Shipped work · Live demos · Case studies
Everything on this page is real work. Where the tool can be shown, there is a live demo. Where it runs on a client’s private data, there is a case study instead. Clients stay anonymous, and every label tells you exactly what you are looking at.
Numbers moved between tools, checked by rules you can see and edit.
For a 50-person landscaping company. Every morning at 8:30 the tool reads the field app and the payroll software, finds the previous day’s timesheet errors and drafts one SMS per person. A manager reads each message before clicking send, and can change the rules on screen. The demo runs on fictional data.
Open the demo →For a 100-person cocoa exporter. Inspectors survey thousands of farmers and their plantations. The ERP checks every ID and every plot against the same rules, shows who can be bought from under the EU deforestation rules, and confirms every purchase at the collection point. The demo runs on fictional data.
Open the demo →For IT teams that support SAP users. A ticket comes in, the agent sorts it, finds the cases already solved and writes the procedure. A technician approves it or escalates it, and nothing is applied before that click. The demo runs on fictional data.
Open the demo →A 50-person landscaping company logs hours in a field app and pays through payroll software. Making the two agree used to take days of manual checking every month. The tool I built runs the same rules every time, flags anomalies before the month closes, and the CEO edits the rules himself, on screen. The live tool runs on client data, so the link opens the page I hand the CEO with the test files. It covers seven payroll rules and the exact anomaly planted in synthetic data.
Read the explainer →Leads answered, qualified and followed up the same day.
For a group of medical labs in France. Every week the tool reads the company registry, legal notices and job offers for each competitor on your territory. What does not matter is discarded with its reason, and the rest lands in one Monday email, each line linked to its source. It runs as a ready-made tool on a subscription, and this demo uses fictional data.
Open the demo →A Coursera-style learning platform with courses, checkout, certificates and a CRM behind it. I built it and I operate it for a French e-learning company. It sells real courses today.
Visit gptformation.pro ↗The AI seats you already pay for, actually used by the team.
For the same 50-person company as the reconciliation work. Staff were pasting internal documents into personal ChatGPT accounts. I set up a Claude Team plan in ten working groups for sixteen people, with rights mirrored on the existing SharePoint tree, so nobody sees what they should not. The CEO, the assistants and accounting were onboarded first, each with a knowledge base built around their actual week.
For the finance commission of a regional football league. The method for analysing a club’s accounts lived in one person’s prompt. I turned it into a shared app. Each member opens a club’s file, drops in its accounts, reads the report as it is written and downloads it in Word. The demo runs on fictional data.
Open the demo →Start here
I will tell you what I would build first. A short call, no pitch.