Skip to content
Optim-AI

AI maturity test

Where does your company really stand?

Ten questions on your actual use, your framework, your data, your governance and your dependency on suppliers. Your maturity level and a recommended next step are displayed immediately, without asking for anything. The full report, with your three priorities and the profile of your answers, is sent to you by e-mail.

Before you start

Three lists, about twenty seconds. They are used to adapt the recommendation and do not count towards the level.

The four levels

The scoring is public. Eight control questions are worth 0 to 3 points each, i.e. 24 points, scaled to 100. Two questions describe what is happening in your company (exposure, regulatory signal) and do not count towards the score. The result is a self-assessment, declarative and from a single person: a benchmark, not a measurement.

Discovery0 to 7 points out of 24
AI is a blind spot. What your teams do with AI may be escaping you, and you cannot manage what you cannot see.
Experimentation8 to 13 points out of 24
Uses exist, driven by individuals. The value created is real, but it is neither measured, nor protected, nor transferable: it leaves with the person who created it.
Structuring14 to 19 points out of 24
The subject is under control. The challenge now is staying in control over time: supplier dependency, real cost, quality of outputs and in-house ability to check them.
Scaling20 to 24 points out of 24
Your answers describe a subject ahead of most companies of your size. The risk is no longer inaction; it is accumulated complexity and dependency on a few key people.

How the recommendation is decided

The table is public. The rules are read in order: the first one whose conditions are all met gives the recommendation. The level describes, the recommendation advises.

The classes used

  • Exposure classes are read in this order: strong, then unknown, then moderate, then low. The first condition met applies.
  • Strong exposure: a personal or free account used with data on individuals or confidential documents, or at least three situations ticked among the five in the first question.
  • Unknown exposure: "I don't know" to the first question, or "nobody uses it" or "I don't know" to the question on authorised tools.
  • Moderate exposure: one or two situations ticked, or uses that do not all go through authorised tools.
  • Low exposure: no situation ticked, and all known uses go through authorised tools.
  • Weak control: no written rule, no visibility on data, or 11 points or fewer. Intermediate: a verbal instruction only, visibility you cannot check, or 12 to 15 points. Established: a written rule, visibility on the data of authorised tools, and 16 points or more.
How the recommendation is decided
RuleConditionRecommendation
T1You indicate that AI plays a part in a decision that concerns an individual.Audit, after a mandatory conversation
T2Strong exposure and weak control.Audit
T3Strong exposure and intermediate control.Conversation to decide
T4Strong exposure and established control.Diagnostic
T5Unknown exposure.Diagnostic
T6Content made with AI is published, or "I don't know" to the regulatory signal.Diagnostic, with a conversation (mandatory if "I don't know")
T7Three or more "I don't know" answers.Diagnostic
T8No regulatory signal, no exposure situation (or widespread use alone), all known uses go through authorised tools, established control, 20 points or more, rule and measurement at level 2 at least.Verification conversation
T9Moderate exposure.Diagnostic
T10No regulatory signal, no exposure situation, all known uses go through authorised tools, at least a verbal rule, visibility on the data of authorised tools.No engagement for now
T11All other cases.Diagnostic
  • A regulatory signal is always accompanied by a conversation, whatever the recommendation.
  • Fewer than ten employees: an Audit recommendation becomes a two-day diagnostic.
  • When in doubt, the most cautious recommendation is chosen. It remains a hypothesis, to be checked in a first conversation.

Next step

Not keen to answer ten questions?

A thirty-minute conversation covers the same ground, in less time, and with a real person.

Write to usSee our services