The AI audit of an SME fits in four weeks, deliverables included.

An AI audit maps what actually happens in the company: what your teams use, the AI features already active in your tools, and the data leaving through both channels.

It is not an infrastructure audit. The question is not whether your servers hold up, but who uses what, on which accounts, with which documents, and what is covered by a written rule.

The subject concerns almost every company: an Odoxa survey in France measures that 12 % of employees already use AI at work, and that nearly one in two has never been trained for it. In an SME, these uses almost always exist before the first written rule.

The audit advances in four fixed weeks. The format is short by design: a mapping that drags on goes stale as it is being written.

The fourth week ends with a restitution to management: what was found, what it costs, what deserves a project, and what does not.

Everything the audit produces belongs to you: the usage register, the data flow inventory, the search cost measurement, the drafted AI usage policy and the costed action plan.

These documents remain usable without us. You can hand the follow-up to your usual provider, run it in-house, or stop there.

That is our engagement rule applied: proof before risk. You never fund one stage on the promise of the next.

Since February 2025, the EU AI Act requires AI literacy measures, and the ability to demonstrate them, from any company whose teams use these tools.

The audit answers by construction: the usage register and the signed policy are precisely the documents that demonstrate the subject is governed, and the documented restitution counts among those measures.

The transparency obligation of 2 August 2026, for its part, only concerns companies that expose a conversational agent or generated content to the public: the audit establishes whether you are concerned, without turning it into a scarecrow.

The restitution leads to a decision, not to a commitment. Three paths are possible: train your teams and govern usage (our Déclic offer), install a private AI over your documents (Bastion), or stop with the dossier.

If the scoping shows the project is not worth it, we tell you so and the dossier stays with you. It has happened, and it is the normal way this works.

Our pay comes from the project, never from reselling licences: the final recommendation carries no conflict of interest.

The audit method comes from what we operate ourselves: ten internal vertical SaaS products in production, more than fifty sites and platforms delivered, and DocAgora, our e-health platform in Portugal, where supervision of usage is not negotiable.

More than seventy clients have worked with us between France and Portugal since 2024. Detailed case studies are available on request, during the diagnostic.