Artificial intelligence
The right question isn’t "where to put AI" but "which problem do we want to solve".
AI is only worth something when tied to a specific, measurable, controlled use. We help organisations identify where it genuinely adds something, integrate it there with discernment, and keep control — over quality, cost, security and the final decision, which stays human.
Discuss a situation
Your teams are already using AI — or you’re being pushed to "do AI".
You’re asked to have an AI strategy. The pressure is real: competitors, board, teams. But between the hype and the uses that genuinely matter for you, the line isn’t clear.
Your teams are already using AI tools — with no framework. Sensitive data may be passing through services you don’t control. Use came before the rules.
You’ve identified a use, but you have doubts. Is it reliable? At what cost? What happens when the model gets it wrong? Who remains accountable for the decision?
In all three cases, the problem isn’t technical. It’s one of discernment: telling what warrants AI from what doesn’t need it.
AI doesn’t solve a problem. It amplifies an intention — good or bad.
Applied to a clear process, it saves real time. Applied to a muddled process, it produces confusion faster. That’s why we never start with the technology, but with the problem: what are we trying to achieve, at what quality, at what cost, with what acceptable level of control?
A useful AI is one whose contribution can be measured, whose outputs can be checked, and where you can take back control at any moment. Our role isn’t to put AI everywhere, but to make you able to decide where it belongs — and above all where it doesn’t. The end goal isn’t your dependence on a tool: it’s your teams' autonomy.
Five types of engagement, from framing to handover
- 01Framing & identifying use cases. We start from your real processes to spot where AI adds measurable value — and where it would be a costly gimmick. You come away with a prioritised list, not a general promise.
- 02Training & awareness. We train your teams in a clear-eyed use of AI: what it can do, what it can’t, how to recognise a wrong answer, which data must never be exposed to it.
- 03Framework & governance of use. We set the rules: which tools, for what, with which data, what controls. Enough to turn unmanaged use into managed use.
- 04Automation & assistants. Once a use case is validated, we build the solution — task automation, workflows, assistants, document search systems (RAG) based on your own content — with human oversight built in.
- 05Evaluation & handover. We measure what the solution actually delivers, document it, and hand your teams what they need to maintain it. No black box for you to depend on.
Our logic: problem → use case → feasibility → build → measurement → autonomy. The technology only comes halfway along, never at the start.
Typical cases, representative of what we work on
- You’re asked for an "AI strategy" and don’t know where to start.
- → Framing workshop + prioritised list of use cases.
- Your teams use AI with no rules.
- → Acceptable-use policy + awareness session.
- A repetitive task costs time every week.
- → Automation or dedicated assistant, with human oversight.
- You’re buried under documentation no one can find.
- → Search system (RAG) over your own documents.
- You’ve launched an AI tool but doubt its real contribution.
- → Independent evaluation: benefit, cost, risks.
What you receive: a framing note with prioritised use cases, an acceptable-use policy, a documented solution where the mandate goes as far as building, a measured evaluation, and a transfer of skills to your teams.
A limited initial framing lets you start without committing to everything.
What we do — and what we don’t promise
What we do: help you decide where AI belongs, implement it where it adds measurable value, and make you autonomous with it.
What we don’t promise: that AI automatically solves your problems, that it is 100% reliable, or that it replaces human judgment. Models get things wrong; we design so that the error is detectable and the decision is always taken back up by a human.
What we expect from you: a real problem to address and access to the people who know the process in question. A good AI solution is built with the people who do the work, not alongside them.
Every engagement is covered by a confidentiality agreement — all the more important when your data feeds a system.
Frequently asked questions
- Do you already have to "do AI" to engage us?
- No. Ruling out the false uses is as much part of the work as identifying the real ones.
- Is our data safe?
- It’s a central question, which we address from the framing stage: which data can be used, with which tools, and which must never leave your premises.
- Will AI replace our teams?
- Our approach aims for the opposite: to augment your teams and their autonomy, keeping the decision human. A well-set-up AI saves time on tasks, not on judgment.
- What are the running costs?
- The running cost is part of the evaluation. A solution that delivers little for a high cost does not deserve to be deployed — and we will tell you so.
- Can we start small?
- Yes, and it’s recommended: a framing workshop or a first limited use case is better than a large, ill-defined project.
A situation to clarify?
You can engage us on this area — or simply outline it, and we’ll tell you how to approach it.
Discuss a situation