Router, Server Components, and the latest performance optimizations. Every page is rendered server-first for fast first paint and strong SEO, with client-side interactivity layered on top where it matters.
- **TypeScript**: end-to-end static typing for maintainability and for reducing the class of bugs that plague loosely-typed e-learning platforms over time.
- **PostgreSQL and Prisma**: a robust relational database with a modern ORM. Course, lesson, module, user, progress, and achievement data all live in a properly designed schema, not in a pile of JSON blobs.
- **NextAuth.js**: a secure, complete authentication system supporting credential login, social providers, and session management that respects modern security practices out of the box.
- **Vercel**: deployment on a platform built for Next.js, with edge computing, global CDN, automatic previews per branch, and production-grade observability.

### The main product features

- **Advanced course management**: a rich **EditorJS** editor with **Markdown** support enables instructors to create structured, interactive educational content — text blocks, code snippets, images, embedded media, quizzes. No WYSIWYG chaos, no dead-end HTML editor.
- **Gamification system**: points, badges, streaks, and visible progression rewards turn passive learners into active ones. The gamification layer is backed by proper domain modeling, not bolted on — so it scales to new course types without rewrites.
- **Integrated analytics**: detailed tracking of learner progress, completion rates, time spent per module, and course-level performance. Instructors see what works and what does not, at a glance.
- **JSON import/export**: easy migration and backup of courses. A instructor can export their course as a structured JSON file, version it, re-import it, or hand it to a colleague. Content is not locked into the platform.
- **Interactive questions**: an integrated quiz and assessment system that supports multiple question types and produces graded reports.
- **Responsive interface**: modern, adaptive design that works across mobile, tablet, and desktop with a cohesive experience at every breakpoint.

### Complete SEO optimization

For an e-learning platform, SEO is not optional — organic traffic is often the single largest acquisition channel. Aura Academy was built from day one with SEO as a first-class concern.

- **Schema.org structured data**: rich JSON-LD markup for courses, educational occupational credentials, organization, and breadcrumbs. This unlocks rich results in Google and significantly boosts click-through rates from the SERP.
- **Dynamic sitemap**: automatic sitemap generation covers every course and page, so new content is discovered and indexed fast.
- **Optimized robots.txt**: precise directives for every major search-engine crawler, with allowances configured for AI-focused crawlers too.
- **Open Graph and Twitter Cards**: full social metadata for premium link previews when courses are shared anywhere on the web.
- **Performance optimized**: fast initial load times, route-level code splitting, lazy loading, and image optimization that keeps Core Web Vitals in the green.

### The delivered result

Aura Academy is a production-ready, scalable, performant e-learning platform delivering a modern and engaging learning experience. The platform is live and serving real learners at [https://aura-academy.bovo-digital.tech/](https://aura-academy.bovo-digital.tech/), and demonstrates what a complete, engineered, SEO-aware LMS looks like in 2026.

### Technology stack

- **Next.js 15** and **React 19** for the full-stack app layer.
- **TypeScript** for type safety.
- **PostgreSQL** and **Prisma** for the data layer.
- **NextAuth.js** for authentication.
- **EditorJS** and **Markdown** for content creation.
- **Schema.org** for structured data.
- **Vercel** for deployment, preview environments, and edge caching.

---

**Ready to build your premium website?** [Discover our Premium Web Development service →](/expert-site-web-premium)

Aura Academy: e-learning platform Next.js 15 + PostgreSQL + Prisma — gamification, quizzes, Schema.org SEO, analytics, JSON import/export. Vercel.

Aura Academy: E-Learning Platform Next.js 15

Complete E-Learning Platform with Next.js 15, TypeScript & SEO Optimized

Tontineo's mission control: a web admin dashboard to manage users, tontines, transactions, and Stripe/Twilio settings — all from one secure, multilingual interface.

## Tontineo's mission control — a complete admin application to run the platform from one screen

Behind every serious consumer product is a powerful internal tool that most users never see. For Tontineo, that tool is the administration web application: a centralized dashboard that lets the operations team supervise, debug, configure, and grow the platform safely. Built as a separate, hardened interface from the user-facing mobile app, it gives administrators the leverage to manage thousands of tontines, tens of thousands of users, and millions in transaction volume — without losing visibility or control.

### The main features

- **Centralized dashboard**: a real-time overview of platform activity — active users, live tontines, transaction volumes, support tickets, key performance indicators — all visible at a glance on the first screen after login.
- **User management**: a complete interface to view, search, filter, and manage every user of the mobile app. Profile details, account history, KYC status, and audit trail for any action performed by or on the account.
- **Tontine management**: a supervision tool to track every active tontine — members, contribution schedules, current beneficiary, cumulative volume, status — in real time. The ops team can intervene when a group reports an issue, without ever touching the underlying data directly.
- **Transaction management**: detailed tracking of every financial transaction, with complete history, Stripe-side cross-reference, refund tooling, and verification workflows for compliance review.
- **Waitlist management**: a dedicated interface to curate the waiting list, approve access in batches, and give specific users early entry when appropriate.
- **Administrator management**: a role-based access control (RBAC) system to manage staff permissions, so each team member sees only what their role requires. Audit logs record every administrative action for accountability.
- **Service configuration**: a settings area to configure essential third-party integrations — Twilio credentials for SMS, Stripe keys for payments, support contact email, feature flags — without ever redeploying code.
- **Modern interface**: an elegant design with a dark navigation sidebar, clear content areas, and a UX tuned for long daily sessions of operational work.
- **Multilingual support**: the interface is available in both French and English, for international operational teams.

### Why a dedicated admin app matters

Relying on database consoles or direct SQL queries is how fintech platforms get compromised or make unrecoverable mistakes. A dedicated admin application enforces safe, auditable, role-specific workflows around every sensitive action. It is the difference between an amateur operation and a mature one.

### Technology stack

- **React** and **TypeScript** for the component architecture.
- **Web application** delivery with modern build tooling and deployment.
- **Dashboard patterns** with real-time data updates and filter-heavy interfaces.
- **Admin panel** architecture with RBAC, audit logs, and safe-action workflows.
- **API integration** with the Tontineo backend, Stripe, Twilio, and Firebase.

---

**Ready to build your premium website?** [Discover our Premium Web Development service →](/expert-site-web-premium)

Tontineo backoffice: user management, tontines, transactions, waitlist, admin permissions and Stripe/Twilio config. React/TypeScript dashboard.

Tontineo Admin: Full Management Dashboard

Tontineo - Administration Application

A premium landing page that positions Tontineo as the future of group savings — conversion-optimized sections, smooth animations, and direct store links to drive app downloads.

## Tontineo's digital storefront — a landing page engineered to turn visitors into downloads

A mobile app lives or dies on the landing page that introduces it. For Tontineo, the app that digitizes traditional tontine management, we built a premium web presence that is both a product showcase and a conversion machine. Every section earns its place: it exists because it moves a specific visitor closer to the App Store or Google Play button. Nothing decorative, nothing accidental.

### The main features

- **Premium design**: a modern, elegant interface with smooth animations, well-structured sections, and a responsive layout that feels polished on every device — from narrow phones to wide desktop monitors.
- **Clear product presentation**: dedicated sections explain the app's features, the benefits for different kinds of users (organizers vs participants), and real testimonials that anchor the claims in actual usage.
- **Conversion optimization**: strategically placed calls-to-action, a lead-capture form for prospects who want to hear when new features ship, and direct links to both **Google Play** and the **App Store** so the visitor can download the app in one tap.
- **Optimized SEO**: carefully crafted metadata, semantic HTML structure, clean heading hierarchy, and strong Core Web Vitals performance so the page ranks well for both branded and category queries ("digital tontine app", "tontine management application", and variants).
- **Smooth user experience**: intuitive navigation, subtle animations that reinforce rather than distract, and a funnel that naturally guides visitors from "what is this?" to "I want to download it".
- **Responsive design**: adaptive breakpoints tested on real devices, ensuring the experience never degrades regardless of screen size.

### Why landing pages matter for mobile apps

An app store listing is not enough. In 2026, prospective users almost always evaluate a mobile app on a landing page first — friends share URLs, ads drive to URLs, search engines rank URLs. The landing page is where the app wins or loses the visitor's trust. For a financial app like Tontineo, which handles real money flows between community members, the quality of the landing page directly impacts how quickly users feel comfortable downloading and signing up.

This page is designed to project exactly that trust: serious, polished, transparent, and user-centric.

### What this landing page delivers

- A digital showcase that accurately conveys Tontineo's value, features, and safety guarantees.
- A measurable conversion layer (CTAs, form, store links) that the marketing team can iterate on.
- A solid SEO foundation that keeps the app discoverable in search engines over time.
- A brand-consistent experience from the first pixel to the final page.

### Technology stack

- **Next.js** with modern React and TypeScript.
- **Tailwind CSS** for the design system and responsive breakpoints.
- **SEO** setup with full metadata, OG tags, and semantic HTML.
- **Responsive design** tested on real devices.

---

**Ready to build your premium website?** [Discover our Premium Web Development service →](/expert-site-web-premium)

Tontineo landing page: modern design, conversion-optimized sections, Google Play & App Store links, SEO metadata, responsive all devices.

Tontineo: Mobile Tontine App Landing Page

Tontineo - Premium Landing Page

Tontineo digitizes the ancestral tontine: a Flutter app with Stripe payments, Google/Apple Sign-In, GDPR compliance, and real-time tracking — now live on Google Play.

## Tontineo — digitizing the ancestral tontine, securely, transparently, for the mobile era

Tontineo is a mobile application that transforms how communities run their tontines. A tontine is a centuries-old rotating-savings mechanism where a group contributes regularly and each member receives the pooled amount in turn — a cornerstone of mutual finance in many regions of the world. Traditionally managed by a trusted human organizer with a notebook and a lot of good faith, tontines are ripe for a modern, digital, secure rebuild. That is exactly what Tontineo is. Developed with **Flutter** and available on the Google Play Store, the app lets users create, join, and manage tontine groups with advanced payment, communication, and tracking features that make the whole practice safer and more transparent for everyone involved.

### Tontine management

Creating a tontine on Tontineo is flexible and intuitive. The organizer configures the number of members, the contribution amount, and the frequency (weekly, monthly, or any custom cadence). Several drawing modes are available: a pre-agreed fixed order, a random draw that happens on schedule, or a member vote to decide the next beneficiary. Throughout the lifecycle of the tontine, contributions and beneficiaries are tracked in real time, so every participant always knows exactly where the group stands.

### Secure payments

**Stripe** integration handles every transaction with the same security standards a bank would demand. Automatic contribution payments are available for Premium users — "set it and forget it" tontine participation, without the friction of monthly manual transfers. Multiple payment methods are supported, and a detailed transaction history is always one tap away for every user.

### Group management

A **code-based invitation** system makes joining a tontine effortless and prevents strangers from sliding into private groups. An integrated chat keeps communication inside the app, so discussions about rules, schedule, or individual circumstances stay in one organized place. Premium users unlock a **digital vault** for group documents — rules, agreements, ID uploads — along with advanced statistics and exportable reports.

### Security and compliance

Security is not an afterthought here; it is a foundational pillar. **Firebase Authentication** secures user accounts. GDPR compliance is enforced with explicit consents, data export rights, and account deletion flows. **Age verification (18+)** is required at sign-up for legal reasons and for the integrity of the financial relationship inside each group. All legal documents — Terms of Service, Terms of Sale, Privacy Policy — are integrated into the app and kept up to date.

### Internationalization

Multilingual support in **French and English** from day one, with on-the-fly language switching and a fully localized interface. Dates, currencies, and micro-copy all adapt to the user's preference.

### Social authentication

Quick login via **Google Sign-In** and **Apple Sign-In**, automatic user profile creation, and a smooth experience across every authentication mode — the user never has to think about which account they used.

### Discovery mode

Access without an account is available for users who just want to explore the app before committing (this is also an Apple App Store requirement). Authentication is only enforced for advanced features, and intelligent authentication guards make the transition from guest to logged-in feel natural.

### Advanced monitoring

Production stability was a non-negotiable. **Firebase Crashlytics** captures crashes in real time with symbolicated stack traces, email alerts fire automatically for critical crashes, and advanced analytics feed the product team with detailed usage and reliability metrics. Every issue is surfaced fast; most are fixed before users even notice.

### Why Tontineo matters

Tontineo is more than a utility app: it is a modernized version of a financial practice that has supported families and communities for generations. By making tontines safer, transparent, and easier to manage, Tontineo empowers communities to keep this tradition alive in the age of mobile finance, without relying on fragile spreadsheets or risky informal setups.

The application is live on the Google Play Store and delivers a premium user experience for running tontines in a modern, secure, and transparent way.

### Technology stack

- **Flutter** and **Dart** for the cross-platform mobile codebase.
- **Firebase** for authentication, database, and real-time syncing.
- **Stripe** for secure payment processing.
- **Google Sign-In** and **Apple Sign-In** for social authentication.
- **Firebase Crashlytics** for production monitoring.

---

**Ready to launch your mobile app?** [Discover our Mobile App Development service →](/expert-application-mobile)

Tontineo: Flutter app for digital tontines — Stripe payments, integrated chat, random draws, GDPR compliance, Google Sign-In. On Google Play.

Tontineo: Digital Tontine Flutter App

Tontineo - Digital Tontines Mobile Application

A Next.js 16 dark-mode landing page for FocusLock with FR/EN routing, dynamic OG images, lead capture to n8n, and complete legal pages — production-ready and globally optimized.

## FocusLock's web presence — a premium Next.js 16 landing page built for conversion and SEO in two languages

A mobile app is only half of the product. The other half is the landing page that convinces a visitor to download it. For FocusLock — a professional-grade concentration operating system — we built a premium web experience on Next.js 16 (App Router) and React 19, with full internationalization, a production-grade SEO layer, and a visually consistent dark design that matches the app's identity from browser to home screen.

### The technology and architecture

- **Modern stack**: Next.js 16 with App Router, React 19, TypeScript, and Tailwind CSS 4 — the contemporary stack for performant, maintainable marketing sites that feel fast on every device.
- **Full internationalization**: complete FR/EN support via **next-intl**, with localized routing (/fr, /en), language detection, and clean URL structure for both SEO and shareable links.
- **Exclusive dark theme**: the brand identity requires dark-only design. We enforced this at the architecture level via next-themes with the light theme deliberately disabled, keeping every touchpoint on brand.
- **Optimized SEO**: dynamic Open Graph metadata, OG images generated on the fly through Next.js API routes, language-specific SEO, and canonical URL handling for each locale.

### The main features of the site

- **Premium landing page**: a modern dark design with smooth animations and sections engineered for conversion — hero, social proof, feature tour, testimonials, final CTA.
- **Lead capture form**: connected to **n8n** for automatic routing of new leads into the marketing funnel, without any manual handling.
- **Complete legal pages**: privacy policy, terms of use, cookies, legal notice, security, compliance, DPA, GDPR and CCPA rights, acceptable use policy. A mobile app that handles focus data and integrates deeply with the OS is held to a high legal bar, and the site reflects that.
- **Information pages**: service status page, contact, FAQ, support, downloads — each one indexed and optimized individually for discovery.
- **Custom error pages**: 404, 500, and other HTTP error codes, all designed in the same visual language so an error screen still feels like part of the product.
- **Responsive design**: every breakpoint is tested and polished, from narrow Android devices to wide desktop monitors.
- **Performance optimizations**: Server Components for static pre-rendering, automatic image optimization, lazy loading, and code-splitting, so every page meets modern Core Web Vitals thresholds.

### Technical architecture

- Modular structure with clear separation between landing, layout, and UI components.
- Server Components for static pre-rendering, with explicit translation loading for each locale — avoiding the common pitfall of shipping duplicate translation bundles to the client.
- Client Components for interactive islands, using the useTranslations() hook from next-intl to access localized strings cleanly.
- Vercel configuration tuned for optimized deployment: preview environments per branch, edge caching, analytics integration, and automatic image optimization at the CDN layer.
- Environment variables for every integration (n8n webhook URLs, download URLs per platform, support endpoints, concierge contact), so no secret ever leaks into the repository.

### Advanced internationalization

- Complete, linguist-reviewed translations across every page and section.
- Language-specific SEO with metadata adapted per locale — not just mechanically translated, but rewritten to match search intent in each language.
- Dynamic Open Graph images generated via API routes, producing visually rich link previews in both French and English on social platforms.
- Transparent localized routing that preserves user intent across language switches — switch from /fr to /en on the pricing page and you land on the pricing page, not the home page.

### What this project demonstrates

This landing page is a showcase of what a modern web presence should look like in 2026: a Next.js 16 App Router architecture, React 19 server components, a strict internationalization strategy, real technical SEO (not just meta tags), and a premium visual language that is consistent from the app icon to the legal footer.

### Technology stack

- **Next.js 16** (App Router) and **React 19**.
- **TypeScript** for end-to-end type safety.
- **Tailwind CSS 4** for the design system.
- **next-intl** for internationalization and localized routing.
- **next-themes** for enforced dark theme.
- **Server Components** for performance and SEO.
- **Vercel** for deployment and edge caching.
- **n8n** for the lead-capture pipeline.

---

**Ready to build your premium website?** [Discover our Premium Web Development service →](/expert-site-web-premium)

FocusLock landing page: Next.js 16 + React 19, FR/EN internationalization (next-intl), dark design, Open Graph SEO, lead capture form connected to n8n.

FocusLock: Next.js 16 Landing Page FR/EN

FocusLock — Premium Landing Page

FocusLock is a premium Flutter concentration app with app blocking, smart notifications, proximity detection, and a full analytics dashboard — built to spec, top-tier quality.

## FocusLock — a premium Flutter 3.x concentration app built to spec, shipped on every platform

FocusLock is a premium mobile concentration-management application developed with Flutter 3.x and delivered in full compliance with a detailed client specification. The app delivers a complete, polished user experience from the first launch to the tenth session, following a clean flow: onboarding, session configuration, active session, statistics, settings. Every screen was designed with the attention to detail you would expect from a paid App Store product, not a side project.

### The user journey, in five acts

**Onboarding**
A three-screen introduction welcomes new users to the app's philosophy and features. All resources are embedded for a fully offline experience — the app never waits on a network request to show its first screen. A premium LaunchScreen sets the tone from the very first tap on the icon.

**Session configuration**
A rich configuration layer lets users tailor every focus session: presets for common durations, fully customizable session length, selective app blocking across the user's Android apps, notification silencing, proximity detection (so moving the phone closer pauses or alerts), adjustable sensitivity for the proximity sensor, alarm control, and volume control. No detail was compromised.

**Active session**
During a focus session, FocusLock becomes a calm, focused space: an animated timer, localized motivational messages that adapt to the user's language, a blocking overlay that intercepts blocked-app attempts, automatic notification muting on Android via the system NotificationManager, and haptic feedback tied to proximity events. Every interaction is designed to be unobtrusive — the goal is deep work, not app engagement.

**Statistics**
A full analytics dashboard with day / week / month filters, 7-day graphs, dynamic goal indicators, session breakdowns, a detailed session history, and a badge system that rewards streaks and milestones without becoming gimmicky.

**Settings**
Advanced configuration at the user's fingertips: language management (FR/EN with persistence of the user's choice), light and dark theme support, configurable time units, app blocking via the native Android selector, default emergency-stop behavior, granular permission management, and full data export and management — because respecting the user's data is a feature, not a checkbox.

### Technical architecture

- Modular architecture with clear feature separation: onboarding, home, session_config, session_active, statistics, settings, permissions, premium.
- A complete Design System with color tokens (indigo/violet theme), Inter typography, spacing scale, radii, and shadow tokens — delivering visual consistency across every screen.
- Persistence via SharedPreferences for settings and session history, so no external dependency is required for the app to function fully offline.
- Native Android integrations: app blocking via Usage Stats and Overlay permissions, notification muting via NotificationManager, proximity detection with HapticFeedback for tactile cues.
- Premium branding: embedded logo, native Android 12 splash screen, custom icons tuned per platform for Android, iOS, and web.
- Complete FR/EN localization with full persistence of the user's language preference, including right-to-left readiness for future language additions.

### Quality and documentation

- Complete functional documentation, including detailed user-flow specifications and a full QA checklist used during delivery.
- Structured, maintainable code organized for future contributors.
- Multi-platform support — the Flutter codebase runs cleanly on Android, iOS, Web, macOS, Linux, and Windows, with platform-specific adjustments where needed.

### What this project demonstrates

FocusLock showcases complete mastery of Flutter app development, deep Android native integrations (Usage Stats, NotificationManager, Overlay permissions), disciplined state management, structured data persistence, and premium mobile app design. It is the kind of reference project that proves an application can be both technically rigorous and emotionally polished.

### Technology stack

- **Flutter 3.x** and **Dart** as the core framework.
- **Material 3** for the design foundations.
- **Android Native** APIs for the system-level features (Usage Stats, NotificationManager, Overlay).
- **SharedPreferences** for local persistence.
- **Localization (i18n)** for complete FR/EN support.
- **Custom Design System** for visual consistency.

---

**Ready to launch your mobile app?** [Discover our Mobile App Development service →](/expert-application-mobile)

FocusLock: Flutter 3.x app — Android app blocking, notification silence, animated timer, day/week/month stats, badges. iOS & Android.

FocusLock: Premium Flutter Focus App

FocusLock — Flutter Mobile Application

JWT auth, IP verification, and per-user credit management in one n8n workflow — the security backbone that keeps the Echo AI platform abuse-free and scalable.

## The invisible backbone that keeps a generative-AI SaaS from being drained by abuse

Generative AI applications have a quietly brutal economic problem: every request costs real money, and without the right infrastructure a handful of malicious or careless users can burn through a month's budget in hours. For **Echo AI** — an innovative product that turns a product photo and an inspiration image into a professional creative brief for advertising, merging style and product into impactful visuals — this risk was existential. My role was to build the security and access-control layer that makes the whole thing sustainable.

### The project context

Echo AI's user experience is effortless by design: upload a product image, upload an inspiration image, get a polished creative brief ready to drive paid-ad production. Behind that simplicity sits a sophisticated pipeline of AI calls, each of them consuming compute credits. Opening the door to that kind of backend without proper guardrails would have invited abuse within days.

### The challenge

The objective was to implement a sophisticated access-control system that would:

- Prevent abusive and unauthorized use of the application.
- Enforce credit quotas per user so that free-tier, trial, and paid users each get exactly what they were allocated — no more, no less.
- Absorb traffic spikes without breaking or manually scaling.
- Remain modular so the product team could evolve policies later without rearchitecting anything.

Without such a system, the Echo AI infrastructure would have been overwhelmed by malicious or excessive requests, leading to unpredictable costs and degraded experience for legitimate users.

### The implemented solution

I designed and deployed a powerful automation workflow built on **n8n**, acting as the intelligent gatekeeper in front of every sensitive endpoint. The workflow integrates several mechanisms, each of them handling a specific class of threat.

**Credit management**
Precise per-user tracking of credit consumption. Every generation request is checked against the user's remaining quota before any AI call is made. When the quota is exhausted, the user is informed cleanly and pointed to upgrade paths, rather than being silently cut off.

**IP address verification**
IP-based controls identify and rate-limit suspicious or excessive traffic sources. Known abuse patterns are blocked upstream, before they ever reach the expensive AI layer. Geographic anomalies, sudden bursts, and VPN/proxy misuse all have their own detection rules.

**JWT authentication**
Every request is authenticated via JSON Web Tokens. Only properly signed, non-expired, non-revoked tokens from authenticated users are processed. The token layer also carries the user's plan and quota information, so downstream nodes do not have to re-query the database on every request.

**Robust data storage**
Integration with Google Sheets for dynamic, easily accessible management of user data and security configurations. This choice is intentional: non-technical operators on the Echo AI team can adjust thresholds, whitelists, and quota policies directly, without touching code or redeploying. A lightweight admin UX, essentially, built on a tool the team already masters.

**Webhooks and integrations**
The workflow leverages webhooks for low-latency communication between the application and the security layer. Real-time responsiveness is critical — an access check cannot take 2 seconds, or the UX of Echo AI becomes unusable. The architecture is engineered to keep checks under a tight latency budget.

### The impact and results

- **Abuse neutralized**: significant reduction in suspicious and malicious request volume.
- **Predictable costs**: per-user quotas enforce the company's unit economics. No more open-ended bills.
- **Fair user experience**: legitimate users are never throttled unfairly, because the rules are deterministic and transparent.
- **Operational flexibility**: thanks to the modular n8n architecture and the Google Sheets configuration layer, the product team can evolve access policies in minutes, not weeks.
- **Scale-ready foundation**: as Echo AI grows, this layer grows with it, because every rule is parameterized rather than hard-coded.

### Why this kind of infrastructure is worth building properly

For any generative-AI SaaS, access control is not an optional feature — it is the spine. Build it poorly, and the business becomes financially uncontrollable. Build it properly, and every other feature can scale without fear.

### Technology stack

- **n8n** for the full workflow orchestration and gatekeeper logic.
- **JWT (JSON Web Tokens)** for stateless, secure authentication.
- **Google Sheets** as the dynamic configuration and user-data layer.
- **Webhooks** for low-latency synchronous checks.
- **IP verification rules** for upstream abuse prevention.

---

**Ready to automate your business?** [Discover our AI Automation service →](/expert-automatisation-ia)

n8n system for Echo AI: user credit management, IP verification, JWT authentication. Secure infrastructure protecting your AI SaaS from abuse.

SaaS AI Access Control — n8n JWT Credits

Access Management and Abuse Prevention for Generative AI (Echo AI)

Personalized WhatsApp campaigns at scale — an AI agent that reads your contact sheet, generates a unique Gemini-crafted message per recipient, and tracks every delivery.

# AI Agent for Automated WhatsApp Campaigns

## Context
In the current digital communication landscape, reaching customers in a personalized way and at scale is a major challenge. Manual WhatsApp campaigns are not only time-consuming but also lack the necessary personalization to effectively engage the audience. This project aimed to overcome these limitations by designing an intelligent automation solution.

## Challenges
The main challenges included:
*   **Scalability:** Sending thousands of messages without manual effort.
*   **Personalization:** Generating unique and relevant messages for each contact, beyond simple templates.
*   **Reliability:** Ensuring proper message delivery and managing failures.
*   **Tracking:** Obtaining a clear report of campaign performance.
*   **Integration:** Linking various platforms (spreadsheets, messaging APIs, AI, notification systems).

## The Solution: An Intelligent Workflow
I developed a robust AI agent, orchestrated via a detailed workflow, to manage the entire lifecycle of WhatsApp campaigns. Here are the key steps:

1.  **Workflow Trigger:** The process starts on demand or at a scheduled time.
2.  **Data Acquisition:** A list of contacts is read and imported from a Google Sheet, serving as a dynamic database for the campaign.
3.  **Loop Processing:** For each contact, a series of actions is executed:
    *   **Cleaning and Formatting:** The phone number is standardized to ensure compatibility with the WhatsApp API.
    *   **Personalized Message Generation:** The Gemini API (Google's Artificial Intelligence) is used to create a unique and contextually relevant message for each recipient, thus increasing engagement.
    *   **Sending via API:** The generated message is sent to the contact via the Evolution API, a reliable solution for WhatsApp communications.
4.  **Result Management:**
    *   **Real-Time Update:** The sending status (success or failure) is immediately recorded in the Google Sheet.
    *   **Error Handling:** In case of failure, errors are captured and documented for later analysis, allowing for continuous optimization.
5.  **Final Report and Notifications:**
    *   **Result Compilation:** Once the campaign is finished, an aggregated report is compiled, summarizing the overall performance.
    *   **Automated Alerts:** This report is then automatically sent via Telegram and email to stakeholders, ensuring maximum transparency and responsiveness.

## Technologies Used
*   **Google Sheets:** Contact management and status tracking.
*   **Google Gemini API:** AI-powered personalized content generation.
*   **Evolution API:** WhatsApp message sending interface.
*   **Telegram API:** Campaign completion notifications.
*   **Email Sending Service:** Additional email notifications.
*   **Workflow Automation Tools:** Process orchestration and execution.
*   **Regex / String Processing Functions:** For number cleaning and formatting.

## Impact and Benefits
This project transformed the management of WhatsApp campaigns, offering:
*   **Massive Time Savings:** Drastic reduction in manual effort.
*   **Increased Engagement:** Thanks to advanced personalization by AI.
*   **Increased Reliability:** With accurate tracking and proactive error management.
*   **Data-Driven Decisions:** Thanks to detailed reports.
*   **Scalability:** Ability to manage unlimited message volumes.

This AI agent represents a complete solution for businesses wishing to optimize their customer communication on WhatsApp, combining the power of AI with the efficiency of automation.

---
**Ready to automate your business?** → [Discover our AI Automation service](/expert-automatisation-ia)

AI agent sending thousands of personalized WhatsApp messages via Gemini. Google Sheets tracking, Telegram report, error handling. Scalable and reliable.

Automated Personalized WhatsApp Campaigns — Gemini

Intelligent AI Agent for Automated and Personalized WhatsApp Campaigns

Régina never sleeps: this WhatsApp AI assistant qualifies leads, books appointments, sends quotes, and manages your full sales pipeline 24/7 — completely on autopilot.

## Régina on WhatsApp — an AI assistant that runs your sales pipeline at 3 a.m.

At 3 a.m., while the competition's inbox is silent and a potential customer is scrolling in bed, our WhatsApp AI assistant can qualify the lead, book the meeting, and even send a personalized quote. That is the new baseline for serious businesses in 2026: a conversational AI that runs the full customer relationship cycle 24/7, not a landing page that says "we will get back to you within 24 hours".

After the success of Régina on our own website, we adapted her for WhatsApp — a direct, instant, global communication channel that every prospect already uses. The result is a tireless AI agent that grows the business while the human team focuses on strategy and closing.

### The key capabilities of the WhatsApp assistant

**Complete 24/7 customer relationship management**

- **Automatic lead qualification**: the AI asks the right calibrated questions to identify the most relevant prospects — budget range, urgency, decision authority, scope — without sounding like a survey.
- **Intelligent appointment booking**: connected to team calendars, the assistant proposes real-time slots and books them instantly. Timezone handling, buffer times, working-hours rules are all respected automatically.
- **Instant FAQ answers**: pricing, timelines, services, payment terms, onboarding process — every recurring question is handled immediately, with consistent messaging that is never off-brand and never impatient.
- **Personalized quote generation**: for simple scopes, the assistant can produce a structured quote in the chat itself, already branded and ready to sign.

**Support for every communication format that matters on WhatsApp**

- Handles text messages, images, voice notes, audio files, and document attachments.
- Voice messages are transcribed and understood, so a prospect who prefers talking over typing is not forced to change their habit.
- The conversation history is maintained per contact for personalized, context-aware follow-up. Régina never asks the same question twice, and picks up conversations exactly where they were left off, days or weeks later.

### The project's real impact

- **Continuous conversion**: prospects become clients even when the team is offline — nights, weekends, holidays.
- **Higher customer satisfaction**: clients get instant, accurate answers to their questions, which is increasingly what they now expect from serious providers.
- **Genuine business automation**: the company runs on autopilot at the top of the funnel, freeing human attention for the strategic, higher-leverage tasks that actually move revenue.
- **Structured data capture**: every conversation feeds structured signals back to the CRM — intent level, service interest, objections, pricing sensitivity — which the Sales team uses to prioritize their day.

### Why WhatsApp specifically

WhatsApp is the single most intimate, most-used messaging platform in many markets. It has a 98% open rate, near-zero email-style fatigue, and prospects feel comfortable sharing context there that they would never type into a website form. Building an AI assistant directly on WhatsApp meets prospects exactly where they already are, with zero friction to start a conversation.

### How it is built

The assistant combines **nocode** components (for the conversational flows and integrations) with **custom development** (for the logic that nocode cannot express cleanly: quote generation, advanced calendar rules, exception handling, CRM sync). This hybrid architecture makes the assistant both fast to iterate on and solid enough to power a real commercial operation.

The question is no longer whether a business needs an AI assistant. It is why it does not have one yet.

### Technology stack

- **WhatsApp Business API** for the messaging channel.
- **AI / chatbot layer** (large language models) for conversation, qualification, and quote generation.
- **Nocode platform** for conversational flows and integrations.
- **Custom development** for business-specific logic and CRM synchronization.
- **Calendar integration** for real-time appointment booking.

---

**Ready to automate your business?** [Discover our AI Automation service →](/expert-automatisation-ia)

Régina, your 24/7 AI WhatsApp assistant: qualifies leads, books appointments, sends quotes, answers FAQs. Even at 3 AM. Nocode + custom development.

AI WhatsApp Assistant 24/7 — Customer Relations

AI Assistant on WhatsApp for 24/7 Customer Relationship

A custom chatbot that books appointments, qualifies leads, and answers FAQs 24/7 on the BOVO Digital site — built in one morning, converting prospects ever since.

## Our own chatbot, built in a single morning — and converting prospects ever since

This project is a textbook example of what every agency should do: build for yourself what you build for clients. For months, the creation of a chatbot for our own BOVO Digital website had been pushed back, again and again, because client work always came first. One dedicated morning, we shipped it. Within days it was already qualifying prospects, booking calls, and answering FAQs while we slept. It became an instant case study in the power of internal tooling.

### What the chatbot actually does

The assistant was designed from day one as a **real virtual sales assistant**, not a "hello how can I help you" toy. It handles the critical stages of the prospect journey end to end.

- **Automated appointment booking**: the bot is connected to the team's calendars and proposes genuinely available slots in real time. No more email ping-pong to find a 30-minute window; the prospect picks a slot and it is booked, confirmed, and added to both calendars instantly.
- **Smart availability checks**: conditional logic cross-references multiple team calendars, respects working hours and time zones, and surfaces the best options for each prospect depending on the service requested. A "discovery call" with the founder does not surface the same slots as a "technical scoping call" with an engineer.
- **Dynamic FAQ module**: the bot immediately handles recurring questions (pricing ranges, typical timelines, list of services, how we work, where we are based, collaboration process). Prospects get an instant answer and the team stops repeating the same paragraph five times a day.
- **Lead qualification**: before a human ever sees the conversation, the bot asks strategic, calibrated questions to judge whether the prospect is a real opportunity. Unqualified leads get the information they need without hijacking the team's calendar.
- **Commercial pre-negotiation**: the bot gently steers prospects toward realistic price ranges based on scope, so the first call with a human starts from a healthy baseline instead of from a twelve-thousand-euro gap between expectation and reality.

### How it is built

The implementation combines **nocode** tools and **custom development**. Nocode was used for the rapid bootstrap of the conversational interface and the simpler logic branches. Custom development was layered on top for everything nocode alone cannot handle cleanly: team calendar integration, advanced qualification scripts, exception handling, and tracking each conversation so we can continuously improve the scripts based on real data.

This hybrid approach is the honest answer to a tradeoff every agency faces: pure nocode is fast but hits walls; pure custom is powerful but slow to ship. Using both, with a clean boundary between them, gives the best of each world.

### Project impact

- Drastic reduction of the time spent dealing with unqualified prospects.
- Sharply increased responsiveness to inbound requests, including outside business hours.
- Better user experience, with 24/7 availability and instant answers.
- Strategic productivity gains for the team, now spending their time on high-value conversations rather than on triage.

### A broader lesson for entrepreneurs

Beyond the immediate metrics, this project highlighted an essential truth for any operator: **take the time to build your own internal tools**. The leverage is extraordinary. An optimized internal process does not just improve your operations — it becomes a visible proof of capability that every prospect on your site can experience firsthand.

At BOVO Digital, this chatbot is not a gimmick. It is a live demonstration of the kind of automation and AI we apply to our clients' businesses. Every visitor who talks to it is already seeing our work. That is the ultimate form of marketing: showing, not telling.

### Technology stack

- **n8n** for the orchestration backbone and integrations.
- **Nocode conversational layer** for rapid iteration on the chat scripts.
- **Custom development** for calendar integration, qualification scripts, and exception handling.
- **Calendar API** for real-time availability and booking.
- **Internal database** for response management and continuous script improvement.

---

**Ready to automate your business?** [Discover our AI Automation service →](/expert-automatisation-ia)

BOVO Digital AI chatbot: automated appointment booking, FAQ, lead qualification and pre-negotiation. 24/7 without human intervention. n8n + custom dev.

AI Chatbot: Book Appointments + FAQ + Qualify

Intelligent Chatbot for BOVO Digital

Calendly link sent but no booking yet? Automated follow-ups at 24h and 48h recover hesitant leads — with Sales team escalation if they still don't act.

## Calendly link sent but not clicked? The system takes over, automatically

One of the most frustrating leaks in a modern sales funnel is the warm prospect who asks for the booking link, receives it, says "great, thanks" — and then never actually books. The intent was there; the follow-through was not. This workflow was built specifically to plug that leak, because losing a booking-ready prospect for the price of one forgotten click is unacceptable.

### The follow-up sequence

**Unclicked link detection**
Make.com regularly checks the Google Sheets CRM to identify prospects who received their Calendly link but have not yet used it to schedule a call. The check is automated, continuous, and invisible — no one has to remember to run it.

**First follow-up at 24 hours**
A gentle reminder goes out after 24 hours: the link again, a friendly tone, zero pressure, minimal friction. Often this single message is enough — the prospect simply forgot or got distracted, and the nudge brings them back.

**Second follow-up at 48 hours**
If still no booking, a second, more direct message goes out with an alternative offer: a specific suggested time slot rather than a generic "pick anything" link. Humans convert much better on concrete proposals than on open-ended calendars, and this variation is intentional.

**Escalation to the Sales team**
After 48 hours and two automated attempts, a Telegram alert is pushed to the Sales team. A human now steps in — a phone call, a personal LinkedIn message, a warm "I just wanted to make sure everything is OK on your side". This is where automation hands over to a human, with full context on everything the prospect has already received.

### The business logic behind the delays

The 24h / 48h / human spacing is not arbitrary. It is calibrated to feel like a patient, professional follow-up — not like spam. Prospects notice the difference. Aggressive follow-ups damage the relationship; thoughtful ones rebuild intent. The workflow treats the prospect the way a good human sales rep would, just at scale and with perfect reliability.

### What this delivers for the client

- Prospects who would otherwise have silently disappeared actually book the call.
- The Sales team only intervenes on genuinely stubborn cases, saving their energy for the conversations where it matters.
- The whole pipeline becomes visible: at any time, the client can see exactly how many prospects are stuck at the "link sent but not clicked" stage and how many recovered.
- Recovery-rate reporting is automatic — the business can measure, in dollars, how much revenue this single workflow rescues every month.

### Technology stack

- **Make.com** for orchestration, scheduling, and escalation logic.
- **Twilio SMS** and **Twilio WhatsApp** for the automated follow-ups.
- **Google Sheets** as the live CRM and audit log.
- **Calendly** as the booking platform and source of truth for click events.
- **Telegram Bot API** for escalating to the Sales team in real time.

---

**Ready to automate your business?** [Discover our AI Automation service →](/expert-automatisation-ia)

Calendly link not clicked → follow-up at 24h, then 48h with alternative, then Telegram Sales escalation. No hot prospect lost.

Unclicked Calendly Link: Auto Follow-Up

Active Follow-up of Unclicked Calendly Links

Slash your no-show rate with SMS and WhatsApp reminders at 24h, 6h, and 15min before each appointment — automated, reliable, and zero extra effort.

## A no-show rate cut in half — without asking the Sales team to lift a finger

A missed appointment is a lost deal. For any business whose pipeline relies on booked calls — coaches, agencies, real-estate brokers, financial advisors — a no-show is not just wasted time, it is wasted pipeline value. This workflow ensures that every prospect who books shows up, through a calibrated sequence of reminders across SMS and WhatsApp that runs without human supervision.

### The reminder system, step by step

**Reminder #1 — 24 hours before the appointment**
A personalized message is sent exactly 24 hours before the booked time. It reconfirms the appointment, restates the benefit the prospect is going to get from it, and explicitly asks for a confirmation reply. Phrased correctly, this single message already lifts the show-up rate by a meaningful margin.

**Reminder #2 — 6 hours before**
A second reminder is sent 6 hours before. This one includes a polite reschedule link in case plans have changed. The goal is not to guilt the prospect into showing up; it is to make rescheduling effortless so the lead does not simply ghost.

**Reminder #3 — 15 minutes before**
A final reminder is sent 15 minutes before the appointment: the direct call link, simple connection instructions, a warm human tone. This is the one that catches the prospect while they are already at their desk and nudges them into action.

**Orchestration with Make.com and Calendly**
The scenario synchronizes with Calendly to know the exact timing of every booking. Twilio handles the delivery of SMS and WhatsApp messages with guaranteed tracking: delivered, read, replied, failed. If a message fails to deliver (bad number, opt-out), the Sales team receives an alert instead of silence.

### What this system delivers

- A no-show rate that drops materially — typically cut in half for well-configured accounts.
- A completely hands-off experience for the Sales team: they focus on the calls, not on the reminders.
- A consistent, on-brand communication tone across every reminder, every prospect, every channel.
- A measurable, reportable process where every sent and received message is logged for later analysis.

### Why reminders deserve their own engineered workflow

Most teams treat reminders as an afterthought — a quick email template scheduled in the calendar tool. But the delta between a 50% show-up rate and an 85% show-up rate is measured in revenue, not comfort. Investing in a dedicated multi-channel reminder workflow is one of the highest ROI automations a sales-driven business can run.

### Technology stack

- **Make.com** for the scheduling and orchestration logic.
- **Calendly** as the source of truth for appointment times.
- **Twilio SMS** and **Twilio WhatsApp** for multi-channel delivery.
- **Google Sheets** as the log and dashboard of every sent reminder.

---

**Ready to automate your business?** [Discover our AI Automation service →](/expert-automatisation-ia)

Automatic reminders at 24h, 6h, and 15 min before the appointment via SMS/WhatsApp. No-show rate slashed effortlessly. Make.com + Calendly + Twilio.

Automatic Appointment Reminders — SMS + WhatsApp

Automated Appointment Reminders

Every WhatsApp reply is instantly captured, the lead identified, and routed to the exact AI scenario — no manual sorting, no missed opportunities.

## Every WhatsApp reply identified, qualified, and routed to the right AI scenario

In a multi-prospect automated funnel, an incoming WhatsApp reply is never a generic event. The same message — "Tell me more" — can mean completely different things depending on where the prospect is in the journey: is it a brand-new lead curious about your offer? A qualified prospect asking for the demo link? A booked client about to cancel? The value of this automation is to **understand the context in real time** and route the message to the exact AI scenario designed to handle it — no human triage required.

### What this workflow does

**Webhook-based reception**
Every incoming WhatsApp message fires a Make.com webhook in real time. There is no polling, no delay, no batch job. The scenario activates within seconds of the prospect pressing send, which is critical for conversion — prospects expect near-instant replies on WhatsApp.

**Contact identification**
The system queries the Google Sheets CRM to identify who this prospect is, at which stage of the funnel they currently sit, which scenario they came from, and which closer (if any) is assigned to them. This contextual lookup is what makes the rest of the workflow work: without it, the AI would be forced to restart the conversation from scratch each time.

**Smart routing to the right scenario**
Based on the prospect's current status, the reply is routed to a specific downstream Make.com scenario:

- **Qualification** if the lead is new and in the middle of the NEPQ flow.
- **Objection handling** if the lead is qualified but hesitant.
- **Appointment booking** if the lead is ready and needs the Calendly link.
- **Reminder / reschedule** if the lead already has a booking coming up.
- **Human escalation** if the case is flagged complex or commercially sensitive.

Each route is a separate scenario with its own prompts, its own tone, and its own next steps. One router, many specialists.

**GPT-4 intent analysis**
The message content is analyzed by GPT-4 to detect intent: interest, objection, question, confusion, cancellation, booking confirmation, and so on. The classification is injected into the downstream scenario so the AI responds in the right register — not a canned reply, but one tuned to the real emotional state of the prospect in that moment.

### Why this architecture matters

Without this router, your inbound WhatsApp stream becomes noise: a single flat queue where every message is treated the same way. With the router, every inbound message enters the funnel as a **labeled, identified, contextualized event**. The rest of the system can then behave like a team of specialists instead of a single overwhelmed junior agent.

This is the kind of infrastructure choice that separates a toy WhatsApp automation from a real commercial system capable of handling thousands of concurrent conversations without dropping a single lead.

### Technology stack

- **Make.com** as the routing backbone and scenario orchestrator.
- **Twilio WhatsApp** for the messaging channel.
- **Google Sheets** as the live CRM and routing table.
- **OpenAI GPT-4** for intent detection and classification.
- **Webhooks** for the real-time trigger layer.

---

**Ready to automate your business?** [Discover our AI Automation service →](/expert-automatisation-ia)

Every WhatsApp reply identified and automatically routed. GPT-4 analyzes intent (interest/objection/question) and adapts the response. Make.com.

Smart WhatsApp Response Routing — GPT-4

Intelligent WhatsApp Response Processing

Silence from a lead? Automatic 6 AM and 12 PM re-engagement messages go out instantly — with a Telegram alert for your Sales team if the prospect stays cold.

## No more warm leads lost in silence

A prospect who does not reply after the first message is rarely a rejection. They are usually distracted, traveling, in a meeting, or simply catching the message at the wrong moment. Letting that silence settle in is one of the most costly mistakes in any conversion funnel — the lead cools off, interest fades, and the deal is gone. This system treats silence as a signal to follow up, not as a signal to give up.

### How it works

**Silence detection**
After the first WhatsApp or SMS qualification message, Make.com monitors each prospect's status in the Google Sheets CRM. If no reply is registered within a configurable delay (typically a few hours), the re-engagement scenario activates automatically. Every prospect is tracked individually, so late sign-ups are never forgotten and early responders are never spammed.

**Smart reminders at 6 hours and 12 hours**
A first reminder is sent at the 6-hour mark, then a second at the 12-hour mark if the prospect is still silent. The messages are different from the first contact — different tone, different hook, different angle — so the prospect never feels like they are receiving the same message twice. The goal is to feel like a thoughtful human nudge, not an automated loop.

**Telegram alert for the human team**
If the prospect remains silent after both automated reminders, a Telegram notification is pushed to the Sales team. A human now takes over, armed with full context: the original message, the two reminders sent, the timestamps, the CRM status. The handoff is clean — the human does not start cold, they resume a conversation that has already been initiated.

**Real-time Google Sheets tracking**
Every action — first message, reminder 1, reminder 2, reply received, escalation to human — is logged in the Google Sheets CRM in real time. Management always has a live view of the re-engagement pipeline: how many prospects are being nudged, how many recover after automation, how many escalate to the team, and what the recovery rate looks like week over week.

### The commercial impact

- **Fewer lost leads**: prospects who would have disappeared after one ignored message often re-engage after the 6h or 12h reminder.
- **Better Sales team focus**: the team only takes over when automation has exhausted its polite attempts, so their time is spent on genuinely uncertain cases.
- **Calm, consistent follow-up**: no aggressive spam, no irritating tone, just a structured second and third touch that respects the prospect.
- **Measurable recovery rate**: because everything is logged, the client can actually see — in numbers — how much revenue this automation saves every month.

### Why silent-lead recovery deserves its own scenario

Most funnels treat follow-up as an afterthought, buried inside a generic email sequence. But silent prospects are a specific category with a specific behavior, and they deserve a dedicated scenario with channel-aware timing, message variation, and a clean escalation to humans. Isolating this logic in its own Make.com scenario makes it easy to tune, easy to A/B test, and easy to improve over time.

### Technology stack

- **Make.com** for the full re-engagement orchestration.
- **Twilio WhatsApp** and **Twilio SMS** for the reminder messages.
- **Google Sheets** as the CRM and audit log.
- **Telegram Bot API** for operational alerts to the Sales team.

---

**Ready to automate your business?** [Discover our AI Automation service →](/expert-automatisation-ia)

No more lost silent leads: automatic follow-ups at +6h and +12h (WhatsApp/SMS), then Telegram alert for Sales team. Make.com + Twilio + Google Sheets.

Automatic Follow-Up for Silent Leads

Automatic Follow-up for Silent Leads

Cold lead to booked appointment without a single human touchpoint — a multi-channel AI funnel (WhatsApp, SMS, Email) with NEPQ qualification and auto-scheduling.

## From capture to signature — a multi-channel funnel that runs itself

Cold leads rarely turn into clients on the first touchpoint. The difference between a conversion rate of 2% and a conversion rate of 12% is almost entirely about what happens **after** the sign-up: how fast, how consistent, how personalized the follow-up is across channels. This project is a full-stack conversion funnel that handles the entire journey from the opt-in form to the signed deal, with AI qualification at the center and four channels working in concert.

### The automated journey, end to end

**Capture and trigger**
The moment a prospect registers through Systeme.io, a webhook fires Make.com. The workflow starts immediately — no batch job, no delay, no risk that the prospect cools off before the first message arrives.

**AI qualification (NEPQ framework)**
A conversational AI agent opens the dialogue with each new prospect on WhatsApp. It follows a NEPQ-inspired question flow (Neuro-Emotional Persuasion Questioning): problem-awareness questions, consequence questions, solution-awareness questions, commitment questions. The goal is not to "sell" in a cold DM — the goal is to qualify in real time whether this prospect is worth a call with a human closer. GPT drives the conversation, interpreting free-text answers and classifying intent rather than forcing the prospect down a rigid decision tree.

**Smart multi-channel follow-up**
If the prospect goes silent, the system automatically re-engages: WhatsApp first (because it is where the conversation started), then SMS after a calibrated delay, then email as a safety net. Each channel has its own message, its own tone, its own timing. No prospect is ever abandoned after a single missed reply.

**Automated Calendly booking**
When the AI detects commitment, it sends the Calendly link at exactly the right moment. If the prospect books, the system fires off confirmation and reminder sequences automatically — 24 hours before, 1 hour before, 10 minutes before. Calendar events, reminders, timezone handling: all invisible to the prospect, all flawless.

**Segmentation and Google Sheets CRM**
Every prospect is tagged in real time with a status: qualified, unqualified, silent, no-show, closed, lost. A lightweight Google Sheets CRM serves as the single pane of glass for the Sales team. Filters, pivot tables, and per-closer dashboards are built in. A weekly report is generated automatically per closer, so management sees performance without asking for spreadsheets.

### Why this stack beats a classic CRM-only approach

A traditional CRM tracks what happened. This funnel **makes things happen** — it is an operational system, not just a reporting system. The AI qualification layer is the key: it transforms the top of the funnel from "form fills that someone will eventually call" into "prequalified conversations the Sales team can triage in minutes".

### Technology stack

- **Systeme.io** for landing pages and lead capture.
- **Make.com** as the orchestration backbone across every channel.
- **Twilio WhatsApp** and **Twilio SMS** for multi-channel outreach.
- **OpenAI GPT** for the NEPQ qualification assistant.
- **Calendly** for booking and automated reminders.
- **Google Sheets** as the CRM, dashboard, and reporting layer.
- **Telegram** for real-time operational alerts to the Sales team.

---

**Ready to automate your business?** [Discover our AI Automation service →](/expert-automatisation-ia)

100% automated multichannel sales funnel: AI NEPQ qualification, WhatsApp/SMS/Email follow-ups, Calendly scheduling. Capture to close without humans.

Automated Multi-Channel Funnel — AI + Calendly

Multi-Channel Automated Funnel with AI Qualification and CRM Tracking

Every morning at 8 AM, a Make scenario reads Airtable, generates a full GPT article, structures it in Divi, and publishes it live on WordPress — zero manual effort.

## Every morning at 8 a.m., your WordPress blog enriches itself — with zero human clicks

This is a complete editorial production pipeline: from Airtable to WordPress, through GPT and Divi, with Yoast SEO metadata fully configured — and nobody ever pressing the "Publish" button. Airtable is the editorial calendar; Make.com is the conveyor belt; GPT is the writer; Divi is the designer; Yoast is the SEO department; WordPress is the press. All of it, running on a silent schedule.

### The workflow, step by step

**Daily automatic trigger**
Every day at 8 a.m., Make.com reads the Airtable base to detect which articles are scheduled for that day. The calendar is entirely owned by the editorial team — they drag, drop, tag, and prioritize inside Airtable — and the automation simply executes whatever the calendar says. No forgotten article, no missed day, no manual check.

**GPT-powered content generation**
For each scheduled article, Make.com retrieves the title, the brief, a custom prompt tailored to the topic, and the set of images to be used. GPT generates a complete, structured article following the parameters defined in the brief: tone, depth, audience, target keyword, length range. Because every article has its own prompt, quality stays high across very different editorial angles without re-engineering the pipeline.

**Divi formatting and Yoast SEO injection**
The generated content is serialized into JSON, merged into a Divi template using shortcodes, and enriched with all the Yoast SEO metadata required for the article to rank: meta title, meta description, focus keyphrase, OpenGraph tags, Twitter card fields. The article arrives in WordPress already looking like a finished, on-brand post — not a raw GPT dump that still needs two hours of layout work.

**Automatic WordPress publication**
The fully assembled article — images, body copy, metadata, categories, tags, featured image, author — is published to WordPress via the REST API at exactly the scheduled hour. The client wakes up to a new article live on the blog, indexed for SEO, and ready to be shared on social channels.

### What this means for the business

- **An active blog every single day**, without any editorial team burning hours in the back office.
- **Consistent quality**, thanks to reusable templates, structured prompts, and a brand-voice layer baked into the pipeline.
- **SEO optimized from publication**, with Yoast metadata populated before the article goes live instead of being an afterthought a week later.
- **Frictionless scaling**: going from 1 article a day to 5 articles a day is a configuration change in Airtable, not a change in the workflow.
- **Auditability**: every run writes back to Airtable with the final WordPress URL, publication timestamp, and any error, producing a complete history of the editorial pipeline.

### Why this stack works together

Airtable is where humans think; Make.com is where workflows are visible; GPT is where content is produced; Divi is where design is enforced; Yoast is where SEO is applied; WordPress is where the audience reads. Each tool does what it does best, and the glue is engineered — not improvised — so the whole thing is stable, maintainable, and easy to evolve as the editorial strategy changes.

### Technology stack

- **Airtable** as the editorial calendar and content brief repository.
- **Make.com** as the automation and scheduling engine.
- **OpenAI GPT** for article generation.
- **Divi** (WordPress theme framework) for consistent on-brand formatting.
- **Yoast SEO** for metadata injection.
- **WordPress REST API** for end-to-end publication.

---

**Ready to automate your business?** [Discover our AI Automation service →](/expert-automatisation-ia)

Every morning at 8 AM: Make.com reads Airtable, GPT writes, Divi formats, WordPress publishes. Fully autonomous editorial chain with Yoast SEO.

Daily WordPress Blog Auto — Make + GPT

Automated WordPress Article Publishing with Make & GPT

Hundreds of PrestaShop product sheets rewritten for SEO by AI in one workflow — unique titles, meta-descriptions, and anti-duplicate checks powered by n8n + GPT.

## Hundreds of PrestaShop product sheets optimized for SEO in a matter of hours, not weeks

A PrestaShop catalog with no SEO is, essentially, an invisible catalog. Product pages rank on long-tail queries only if titles, descriptions, and metadata are unique, structured, and keyword-aware. Writing that content by hand for hundreds of SKUs is the kind of job that never gets finished. This project turned it into a repeatable, automated pipeline built around an **n8n workflow** that produces unique, SEO-ready content for every product in the catalog.

### The full pipeline

**Catalog export**
The existing PrestaShop catalog is exported to CSV and loaded into Google Sheets. This intermediate sheet becomes the shared working base between the client, the editorial reviewer, and the automation: a single source of truth for every product's "before" and "after" state.

**AI-driven enrichment**
For every product row, n8n fires a call to GPT or Gemini with a carefully engineered prompt. The prompt is structured to produce four distinct outputs per product:

- A commercial, keyword-aware product title.
- A unique, structured commercial description that highlights benefits, materials, and use cases.
- An SEO meta-title optimized for the target query (with the primary keyword placed early).
- An SEO meta-description that matches search intent and pushes click-through.

Prompts include brand voice constraints and category-specific instructions, so products in different categories receive genuinely different treatments instead of cookie-cutter content.

**Anti-duplicate verification**
A similarity-check layer compares each newly generated description against everything already produced in the batch. When two outputs look too close — a real risk with AI generation at scale — the offending row is flagged for regeneration with an adjusted prompt. This step is what separates a production-grade pipeline from a toy: Google penalizes duplicate content within the same catalog, and without this safeguard the whole effort would be counterproductive.

**Reassembly and bulk reimport**
The enriched sheet is reformatted into the exact CSV schema PrestaShop expects and reimported in bulk. Zero manual editing of individual product sheets inside the PrestaShop back office, which is typically where projects like this die.

### The business outcome

- **Hundreds of product sheets** turned from generic, often duplicated, content into unique and structured SEO copy.
- **Consistent tone and structure** across the whole catalog — a critical signal for both users and search engines.
- **Hours instead of weeks**: a task that would take a full-time content team a month is compressed into a single workflow run plus a short editorial review pass.
- **A reusable asset**: every time new products are added to the catalog, the same workflow enriches them instantly, keeping SEO quality constant as the catalog grows.

### Why n8n for this workflow

n8n is the perfect orchestrator for this kind of job: it handles rate limits on AI APIs gracefully, exposes every step visually for audit, and lets the client own the workflow internally if they later want to tweak prompts or inject new product categories. Hosting can be self-managed, which keeps sensitive product data inside the client's infrastructure.

### Technology stack

- **n8n** for workflow orchestration, prompt handling, and batch processing.
- **Google Sheets** as the working ledger and editorial control panel.
- **OpenAI GPT** and **Google Gemini** for content generation — interchangeable at the workflow level.
- **PrestaShop CSV import** for reliable, transactional reimport of the enriched catalog.

---

**Ready to automate your business?** [Discover our AI Automation service →](/expert-automatisation-ia)

Hundreds of PrestaShop product sheets SEO-optimized in hours: unique titles, descriptions, meta tags generated by GPT/Gemini via n8n. Anti-duplicate included.

PrestaShop SEO Automation — n8n + AI

PrestaShop Product Sheet SEO Optimization with n8n & AI

A custom WordPress plugin that fires a partner API callback the instant a prospect submits your Elementor form — secure, conflict-free, sandbox-tested, and fully documented.

## A custom WordPress plugin that connects an Elementor form to a partner API — cleanly, securely, and without conflicts

The client operated a WordPress landing page built with Elementor and had a commercial partner whose callback API had to fire automatically on every submission of the "demande_rappel" (callback request) form. Off-the-shelf connectors were either generic, clunky, or incompatible with the specific authentication and payload requirements of the partner. The solution was a **purpose-built PHP plugin**, designed as a proper WordPress extension rather than a fragile snippet pasted into functions.php.

### The technical challenge

Generic plugins fall short when you need:
- Specific API authentication with tokens rotated at runtime.
- Precise data formatting the partner's backend will accept (including strict phone number normalization and country codes).
- Proper error handling so a failed API call never leaves the client with a lost lead and no trace.
- Zero interference with the dozens of other plugins and themes running on a typical production WordPress stack.

Meeting all four at once requires custom development, and a clear separation between WordPress concerns and integration logic.

### What the plugin does

**Automatic triggering on form submission**
The plugin hooks into Elementor's native submission event for the "demande_rappel" form only. As soon as a visitor submits, the plugin intercepts the data and fires the API call in real time — fully server-side, so no client-side tampering can break it.

**Secure authentication**
The partner's API credentials are stored using the WordPress Options API with appropriate sanitization, never exposed in JavaScript, and transmitted correctly on every request (signed header, bearer token, or HMAC depending on the partner's protocol). Token rotation is supported without taking the site down.

**Phone number formatting**
User-submitted phone numbers are automatically normalized (E.164 or the partner's specific format), stripping spaces, dashes, and parentheses, adding the correct country code, and rejecting obviously invalid numbers before the API is ever contacted. This eliminates a huge class of silent failures the client previously experienced.

**Sandbox environment testing**
Before going live, the plugin was fully validated against the partner's sandbox environment. Every payload, every edge case, every error code path was exercised so production rollout carried zero surprise.

**Error handling and logging**
When the API call fails (network error, authentication error, business-rule rejection), the failure is captured, logged in a dedicated WordPress admin log table, and the team is notified. No prospect is ever lost silently. The original form submission is retained so the Sales team can follow up manually if needed.

**Clean integration, zero conflict**
The plugin is built following WordPress coding standards: namespaced classes, prefixed hooks, enqueued scripts only where needed, internationalization-ready strings, and an uninstall routine that cleans up cleanly. It coexists with Yoast SEO, WooCommerce, caching plugins, and the dozens of other plugins a typical site runs.

**Complete documentation**
The plugin ships with a README covering installation, configuration, admin screen walkthrough, troubleshooting, and API specification references. Handing it over to the client's internal team or any future developer takes minutes, not days.

### Why a custom plugin beats a no-code integration

On a marketing-critical form, silent failures are unacceptable. A custom plugin gives the client full observability (logs, retry, alerts) and full control (the code is theirs, hosted on their server, no third-party dependency that can break or increase in price overnight). For a team that cares about lead hygiene, the small upfront cost pays back within weeks.

---

**Ready to build your premium website?** [Discover our Premium Web Development service →](/expert-site-web-premium)

Custom PHP plugin: Elementor form → partner API triggered instantly. Secure auth, phone formatting, error logging. Zero lost prospect.

WordPress Plugin — Custom Partner API Integration

WordPress Plugin for Partner API Integration

Hyper-targeted B2B lead list of French beauty institutes outside major cities — names, phones, addresses, and website-verified services, extracted automatically.

## Qualified B2B prospects delivered ready to contact — no manual research

The client had a very specific problem: find independent beauty and wellness institutes in France, located **outside the large cities**, offering one or more of a short list of specialized treatments. Nothing generic, nothing that a random LinkedIn search could surface. The kind of list that normally takes weeks of assistant work to build — and a custom scraping pipeline solved it in hours.

### The problem statement

The target was not "any beauty salon". The brief demanded institutes that actively offer at least one of the following specialized services:

- Lipocavitation (ultrasonic body contouring)
- Radiofrequency
- Renata França massage
- Maderotherapy (wooden-tool massage)
- Tesla Therapy
- Body Sculpt
- LPG Endermologie
- Cellular stretching

And, just as importantly, these institutes had to be located outside major metropolitan areas — the commercial angle was to reach secondary cities and smaller towns that are typically under-served by supplier prospecting.

Doing this with Google searches and a spreadsheet would have taken months and would have missed most relevant results, because the specialized services are often mentioned only deep in a service page, not in the business name or Google Maps snippet.

### The technical solution

**Multi-source search**
The script queries Google and Google Maps with a combinatorial set of targeted queries (service × geography × descriptor) to identify candidate businesses across every target region. This catches establishments that would never surface on a single query.

**Service verification**
For each candidate, the script downloads the business's website and analyzes its content to confirm the actual presence of the required services. The verification layer understands variations in how these services are named, handles typos, and recognizes both French and English terminology. Only verified institutes are kept in the final list — no false positives padding the deliverable.

**Contact data extraction**
For each verified business, the pipeline extracts: legal name, trading name, full postal address, phone number, website URL, email address when publicly listed, and the specific services confirmed on that site. All fields normalized, deduplicated, and enriched with the region and postal code.

**Geographic filtering**
An exclusion layer removes businesses located inside predefined metropolitan areas, so the final list contains only the secondary-market institutes the client actually wants to call.

### The delivered result

- A clean, qualified, exportable list, ready for the Sales team to use in a prospecting campaign on day one.
- Every row is traceable: the exact source page that proved the service is offered is linked, so the Sales team can reference it when opening a conversation.
- Zero manual research, zero false positives, zero wasted call minutes on non-qualifying institutes.
- A reusable framework: the same pipeline can be retargeted to any other niche or geography in a few hours of configuration.

### Why outsource this kind of list-building?

A B2B sales team's most expensive resource is its time on the phone. Feeding that team with a scraped, verified, filtered list is the single highest-leverage investment in any outbound motion. Instead of spending 30 minutes per hour on Google, the reps spend 55 minutes per hour actually talking to qualified prospects.

---

**Need targeted B2B data or a catalog migration?** [Discover our Web Scraping & Data service →](/expert-scraping-data)

Qualified B2B leads: beauty institutes outside major French cities with verified specific services (lipocavitation, LPG). Python scraping + Google Maps.

B2B Beauty Leads France — Targeted Scraping

B2B Lead Generation for Beauty Services in France

End-to-end lead nurturing: Systeme.io capture, instant welcome SMS via Twilio, and a full email sequence in ActiveCampaign — all orchestrated hands-free by Make.com.

## Capture a lead at 11 p.m. — send them a welcome SMS at 11:01

Reactivity is the first lever of conversion. The gap between the moment someone shows interest and the moment you first respond is often the difference between a closed deal and a forgotten lead. This workflow guarantees that every prospect who signs up through Systeme.io receives an immediate, multi-channel welcome — email and SMS — no matter the time of day or night.

### The complete system

**Capture via Systeme.io**
As soon as a prospect submits the registration form, Systeme.io fires a webhook to Make.com. The workflow activates in real time, so there is never a delay between intent and acknowledgment on the prospect's side.

**Instant welcome SMS via Twilio**
Within seconds of sign-up, the prospect receives a personalized SMS on their phone. First touchpoint, warm tone, immediate professional impression. This single detail often carries more weight than any later marketing asset, because it is unexpected and fast.

**Email sequence via ActiveCampaign**
In parallel, the contact is added to an ActiveCampaign list and automatically enters the challenge's email sequence: confirmation email, value content, daily reminders, a "final day" push. The sequence is pre-authored; automation handles enrollment and timing flawlessly.

**Make.com orchestration**
The workflow handles conditional logic (data validation, duplicate detection, email format checks, phone format normalization), error handling with retries, and parallel synchronization across both channels so the SMS and the first email arrive essentially simultaneously, not minutes apart.

### Why combine SMS and email?

- **Email** offers guaranteed deliverability, rich content, and direct links to the challenge platform, the calendar, or the resource bundle.
- **SMS** offers a 98% open rate and is read in under 3 minutes on average.

Used together, the two channels cover the full spectrum of prospect behavior. Some people live in their inbox; others barely open emails but never miss an SMS. Sending on both maximizes the chance that the first meaningful interaction happens within minutes of sign-up — which is exactly when intent is highest.

### What this project proves

Lead capture is not a finish line. What actually moves the needle is **post-sign-up reactivity**. This workflow enforces that reactivity as a system, not as a hope. The client knows for a fact that every single lead, regardless of time zone, device, or day of the week, gets the same high-quality welcome within the same tight window.

### Technology stack

- **Systeme.io** for landing pages, forms, and lead capture.
- **Make.com** for real-time orchestration.
- **Twilio** for transactional SMS delivery.
- **ActiveCampaign** for the ongoing email nurturing sequence.

---

**Ready to automate your business?** [Discover our AI Automation service →](/expert-automatisation-ia)

Lead signs up on Systeme.io → Twilio welcome SMS + ActiveCampaign email sequence in 60 seconds. Maximum reactivity, duplicate handling, synced channels.

Lead Capture: Automatic SMS + Email

Lead Capture and Email Campaign Automation

Triple daily SMS touchpoints over 5 days for a real estate challenge — a Twilio + ActiveCampaign + Make.com engine that kept leads engaged and conversions high.

## A fully automated 5-day SMS campaign — 15 messages, hundreds of prospects, zero manual sends

Fifteen text messages. Five days. Hundreds of prospects. Zero manual sending. That is what this workflow delivered for a real-estate training challenge where every minute of the Sales team needed to go into closing calls, not into copy-pasting SMS content.

### The context

The client was running a 5-day intensive nurturing challenge. The strategy required **3 SMS per day**, timed at morning, midday and evening, delivered to each participant at the right moment of their own journey, regardless of when they signed up. Done manually, this would have taken one person full-time for a week. With automation, it took zero human hours after setup.

### The Make.com architecture

**Contact management with ActiveCampaign**
Leads are retrieved and exported from ActiveCampaign, the client's CRM. Make.com keeps synchronization real-time: a new sign-up in ActiveCampaign becomes an active participant in the SMS sequence within seconds. Tags and list membership control which prospects are eligible, so late registrations never break the sequence.

**SMS delivery via Twilio**
Each SMS is sent through the Twilio API with precise timing (morning slot, midday slot, evening slot) across five consecutive days. Twilio guarantees delivery tracking, carrier-level reporting, and fallback routing when a carrier misbehaves.

**Smart sequencing logic**
Make.com handles the sending calendar independently for each prospect: someone who signs up on day 2 of the challenge still receives the full 5-day sequence starting from their own day 1. Duplicate prevention, retry logic, and anti-spam cooldowns are built in at the scenario level.

**Real-time monitoring**
Delivery statuses (delivered, failed, carrier bounce, opt-out) are pushed back into the CRM and into a monitoring sheet so the Sales team can see the health of the campaign at a glance.

### Impact on the challenge

The client maintained a constant presence in every prospect's pocket for five days without consuming a single hour of the Sales team's time. The sequence ran autonomously while the team focused on the actual conversions: calls, demos, closing. Completion rates and show-up rates for the challenge's live events both increased materially, because the reminders were reliable, never missed, and never sent at the wrong time.

### Why SMS for this use case?

- **98% open rate**, versus 20–30% for email marketing in the same industry.
- **Read within 3 minutes** for the vast majority of mobile users.
- **No algorithmic gatekeeper** between your message and the prospect — unlike social platforms.
- **Compliant by design**: the workflow respects opt-out requests automatically and never sends to blacklisted numbers.

### Technology stack

- **Make.com** for the full orchestration and scheduling layer.
- **ActiveCampaign** as the CRM source of truth.
- **Twilio** for SMS delivery, deliverability tracking, and compliance.

---

**Ready to automate your business?** [Discover our AI Automation service →](/expert-automatisation-ia)

Fully automated 5-day SMS campaign: 3 SMS/day, precise timing, deduplication. Make.com + Twilio + ActiveCampaign. Zero manual sending.

Automated 5-Day SMS Campaign — Twilio

Lead Capture and SMS Campaign Automation

A Telegram bot that doubles as your content team — send a topic, get a full AI-written article published on your WordPress blog within minutes.

## A Telegram bot that writes and publishes articles in your place

This is the simplest, most accessible version of editorial automation: a Telegram bot connected to an AI writer, which publishes directly to your WordPress site in a matter of seconds. No dashboard, no back office, no logging in — just a chat message and a published article.

### The user journey, end to end

**1. You send a message**
One line in Telegram is enough: "Write an article about e-commerce trends for 2026". There is no complicated interface, no form to fill in, no settings to tweak every time. The topic goes straight from your thumb to the automation layer.

**2. The AI writes**
Your message is forwarded to ChatGPT or the Anthropic API. The AI generates a fully structured article — introduction, body, examples, and conclusion — following a pre-configured editorial template that keeps voice and style consistent across every piece.

**3. WordPress publishes**
The finished article is sent to your WordPress site via the REST API, with title, content, category, featured image placeholder, and publication status configured in advance. You can choose between instant publication and a draft state that waits for human review.

### Why Telegram as the interface?

Telegram is available on every device, with no friction. You never need to reach a dashboard or a back office. You send the command from wherever you are — on your phone, in a taxi, between two meetings — exactly the way you would text a friend.

For entrepreneurs and content creators, this changes the economics of publishing. Ideas no longer die in a notes app waiting to be written up "later". They become articles the moment they cross your mind.

### What this project enables

- **Publish on the move**, without a computer, without opening a CMS.
- **Delegate the writing** of recurring or low-risk topics to AI so you can focus on the articles that really need your voice.
- **Keep a blog active** without dedicating hours every week to content.
- **Experiment fast**: try new angles, new categories, new headlines — every test costs only a message.

> Simplicity is a feature. This bot proves that a powerful automation does not have to be complex to use.

### Technology stack

- **Telegram Bot API** for the command interface.
- **Make.com** for orchestration, error handling, and audit trail.
- **OpenAI / Anthropic** for the article generation layer.
- **WordPress REST API** for direct publication.

---

**Ready to automate your business?** [Discover our AI Automation service →](/expert-automatisation-ia)

Send a topic via Telegram → AI writes the article → WordPress publishes. Active blog from your phone, no computer needed. Make.com + ChatGPT.

Telegram Bot → Write + Publish WordPress Articles

Blog Article Writing Automation via Telegram

Send a Telegram message or fire an RSS trigger — this Make.com workflow writes the full article with AI and publishes it to WordPress automatically.

## Trigger an article from Telegram — or from an RSS feed — and let AI publish it to WordPress on its own

Imagine sending a message to a Telegram bot and, a few seconds later, finding a complete blog article live on your WordPress site. That is exactly what this workflow does. It is a full editorial production pipeline that turns either a human Telegram command or an RSS feed event into a ready-to-publish WordPress article, with no manual writing, no manual formatting, and no manual upload.

### Two independent triggers — one unified pipeline

**Trigger #1 — Telegram**
You send a topic or a command to the dedicated Telegram bot. The scenario instantly receives the instruction, forwards it to the AI layer, and kicks off article generation. Perfect for spontaneous ideas captured while commuting, at a client meeting, or in the middle of a conversation — you do not lose them anymore.

**Trigger #2 — RSS feeds**
The scenario continuously monitors one or more RSS feeds chosen for their editorial relevance (industry news, competitor blogs, official announcements). As soon as a new item matching your criteria appears, the workflow automatically drafts an original article on that topic, using the feed item as research material rather than as raw content.

### The end-to-end editorial chain

**Dual-source AI drafting**
Content can be generated by ChatGPT (OpenAI) or Anthropic Claude, depending on your preferences in tone, quality, or cost. Both APIs are integrated and interchangeable at scenario level, so you can A/B test models or switch if one of them has availability issues without rebuilding anything.

**Direct WordPress publication**
The generated article is pushed to WordPress via its REST API. Depending on your configuration, it goes live immediately or lands as a draft for a quick editorial review. Category, tags, author, featured image — everything is populated automatically.

**Make.com orchestration**
Every step is managed by Make.com: trigger reception, AI call, formatting, error handling, and publication. A visual workflow that is maintainable, observable, and easy to iterate on without touching code.

### Who this system is built for

- **High-cadence blogs** that need to publish multiple times a week to stay relevant in their SEO niche.
- **News sites** that want to react to industry events within minutes instead of hours.
- **Solo entrepreneurs and consultants** who want to maintain an active editorial presence without dedicating hours each week to content.
- **Content agencies** that want to offer clients a "blog that updates itself" service on top of their strategy work.

### Why it works in practice

The key is the separation between **trigger**, **generation**, and **publication**. By isolating each layer, the same scenario can serve very different content strategies: reactive publishing from RSS for news sites, on-demand publishing from Telegram for opinion blogs, or both at the same time for hybrid editorial lines.

---

**Ready to automate your business?** [Discover our AI Automation service →](/expert-automatisation-ia)

Trigger article writing via Telegram or RSS feed. Make.com + ChatGPT/Anthropic generates and publishes to WordPress instantly. Zero manual action.

Auto Blog: Telegram + RSS Feed + WordPress

Article Writing Automation via Telegram and RSS Feed

Bulk SEO article factory: scrape Google's top 4 results, draft a full optimized article with ChatGPT, and auto-publish as a WordPress draft — entirely with Make.com.

## From a list of keywords to a stack of SEO-ready WordPress drafts — end-to-end SEO article creation, fully automated with Make.com and ChatGPT

Content marketing at scale runs into the same wall every time: the production pipeline is too manual. Research takes hours per article. Drafting takes more. Formatting, internal linking, meta tags, WordPress upload — every step adds friction and every step is a place where a team slows down. The teams that win at SEO content in 2026 have one thing in common: they have industrialized this pipeline. This project is exactly that industrialization, built as a Make.com scenario that can produce high-quality, SEO-optimized draft articles **in bulk**, with no human touching a keyboard until final review.

### What the automation does, end to end

**Step 1 — Keyword list ingestion**
A Google Sheets (or Google Drive CSV) document serves as the control panel. Each row contains a query, the target WordPress site, and a few meta fields (category, target persona, internal links to reinforce). The scenario reads this sheet row by row, allowing the client to queue up dozens or hundreds of topics at once.

**Step 2 — Live SERP extraction**
For each query, Make.com fires an HTTP request that retrieves the **top 4 organic Google.fr results**. This is the critical research layer — the same thing a human writer would do manually to understand what Google currently rewards for that query. A Text Parser module extracts the structural information from each competing page (headings, bullet points, semantic entities, FAQ blocks).

**Step 3 — Prompt assembly and AI drafting via ChatGPT**
The extracted SERP data is injected into a carefully engineered prompt sent to OpenAI. The prompt enforces:

- A clear H1 with the primary keyword placed naturally.
- A structured outline (H2/H3) that covers what the top competitors cover, plus differentiators.
- Intro, body, and conclusion with coherent transitions.
- Internal linking placeholders aimed at the client's existing content clusters.
- SEO metadata: meta title and meta description matching the target query's intent.

The result is not a "generic GPT blog post" but a researched, competitor-aware, SEO-structured draft.

**Step 4 — Formatting and cleanup**
The raw AI output is passed through Text Parser and custom transformation modules that clean Markdown, fix stray code fences, normalize heading levels, and prepare the HTML expected by WordPress.

**Step 5 — Publication as draft on the target WordPress**
The formatted article is pushed to the correct WordPress site via its REST API, with the right category, author, tags, featured image placeholder, and meta fields. It lands as a **draft**, ready for a human editor's 10-minute review pass before publication — not as an auto-published black box.

### The engineering choices that make it production-grade

- **Per-site configuration**: a single scenario can publish to multiple WordPress sites based on the sheet row's `target_site` column. One control panel, many blogs.
- **Error handling**: each module has retries and graceful-failure branches. Failed articles are logged back into the sheet with the reason, never silently lost.
- **Cost control**: prompts are tuned for token efficiency, and SERP HTTP calls are cached within a session so re-runs don't rack up bills.
- **Quality gates**: a minimum word-count check and a keyword-density check are built in before the draft is sent to WordPress — anything below the threshold is flagged for human attention rather than published.
- **Audit trail**: every article run writes back to the Google Sheet with the WP draft URL, token usage, and runtime, so the client has a full log of what was produced and at what cost.

### Why this matters for a real business

- **Production speed**: what used to take 4–6 hours per article (research + draft + format + upload) now takes minutes of machine time plus a short human edit.
- **Consistency**: every article follows the same SEO template, reducing on-page SEO variance across the blog.
- **Scalability**: the same setup handles 5 articles a week or 50 — only the keyword sheet changes.
- **Editor focus**: the human team stops writing from a blank page and starts editing pre-structured drafts, which is a much better use of expensive editorial time.

### Technology stack

- **Make.com** as the orchestration layer (scenarios, routers, error handlers).
- **Google Drive / Google Sheets** as the control panel for the editorial pipeline.
- **HTTP Request modules** for live SERP extraction.
- **Text Parser** for structural extraction and cleanup.
- **OpenAI (ChatGPT)** for research-aware article generation.
- **WordPress REST API** for direct publication.

---

**Ready to automate your business?** [Discover our AI Automation service →](/expert-automatisation-ia)

Make.com scrapes top Google results, drafts with ChatGPT, publishes as WordPress draft. Dozens of SEO articles per day with fresh, ranked sources.

Automated SEO Articles — Make.com + ChatGPT

SEO Article Creation Automation with Make and AI

1,400+ spare parts references, translated, priced at a 45% margin, and live on PrestaShop — a Python pipeline that turned a raw CSV into a fully operational catalog.

## Turning a raw 1,400-product spare parts file into a translated, margin-adjusted Prestashop store

Catalog data rarely arrives clean. In this project, the client supplied a raw feed of more than **1,400 references** for 4x4 vehicle spare parts — a CSV/XML file with no proper headers, in English, with inconsistent pricing structure. The goal: turn this messy upstream feed into a fully populated French-language Prestashop store, with correct categories, local pricing logic, and search-engine-friendly titles, within a tight deadline.

The solution was a custom Python import pipeline designed to absorb complexity upstream and hand over a clean, production-ready catalog downstream.

### The data challenges to solve

- **Header-less source file**: the CSV/XML had no column names, requiring preliminary analysis to map each field correctly to the Prestashop schema.
- **English-only content**: all product titles and descriptions had to be translated to French before publication, preserving technical accuracy for automotive parts — a domain where mistranslation is immediately spotted by mechanics and DIY customers.
- **Margin application**: the supplier prices had to be marked up by **45%** to produce the retail prices displayed on the store, with rounding rules for psychological pricing.
- **Prestashop 1.7.8 quirks**: title length is capped, certain special characters break URL rewrites, and the import API has rate limits that must be respected.
- **Category assignment**: each product needed to land in the right node of a multi-level category tree (by vehicle brand, by part family, by compatibility range).

### The pipeline architecture

**Stage 1: raw file ingestion and analysis**
The Python pipeline reads the source CSV/XML, infers the structure, detects anomalies (missing fields, duplicate SKUs, unparseable prices), and produces a cleaned intermediate dataset. Every row that cannot be confidently mapped is flagged in a report for human review, rather than silently dropped.

**Stage 2: automated translation via API**
Each English title and description is sent to a translation API, with domain-specific prompt hints to keep automotive terminology accurate ("bush", "bearing", "suspension arm", "strut" and their French equivalents). Translated strings are cached so re-runs do not repeat calls or costs.

**Stage 3: margin application and price formatting**
Supplier prices are parsed, the 45% markup is applied, and the final price is rounded to the client's preferred format (e.g., X.99 or X.90). VAT calculations are applied appropriately for the target market.

**Stage 4: format adaptation to Prestashop 1.7.8**
Titles are truncated where needed to respect the 128-character limit without losing the most important descriptive terms. Special characters that break URL rewriting are cleaned up. Category paths are normalized to match the target taxonomy exactly.

**Stage 5: batch import via API**
Products are inserted into Prestashop in batches sized to respect the platform's limits, with transactional safety (a failure mid-batch doesn't leave the catalog in an inconsistent state). The import log captures every inserted, updated, and skipped record for later audit.

### Why this kind of project is worth outsourcing

Most e-commerce teams under-estimate the complexity of a catalog migration of this scale. They try to do it in Excel, hit the Prestashop import CSV tool, and discover three weeks later that half their products are in the wrong category, translations are broken, and margins are miscalculated. A purpose-built Python pipeline removes all of that risk in a single engineered pass.

### The delivered outcome

- A **fully populated Prestashop store** with more than 1,400 products, ready to sell on day one.
- **Accurate French translations** across all titles and descriptions.
- **Correct pricing** with consistent 45% margin and psychological rounding.
- **Clean category structure** aligned with the client's merchandising strategy.
- **A reusable pipeline** adaptable to future supplier feed updates with minimal reconfiguration.

---

**Need targeted B2B data or a catalog migration?** [Discover our Web Scraping & Data service →](/expert-scraping-data)

1,400 PrestaShop references imported: auto EN→FR translation, 45% margin applied, categories created. Python, batch import, timeout handling. Zero entry.

Import 1,400 Refs PrestaShop — Auto Translation + Margin

Catalog Import on Prestashop with Automated Translation and Markup

585 products with variations, photos, and descriptions migrated to WooCommerce in record time — a custom scraping script that eliminated months of manual data entry.

## Migrating 585 products with full variations from a legacy site to a brand-new WooCommerce store — without a single line of manual data entry

Catalog migration is one of those projects that looks deceptively simple from the outside and turns out to be one of the most error-prone jobs in e-commerce. A store with 585 products, each with multiple sizes, colors, photos, and rich descriptions, represents thousands of individual data points. Done by hand, it takes months of tedious copy-pasting, and every mistake silently damages the new store's SEO or customer experience. Done by script, it takes days — if the engineering is right.

This project is a textbook example of that second path: a custom-built Python scraping and import pipeline that moved the entire catalog of **payperwear.com** to a fresh WooCommerce installation, variations included, categories included, images included, with the design finalized for optimal browsing.

### The scope of the migration

The client needed the complete contents of their existing site re-created on a new WooCommerce platform. Specifically:

- **585 products** in total, across multiple categories.
- For each product: a full set of **variations** (sizes, colors, sometimes material or fit), each variation tied to its own SKU and stock status.
- Rich **product descriptions** including formatting, lists, and product-specific marketing copy.
- **Photos** in multiple resolutions, aligned with each color variation.
- A coherent **category tree** matching the structure of the source site to preserve navigation logic and SEO topical relevance.

### The technical approach

**Custom scraping script**
A Python-based scraper was written to walk the entire source site, extract structured data from each product page, download the associated images, and normalize the whole catalog into a clean intermediate format. Because the source site had inconsistencies between product templates, the parser was built to tolerate variation in HTML structure without skipping records.

**Variation handling**
WooCommerce's variation model has its quirks. Each product needed its attributes (size, color, etc.) registered first, then the parent product created, then every variation inserted one by one with the right SKU, price, stock, and image. The script automated this multi-step dance flawlessly across all 585 products.

**Image pipeline**
All images were downloaded, renamed for SEO, re-uploaded to the WooCommerce media library, and attached to the correct variation. This alone would have taken weeks by hand; the script handled it in hours.

**Category and tag setup**
Categories were re-created on the new site with the same hierarchy as the source, and each product was assigned to its correct nodes. Category-level SEO text and attributes were migrated alongside.

**Storefront design configuration**
Beyond the data migration itself, we configured the WooCommerce theme for smooth navigation: filters by size and color, clear product grids, responsive design, fast-loading category pages. The result is not just "the same catalog elsewhere" — it is an upgraded version of it.

### The hardest technical problems solved

- **Massive data import without timeouts**: the import ran in carefully sized batches to avoid killing the PHP process or exhausting WooCommerce's memory.
- **Variation deduplication**: same-attribute variations coming from different source product pages were matched and merged where appropriate.
- **Category filter configuration**: custom attribute taxonomies were tuned so the filter plugin displayed the right options on the right category pages.
- **Data consistency**: the script includes pre-flight checks comparing the count of imported products and variations against the source, catching any silent drop.

### The delivered outcome

- A fully functional WooCommerce site, visually and functionally equivalent to the original, with a modernized UX.
- **Zero manual data entry** for the client's team during the whole migration.
- **SEO-friendly URLs, alt tags, and metadata** preserved or improved versus the legacy site.
- **A reusable migration framework** that has since been adapted for other catalog migrations on WooCommerce, Shopify, and Magento.

---

**Need targeted B2B data or a catalog migration?** [Discover our Web Scraping & Data service →](/expert-scraping-data)

585 products with variations migrated to WooCommerce via scraping + API. Complex variations, deduplication, error recovery. Zero manual entry.

Scraping + WooCommerce Import — 585 Products

Web Scraping and Catalog Import on WooCommerce with Variations

A Flutter & Firebase marketplace connecting hungry consumers with discounted unsold food — geolocation, real-time offers, and integrated booking in one clean app.

## PANIGASPI: a Flutter marketplace that rescues unsold food and saves local businesses money

Food waste is one of the most striking paradoxes of the modern retail world. Restaurants, bakeries, sandwich shops, and small supermarkets throw away perfectly edible food every single evening — often bread, pastries, prepared meals and produce that are still fresh and safe to eat, but cannot legally or commercially be sold the next morning. At the same time, millions of consumers would gladly pay a discounted price for those same items if only they knew where and when to pick them up.

PANIGASPI bridges that gap. Built with **Flutter** for the mobile front end and **Firebase** for the serverless backend, the application turns unsold food into a win-win transaction: merchants recover part of their losses, consumers eat well for less, and the environmental footprint of the food industry is meaningfully reduced.

### The marketplace experience for consumers

- **Real-time offer display**: users see the list of baskets available around them right now, with the quantity still in stock updating live as other users purchase.
- **Detailed listings**: each offer shows a product description, the original price versus the discounted price, the merchant's identity, the quantity remaining, and the exact pickup cutoff time.
- **Geolocation-first UX**: the app sorts offers by distance by default and displays them on an interactive map so the user picks the option that fits their commute or errands.
- **Built-in reservation and payment**: the user reserves a basket in a single tap, pays inside the app, and receives a confirmation code to show at pickup. No cash handling, no negotiation at the counter.
- **Order history and favorites**: users can save their preferred merchants, repeat orders in one tap, and track the environmental impact of their purchases over time.
- **Account management**: email/phone registration, secure login, password recovery, profile editing — everything expected from a modern consumer app.

### The merchant side of the platform

- **Quick listing creation**: merchants publish a basket in under a minute, specifying the contents, original value, discounted price, quantity, and pickup window.
- **Automatic cutoff management**: unsold baskets expire automatically at the chosen time, avoiding awkward stock management issues.
- **Sales dashboard**: merchants see their daily revenue from anti-waste sales, track trends, and identify their best-performing basket types.
- **Customer communication**: a direct messaging channel is available in case of questions about the pickup.

### The technical stack and its benefits

**Flutter for a single cross-platform codebase**
A single Dart codebase compiles to both iOS and Android apps, cutting development and maintenance costs roughly in half compared to two native stacks. The native look-and-feel of each platform is preserved thanks to Flutter's widget system, so users never feel they are using a "generic" app.

**Firebase as the serverless backend**
Firestore powers the real-time sync of listings and reservations. Firebase Authentication handles secure login flows. Cloud Functions orchestrate the business logic (payment validation, pickup notifications, basket expiration). Firebase Storage holds the images uploaded by merchants. The result is a backend that scales automatically with user growth, with zero server administration for the client.

**Integrated payment flow**
The app embeds a payment solution that collects funds from the consumer, retains the platform fee, and transfers the balance to the merchant — all inside the app experience, without redirecting the user to an external website.

### The impact on the community and the environment

- **Every transaction prevents food waste** at the local level, with a cumulative effect that grows alongside the user base.
- **Merchants convert a loss into revenue** they previously did not capture.
- **Consumers discover new local businesses** they might never have visited otherwise, strengthening neighborhood retail ecosystems.
- **Transparent environmental reporting** inside the app quantifies the CO₂ equivalent saved, turning every user's behavior into a visible, shareable impact.

---

**Ready to launch your mobile app?** [Discover our Mobile App Development service →](/expert-application-mobile)

PANIGASPI: Flutter & Firebase app for selling unsold food. Geolocated map, 1-tap booking, integrated payment, dual merchant/consumer interface.

PANIGASPI: Flutter Anti-Food Waste App

PANIGASPI - Anti-Waste App

Never write a blog post manually again — a Python app that researches topics via Google Search, drafts full articles with GPT, and auto-publishes to WordPress.

## A custom Python tool that researches, writes, and publishes SEO articles to WordPress without human intervention

Content marketing teams spend a disproportionate share of their time on the mechanical parts of the job: searching Google for references, reading competitors' articles, drafting outlines, writing first drafts, adapting them to a CMS, and finally pressing publish. The thinking happens maybe 20% of the day; the rest is plumbing. This Python tool replaces that plumbing with code.

The project delivered a standalone application that takes a list of target keywords, researches them online, drafts full-length articles with GPT, and pushes the result straight into WordPress — hands-free.

### The full content pipeline in one script

**Keyword-based research via Google Search API**
For each target keyword, the tool queries the Google Custom Search API to retrieve the top-ranking pages. Their titles, meta descriptions, and URLs feed the next step, giving the writing model a real-world view of what competitors are saying rather than a purely speculative draft.

**OpenAI-powered article generation**
The tool sends the keyword, the research findings, and a carefully crafted system prompt to the OpenAI API. The prompt encodes the client's tone of voice, target article length, heading structure, and SEO requirements. The result is a complete, coherent article ready for publication.

**Configurable prompt templates**
Every client has a different editorial line. The tool exposes a simple configuration layer that lets non-developers tweak the prompt — tone, length, style, mandatory paragraphs, forbidden words — without touching the source code.

**Direct WordPress publishing via the REST API**
Once the article is drafted, the tool authenticates against the WordPress REST API and publishes the piece with the right category, tags, featured image, and SEO metadata. Drafts can also be scheduled for later publication or sent to a human editor for review.

### Engineering details that make the tool production-ready

- **Rate limiting** on both the OpenAI and Google APIs to stay within quotas during bulk runs.
- **Error handling** baked in: any failed call is logged and retried with exponential backoff.
- **Standalone executable** packaging so the tool can be deployed on any Windows or Linux machine without installing Python manually — useful for clients who want the automation running on their own infrastructure.
- **User-friendly configuration** via a plain text config file, with sensible defaults for users who just want to "run and forget".
- **Extensible design**: the pipeline is modular enough that research sources, writing models, or target CMS can be swapped without rewriting the core logic.

### The strategic value for the client

- **Content production cost collapses**: a multi-article batch runs in minutes, including publishing, instead of spanning several editorial workdays.
- **Reliable editorial cadence**: publication schedules stop depending on whether the content team is overworked.
- **Scalable SEO strategy**: targeting dozens of long-tail keywords becomes realistic, unlocking a segment of traffic that manual processes cannot chase.
- **Portable approach**: the same tool pattern has since been adapted for multilingual content, niche industry blogs, and programmatic SEO campaigns.

---

**Do you want an automated content engine for your site?** [Discover our Automation service →](/expert-automatisation-ia)

Python script scrapes Google, drafts with OpenAI, and publishes directly to WordPress. Dozens of SEO articles produced automatically, zero clicks.

Auto-Publish WordPress Articles Python + OpenAI

Automated Article Generation and WordPress Posting Tool (Python)

Production-grade Django backend ingesting real-time crypto OHLCV data at scale — async Celery processing, robust error handling, and a fully documented REST API.

## A production-grade Django backend that ingests and serves cryptocurrency market data at scale

Fintech and crypto products live or die by their data infrastructure. A trading dashboard that shows stale prices, or a portfolio tracker that misses OHLCV candles, breaks user trust almost immediately. Delivering fresh, reliable, and consistent market data for thousands of cryptocurrencies — across historical windows, technical indicators, and real-time updates — is a non-trivial engineering challenge.

This project delivered that infrastructure: a robust Django backend designed to retrieve, process, store, and expose cryptocurrency market data drawn from the **CoinGecko Pro API**, ready to power trading dashboards, analytics platforms, or automated strategies.

### The engineering requirements addressed

- Serve **historical price data** across multiple timeframes (1m, 5m, 15m, 1h, 1d) for a large universe of cryptocurrencies.
- Expose **aggregated OHLCV** (Open, High, Low, Close, Volume) candles for technical analysis workflows.
- Provide **market analysis indicators** (moving averages, RSI, volatility metrics, market cap rankings) computed from the raw data.
- Guarantee **consistency and freshness** despite the unreliability of upstream APIs and the sheer volume of assets tracked.
- Stand up to heavy read traffic from front-end applications and automated traders without degrading.

### The architectural choices

**Django as the application framework**
Django was selected for its mature ORM, its first-class PostgreSQL support, its tight integration with Celery, and the robustness of its admin interface for operational tasks. The data models were designed to efficiently store time-series records (prices, volumes, market caps) while keeping queries fast through judicious indexing and partitioning strategies.

**Celery for asynchronous task orchestration**
Ingesting data from CoinGecko Pro for hundreds of cryptocurrencies simultaneously would block any synchronous architecture. Celery workers pull data on parallel tracks, respect the API's rate limits, and feed the database without holding up the rest of the system. Scheduled Celery Beat tasks refresh indicators, compute aggregates, and run housekeeping on a rolling basis.

**PostgreSQL for the persistence layer**
PostgreSQL was chosen for its transactional integrity (critical when dealing with financial data), its strong indexing capabilities for time-series workloads, and its maturity in production environments. Careful schema design and query profiling ensured sub-100ms lookups even across millions of records.

### Resilience and error handling

Upstream APIs fail. Networks drop packets. Data arrives late. A serious crypto backend has to plan for all of that:

- **Retry logic with exponential backoff** on every outbound call to CoinGecko, avoiding temporary rate-limit bans.
- **Dead-letter queues** for messages that fail repeatedly, so operators can investigate without losing data.
- **Data consistency checks** that flag missing candles or suspicious jumps between records, triggering automatic re-fetch.
- **Comprehensive logging and monitoring hooks**, making it easy to spot silent failures before they surface in production.

### CoinGecko Pro API integration

The backend handles every detail of the integration:

- **Authentication headers** and API-key rotation ready for multi-environment use (dev, staging, production).
- **Intelligent pagination** that walks through the full coin universe without hitting request caps.
- **Delta updates**: only the candles that are actually new are fetched, minimizing bandwidth and compute.
- **Batching**: several coins are aggregated into a single upstream request wherever the API allows, cutting latency.

### The business value delivered

- **A reliable foundation** for any crypto product that needs fresh, accurate market data.
- **Operational clarity**: operators can inspect any coin, candle, or task through the Django admin with zero additional tooling.
- **Horizontal scalability**: adding Celery workers is enough to handle more coins or tighter refresh intervals.
- **Reusable architecture**: the same backend pattern has been adapted for traditional financial instruments, sports data feeds, and other time-series-heavy use cases.

---

**Do you need a rock-solid backend or API?** [Discover our Web Development service →](/expert-site-web)

High-availability Django infrastructure for crypto data: Celery workers, PostgreSQL, CoinGecko Pro. Real-time OHLCV, technical indicators, auto retry.

103 PROJECTS DELIVERED — WEBSITES, MOBILE APPS & AI

OUR WORK

Latest Delivered Projects — Websites, Apps & AI

Voice AI (Illico) - Vapi Agent Engineering

Voice AI (Illico) - Make & GHL Orchestration Scenarios

Voice AI (Illico) - Web Analytic Dashboard

Explore 103 Projects: Web, Mobile Apps & AI Automation

Voice AI (Illico) - Vapi Agent Engineering

Voice AI (Illico) - Make & GHL Orchestration Scenarios

Voice AI (Illico) - Web Analytic Dashboard

MaxSEO AI - n8n Workflow Engine

MaxSEO AI - Web Pilot Platform

Intelligent Programmatic SEO Orchestration

TCGdex E-commerce Automation

Data Scraping Pipeline "Adrien_78"

Automate your processes with AI

Orchestrated AI Image Workflow

Multi-Channel Automated Funnel with AI Qualification and CRM Tracking

Voice AI (Illico) - Make & GHL Orchestration Scenarios

PrestaShop Product Sheet SEO Optimization with n8n & AI

Mastered Technologies

Automation & AI

Web & Mobile Dev

Backend & Data

Tools & Platforms