Service

AI consulting, from scoping to audit:the best-fit solution, even when it isn’t ours.

We help you choose and integrate AI into your information system, or understand why an existing project falls short. If a third-party tool meets your need better than our platform, we tell you so and we integrate it.

By Jean-Christophe BudinUpdated on

Technology choices

ATG platform, SaaS, open-source agents such as OpenClaw, open-weights models, custom development: we compare the options against your constraints, then integrate the one you pick if you want us to.

Scoping and feasibility

Use cases, available data, confidentiality, budget: we check what is feasible, and at what cost, before the project is committed.

Development practices

Tools, models and shared rules so your developers work with AI while staying in control of their code.

Audit of an existing system

A chatbot or AI project that disappoints: we find out why and propose fixes, backed by benchmarks.

Development teams

AI in your development teams

Part of our work is about how developers use AI: which tools, which models, which rules, and where it actually saves time on your code. The goal is also to reassure the team: AI takes the tedious work, developers stay in control.

Development tools

Code editors and agents (Cursor, Claude Code, OpenCode…) built into the environments your developers already know: that is where they are most productive.

Providers and models

Proprietary or open weights, hosted in Europe or not: chosen on code confidentiality, quality and cost, with capped budgets.

Shared rules

One shared instructions file (conventions, architecture, security) that every tool reads, so AI behaves the same way for everyone.

Audits and quick wins

Automated merge request review, regression checks, documentation, legacy code audit: concrete gains from the first days.

Case study · Ucopia

Around ten developers on a network software product created in 2002: faster delivery, better code quality, robustness preserved.

Read the case study

Audit

A chatbot or AI project that isn’t delivering the expected results?

Vague answers, users drifting away, costs creeping up: we look at what is going wrong and tell you how to fix it. Our advice draws on our own deployments and on benchmarks.

Interviews

With the people who designed the solution, the business teams and a few users: the original goals, the choices made, what is expected today.

Design review

Technical documentation, architecture, models, document processing, prompts: this is often where the results are explained.

Tests on your documents

The solution is tested on your reference documents, with questions whose correct answer is known, to measure rather than guess.

Real usage

Where they exist, logs of the questions asked show what users are really looking for and where the solution fails.

The approach adapts to the project: document assistant, website chatbot, agent, automation. The output is a written diagnosis and recommendations ranked by impact, from quick adjustments to a change of approach if that is the right answer.

See our published benchmarks

Why integrate a solution that isn’t ours

Our platform covers document assistants, support and querying your data. It does not cover everything. For a personal agent wired into your messaging apps, a tool that is already standard in your industry or a very narrow need, an existing tool is often faster and cheaper.

In that case we integrate it with the same standards as our own deployments: access rights, logging, control over where the data lives. That matters all the more for an open-source agent like OpenClaw, powerful but with broad access by default.

Frequently asked questions

Will you always recommend your own platform?

No. We recommend what meets the need. If that is our platform, great. If it is a third-party tool, we integrate it.

How does an engagement work?

A conversation to understand the need, a written recommendation, then the integration if you want it. Expect a few days for scoping and a few weeks for an integration.

Can you audit an AI solution you did not build?

Yes. In-house chatbot, vendor product or integrator project: we need access to the solution and, depending on the case, to its documentation and logs. Expect a few days to a few weeks depending on scope.

Do you help development teams adopt AI?

Yes. We study your practices and tools, choose editors, agents and models with you, set shared rules and deliver quick wins (code review, regression checks, documentation). Expect a few days for a first engagement.

An AI project to scope, audit or integrate?

Tell us about the need and we will tell you which solution looks like the best fit.