AI strategy
Identify and specify your use cases.
An AI strategy engagement, carried out with your leadership and your business teams. Together, we identify and then specify your use cases: for each one, the task, the team that hands it to AI, the data used, the expected result and how it will be measured. The managers who own them rule on each one, keep, frame or ban, and you leave with the roadmap and the company's AI doctrine, approved by leadership: its rules and safeguards, what others call an AI policy. The choice comes before the tools: a team that knows what it hands over to AI, and why, gets more from a few tools than a well-equipped team with no direction.
Questions clients ask me.
“When a client asks ChatGPT which bag to give, what does it say about us?”
We measure it before touching anything: that is what the Agentic Footprint Audit does, in one to two weeks. Then we write what the assistants need to read on your site: the material, the origin, the making, what justifies the price.
“Our best clients of tomorrow already buy from us, and we do not recognise them. Can AI spot them?”
It flags, in your sales, the clients whose purchases change in nature; the advisor chooses whom to call, and when.
“How do we give every advisor the history of our models, at the moment of the sale?”
An assistant answers the advisor from your product sheets and your archives; the advisor speaks to the client in their own words.
“Can we produce our content with AI without it showing?”
Variations, translations and product descriptions, yes, proofread before publication. Creation stays with the artistic direction, and “made with AI” is never a selling point for an object.
“Every complaint starts from scratch. Can the replies be prepared without taking the decision away from the advisor?”
An agent, an AI that receives an objective and chooses its own steps, drafts the reply from the client's history and your terms of sale; the advisor decides and signs.
“We make too many of some references and not enough of others. Does AI forecast better than we do?”
Yes, if it also learns from your waiting lists and your advisors' notebooks, not only from past sales: a piece kept scarce says nothing about what clients want.
The first question has its own engagement: the Agentic Footprint Audit measures what ChatGPT, Claude, Copilot or Perplexity tell your clients about your brand, and says what your site needs to change. One to two weeks, one report.
How an engagement works.
- Scope.
Management first, in one-to-one interviews: what the company protects above all, what it refuses to hand to AI, who answers for it. Then a tour of the business units and entities, with the managers who own each use case and the people who do the work: what costs them time, what loses sales, what already works. They bring their expertise, I bring the method.
- Specify.
Each use case is described in a one-page brief: the task, the team concerned, the data required, the expected result and the measure that will tell whether it keeps its promise, the person who decides and the safeguard that stops the AI. It joins the backlog, the list of cases to be handled.
- Prioritise.
The cases in the backlog are ranked with decision tools that weight several criteria: what the case brings in, what it requires, and whether it strengthens or damages what your clients come to you for. A case that frees up time without saying what that time will be used for does not move up the list. Each manager rules on the cases they own, keep, frame or ban, and sets their launch order. The cases set aside go into the register of refused cases, with the reason: your in-house case law.
At the end of the assignment.
- Your list of use cases, one brief per case.
- The state of the data each one needs.
- The prioritised backlog and the roadmap: the launch order, the tools to plan for, the teams to train.
- For each case, the measure set before launch, to be read again afterwards.
- Formalisation of the company's AI doctrine: the rules, the safeguards and the refused cases drawn from these decisions, approved by management.
Three months on.
- You read again, for each case launched, the measure set before launch: what it delivered, what it did not.
- Dropped cases leave the backlog, with the reason written in one line.
- The roadmap is brought up to date: the next case, the tools to set up, the teams to train.
Fees.
On quotation, tailored: every engagement is calibrated to the ambition, means, resources and scope of the project.
Already chosen your use cases? Go straight to AI solutions or training.