Case studies

Tools that are already running. Five systems in production, two solutions ready to deploy.

Each project started from a concrete problem at a service business. Here is what was built, how it works and what it changes for its users. Most of these clients are based in France; the methods and the stack carry over to US businesses as-is.

7projects, from training to e-commerce
5systems in production with their users
91AI agents in service, under supervision
1,220+learners and clients supported by Libreo agents

Figures as of November 2026

In productionProfessional training

Yoocancare: four AI assistants supporting every learner

The need: give each of its 900+ learners one-on-one support, without overloading the instructors.

The solution: Claire explains the lessons, Clémence grades assignments, Constance checks in every Monday, Prudence helps learners launch their practice.

4assistants
900+learners
< 24 hto grade an assignment
In productionDigital services

Netcoaching: a digital agency where agents produce and humans coach

The need: produce websites, campaigns, articles and Google Business Profiles for dozens of small businesses, without piling on production hours.

The solution: specialized AI agents for each service, a human coach for advice, and sign-off before anything is published.

200+clients supported
4services automated
48 hfor a small-business website
In productionSales pipeline

AI setter: every lead called back at the time they picked

The need: call back the leads from a “free audit” ad campaign without delay, at all hours, without tying up a human setter.

The solution: a voice agent that calls back at the requested time, qualifies the lead, then books the closer's appointment in GoHighLevel with the audit attached.

100%callbacks on time
38%leads qualified
0manual data entry
Read the case study
In productionSensitive data

Private AI on-premise: AI installed at the client's site

The need: use AI on confidential documents without sending anything to an online provider.

The solution: open-source models installed on site, answering only from the company's own documents. Successfully running on a PC that is more than five years old.

0data sent outside
64 GBof RAM on the test machine
12 GBof graphics memory
Read the case study
In production · FranceAccounting

Accounting review agent: discrepancies caught before the books close

The need: review journal entries faster, without re-reading every line by hand before month-end close.

The solution: an agent that goes through the entries, matches the statements and flags discrepancies (VAT, duplicates, unusual amounts). The accountant keeps the final call.

Built on French accounting rules (VAT, French GAAP). A US version for QuickBooks or Xero files (sales tax, 1099s) would need its own rule set: let's talk if you run a bookkeeping firm.

3families of checks
100%of entries reviewed
1summary memo per client file
Ready to deploy · tested or already proven

Built, proven, available for your business.

Tested · ready to deploy

AI receptionist

It answers calls, shares your hours and services, books appointments in your calendar and transfers to a human when needed. Evenings and weekends included.

Tested for a healthcare practice

Clinics · Home services · FirmsRequest a demo →
Ran in production · 2026

Libreo Analytics AI: an online store's numbers, as clear answers

Connected to the data of the Klarcha online store (essential oils and dietary supplements), it answered management questions in plain language: sales, products, margins, inventory. It ran in production until the store closed in summer 2026.

Available for Shopify and WooCommerce stores

E-commerce · ManagementSee the offer →
Your project

Does your need look like one of these projects?

Tell us about your situation: we'll tell you what can be reused, adapted or built.