AI audit & mapping

Consumer assistants, home-made automations, agents built without any framework: AI comes into the company through the back door. It is the new shadow IT, and it carries two risks nobody measures: data leakage and the reliability of decisions.

Mapping actual uses

I take stock of what already exists in the organisation, including undeclared uses — because there always are some. The aim is not to sanction but to measure the starting point: which tools, which agents, which data they are given, and which business teams have moved ahead.

Classifying each use under the AI Act

The EU regulation on artificial intelligence classifies each use by its level of risk, from minimal risk to prohibited use, and sets the obligations that follow. It applies to those who provide an AI system, but also to the companies that use one. Each use identified is matched to its category and to the obligations that apply to it. The AI Act in two minutes →

Securing what needs securing

For each risky use, a clear decision: frame it, replace it or stop it. An agent that touches sensitive data or decisions must be supervisable: an agent proposes, a human approves, and the decision is logged.

For whom: executive management and IT departments who want to know where they stand with AI before an incident or an audit tells them.

The audit informs your decisions; on legal questions, it should be complemented by your lawyer's advice.

Let's talk for 30 minutes →