The method
You decide on the next one from what the last one produced. Nobody signs a six-month commitment against a promise.
€10,000 excl. VAT · fixed fee, one area of the business
Know exactly where generative AI would save you time, what it costs, and what to build first.
Three half-days of interviews with your teams, and read access to your tools.
Eligible for 40% funding by Bpifrance (Diag Data IA) for French SMEs of 10 to 2,000 staff: the company files the request, subject to approval.
from €18,000 excl. VAT · fixed fee, one system in production
A first system genuinely running, used by your teams, with its gain measured.
One person from the business for an hour a week. No training needed first.
on request · depending on what you take over
The system scales, it is monitored, and you take the controls at your own pace.
How much you take over is decided at the end of the pilot, not at the start.
Other ways to start
Not everyone needs to start with a plan. These two formats cover the two most common cases.
fixed fee, one prototype on your data
A prototype running on your own data within a week, so you decide on evidence, not on promises.
You receive
A one-hour scoping call, a data extract, and a demonstration at the end of the week.
Talk it through for 30 minutesthree-month initial commitment
Data and AI leadership over time, without hiring a full-time executive: roadmap, trade-offs, and projects driven all the way to production.
You receive
An executive sponsor, and a monthly review with the leadership team.
All about the fractional format Talk it through for 30 minutesThe journey
On the left, what is usually proposed to a mid-sized company getting started. On the right, what happens when you build first.
The route usually proposed
The route here
An awareness seminar, and someone is named AI lead.
Three half-days with your teams, measuring where the hours actually go, task by task.
The list of use cases goes round the steering committee. Nothing has been costed yet.
You have the plan: levers costed in hours freed, the data note, and the first system to build. You can stop there.
A specification is drafted, suppliers are shortlisted, quotes are compared.
The first system is running. Connected to your tools, used every day, with nobody sent on a course first.
The proof of concept is demonstrated to the committee. It stays in the room.
The gain is measured. On the acquisition build: two inbound enquiries a day, up from one every three days.
A training plan is considered. The budget goes back for arbitration.
Your teams own the system. We extend to the next area, with what the first one taught us.
The left-hand column is not absurd — it is cautious. It simply costs six months before the first gain.
Against the grain
Not for the sake of it: this is what makes a build pay off in a matter of weeks rather than eighteen months.
Train the teams first.
A course dates within months and never says what to automate at your company. Your teams learn on the system that already runs.
List every use case.
A survey takes six months and frees no time. I measure one area and rank the levers by effort and by gain.
A proof of concept to convince.
A demo does not survive the committee. What we build is wired into your tools and used the next day.
Step 01
Eight days spread over three to four weeks, so your teams are never blocked. The deliverable stands up without me: you can hand it to your board or to another supplier.
Interviews with the people who do the work, not only with the people who manage it. We reconstruct how a real week runs, noting every re-keying, every wait and every round trip.
Read access to your tools: CRM, mailboxes, management software, analytics. We count what can be counted. An irritation that adds up to two hours a month is not a build.
Each lever is estimated in hours freed and in effort to build. Alongside that, we decide for each one what data would leave, where it would be processed, and what that implies.
A ninety-minute session with the people concerned, a document that stands up without me in the room, and a clear recommendation on what to build first.
Step 02
One system, put into production and used. Not a mock-up, and not a proof of concept asleep in a test environment.
We write down what will count as success and how it will be measured. Without that sentence, a build can neither be stopped nor judged.
One demonstration a week, on the real system. You watch the build advance and you correct course while there is still time to.
The system goes to the people who will use it. That is when the real problems surface, and that is why this phase sits inside the fixed fee.
We compare the before figure with the after figure. Then, on that evidence, we decide whether to industrialise or to stop there.
The exchange
The load on the client side is what sinks most projects. So it is written down here, and by design it is light.
| What you provide | What you receive |
|---|---|
| Three half-days of interviews with your teams | A map of the time actually spent on the area audited |
| Read access to your tools | Each lever costed in hours freed, ranked by effort and by gain |
| One person from the business, an hour a week during the build | The system built, deployed and connected to the tools you already use |
| A decision at the end of each step | The before and after measurement, and the documentation to take over |
Step 03
A supplier you cannot do without is a risk, not a partner. Independence is built up in stages, and how far you take it is decided at the end of the pilot build, once you know what the system really asks for.
During the build, your teams sit in on the demonstrations and ask questions. Nobody has anything to maintain yet.
The first changes are made together, in your own code repository, with the operating documentation to hand.
You evolve the system and I answer questions. Plenty of clients keep a light monitoring retainer at this point, because watching over automated processing is the first thing to get dropped.
The limits
An offer that never states its limits says nothing at all. Here are my four.
I do not sell licences, and I take no commission on any tool I recommend.
I do not start a build whose gain I cannot measure. If the gain will not be measurable, the build waits.
I do not run catalogue training courses. People come up to speed on your own tool, once it exists.
I do not keep hold of your system. The code sits in a repository you own, from day one.
The method
No, and that is deliberate. A company that has to assemble a file before it can start puts the whole thing off indefinitely. I need to talk to people and look at the tools as they are, not as they would have been tidied up for my visit.
It happens, and it is a useful result. You get the map of where the time goes and a costed demonstration that the subject lies elsewhere, which spares you a build that would have returned nothing. The fee is still due, and I tell you straight rather than selling you the next step.
The diagnostic interviews happen on site, because that is the part that deserves it. The build happens remotely, with a weekly demonstration. I travel at no extra charge within Hauts-de-France, and elsewhere in France with actual expenses added.
You decide. On your premises, on your cloud, or on mine for a transitional period. The choice turns mostly on how sensitive the data is and on whether you have someone in house able to run the system.
That is the stated aim. Handover is gradual: first you watch, then you make changes alongside me, then you lead and I stay in the background. Plenty of clients keep a light monitoring retainer, because watching over automated processing is the kind of job everyone forgets.
The diagnostic is a fixed fee because its scope is fixed. The pilot build depends on what has to be built, and the amount is set against that scope when it starts, once the diagnostic has said precisely what to build.
Next step
The online questionnaire takes five minutes and hands you a map of your levers, whether you take it further or not.
5 minutes, free, no sign-up
Diagnostic
Paid engagement8 days
€10,000 excl. VAT · fixed fee, one area of the business