Home / Use cases / Methods

Methods

AI training for companies : our method, not a catalogue

We train on what we practise every day, on our own projects. Not a tour of the latest tools : how to frame a use case, write instructions that hold, and know what should never be handed to a model.

  • Content built from your own cases
  • Consultants certified by Anthropic and OpenAI
  • Workshop or multi week accompaniment formats

Your AI diagnostic Book a meeting

Key challenges

Many teams have already opened an AI tool without anyone setting a frame : who is allowed to paste what into a model, what a generated answer actually commits you to, and what must always stay checked by a human. Training that only walks through features leaves these questions open. Ours starts from what causes real problems in professional use.

  • Use without a written frame

    Without an explicit rule, everyone decides alone what to paste into a model and what to do with it. The first incident often reveals that no clear limit had been set.

  • Instructions that do not hold

    An instruction given to a model without a closed list or an explicit ban leaves room for invented answers, especially on a price or a guarantee.

  • The confidentiality of your exchanges

    What actually goes into a mainstream model is not always what people assume. Knowing what should stay out of an AI tool is part of framing, before any tool choice.

  • An obligation of means to understand

    The EU regulation on AI includes support for the upskilling of teams that use these tools, without imposing a specific format or level. It helps to know what that actually covers.

What we do

  • 1

    Framing a use case, before opening a tool

    Separating what can be automated from what should not be, and identifying the repetitive tasks that genuinely cost the most.
  • 2

    Writing instructions that hold

    An instruction is judged by what it prevents, not by what it asks for. We replay cases where a model invented a price or a guarantee, and build the filter that stops it.
  • 3

    Working with agents day to day

    Our own production method, passed on as it is : breaking work down by trade, routing each request to the right skill, and keeping a memory that survives across sessions.
  • 4

    Responsible AI and digital sobriety

    What frameworks such as the RGESN and AFNOR Spec 2314 change concretely in a design decision, and how to measure before optimising.

What stays in your hands

  • A written set of usage rules

    What can be handed to a model, what should never be, and who decides in the borderline cases.

  • Sample instructions reviewed with you

    Instructions built on your own cases, with the explicit bans that make them solid.

  • A memo you can pass on

    A support document that remains after the workshop, so people who join later do not start from zero.

  • Practical landmarks on the regulatory frame

    What the EU AI regulation and GDPR concretely ask for, presented in project order rather than in legal jargon.

Expected results

  • Teams that know where the line is

    Everyone knows what can be handed to a model and what should be kept back, without waiting for an incident to find out.

  • Instructions that resist edge cases

    Explicit bans and server side checks reduce the room left for a model to improvise.

  • Autonomy that stays inside the company

    The method we pass on keeps serving after training, on new cases and with new people.

How we move forward

  1. 1

    A framing conversation

    We look with you at the uses already in place or planned, and the questions that come up most often in your teams.
  2. 2

    Content built on your own cases

    The formats described here are not pre catalogued sessions : the content is built with you, from your own situations.
  3. 3

    Workshop or accompaniment

    Depending on the topic, half a day, a full day, or a multi week accompaniment to anchor a way of working with agents.
  4. 4

    A support that stays

    The memo and the usage rules stay inside the company, to be picked up by your teams after we leave.

Who it is for

This training is for owners and teams at small and medium companies that already use AI tools or are considering it, and want to set a clear frame rather than discover the limits afterwards. It suits a team discovering the topic as much as a team that already has habits to correct.

In the same family

Frequently asked questions

How do we train our teams on AI without limiting it to a tools demo ?By starting from what actually causes problems in professional use : what can be automated, what should not be, and the instructions that stop a model from inventing information. This is the approach we pass on, built on your cases rather than a generic overview.
Are we required to train our teams on AI because of the AI Act ?The EU regulation on AI includes support for the upskilling of the people who use these tools. It is an obligation of means, without an imposed training format or a precise level to guarantee. We recommend checking this point with legal counsel for your exact situation.
Our teams use ChatGPT or Claude at work : what should we watch on confidentiality ?The first question is what actually goes into the model : a poorly designed prompt can contain a name, an address or sensitive data without anyone noticing. We work with you on clear rules for what can or cannot be pasted into an AI tool.
Is the training built around your own tools or generic ones ?Around the method, not a specific tool. We pass on our own way of framing a use case and writing instructions, applicable across different models and different business contexts.
Is this a certifying training, or eligible for funding ?No. These are workshop or accompaniment formats built on your cases, not pre catalogued or certifying sessions. For any funding question, please check with your usual funding body.

Let's talk about your context

We always start by understanding the real work. One conversation is enough to know whether we can help, and how.

Tell us what you need Book a meeting