If you run an SME in France, you have probably heard about the Osez l’IA plan and its Diag Data IA. The scheme is genuinely useful: it gives you a structured diagnosis and a prioritised roadmap, partly funded by the State. But there is one thing it does not do, and very few people say it clearly: it does not build anything. Once the report is delivered, the roadmap is yours, and the work of turning it into a tool that actually runs is a separate matter. This article explains the scheme in plain terms, and what comes next.
What the Osez l’IA plan and the Diag Data IA actually are
Osez l’IA is a national plan launched in July 2025, operated by Bpifrance and co-funded by France 2030. Its goal is to spread AI adoption among French SMEs and mid-caps, and it is built around three pillars: raising awareness, training, and hands-on support. The Diag Data IA is the diagnostic tool inside that plan.
In practice, it means eight days of work by a Bpifrance-approved data and AI expert, spread over a maximum of three months. The expert takes stock of your technical and operational situation, identifies concrete AI use cases for your business, prioritises them by value, and hands you a roadmap.
- Total cost: 10,000 EUR excl. tax, with 40 percent covered, leaving roughly 6,000 EUR excl. tax for the company
- Eligibility: SMEs and mid-caps from 10 to 2,000 employees, registered in France, at least 1 million EUR in revenue, more than one year of existence
- Duration: 8 expert days over a maximum of 3 months
- Application: through the Bpifrance diagnostic platform, choosing an approved expert
One practical note: the figures for this scheme have changed several times since launch, and many articles online still quote outdated numbers. Always check the current terms on the official Bpifrance page before you budget anything.
What the scheme funds, and what it does not
This is the part worth reading twice. The Diag Data IA itself covers the diagnosis and the consulting: the audit, the use case identification, the prioritisation, the roadmap. It does not cover the build. Not the technical integration, not the setup, not the connection to your existing systems. At the end of the eight days, you leave with a document, not a tool.
In other words, at the end of the eight days you own a document, not a tool. That document is valuable: it tells you where to start and why. But nothing in your business has changed yet. The gap between a good roadmap and a system that runs every day is exactly where most AI projects stall in an SME.
A good roadmap is not a working tool. Everything is decided in the gap between the two.
One nuance is worth adding, because it is often missed. The Diag Data IA is not the only scheme under the Osez l’IA umbrella. IA Booster France 2030, a broader programme also run by Bpifrance, includes a fourth mission: support for implementation, co-funded at 50 percent, for a service worth up to 60,000 EUR excluding tax. Part of the deployment can therefore be subsidised. Two things to keep in mind: this mission is a consulting engagement designed to kick off the deployment, not an open budget for the full build, and it runs through the network of providers referenced by Bpifrance for that phase. If capturing that subsidy is your priority, pick a provider from that network. If you value speed, a specific technical skill set, or your project runs past the funded ceiling, an external execution partner is the more direct route.
After the diagnosis: the questions SMEs actually ask
Once the roadmap is on the table, the questions become very concrete, and they are rarely answered in the report itself.
- How long does the first use case actually take to build, and who builds it?
- What does the build cost, now that it is no longer subsidised?
- Which tools do we buy, and which do we build? What do we do with the systems we already have?
- Who owns and maintains the result once it is live?
- How do we start small enough to prove value, without committing to a large programme?
Our answer to the last one is always the same: start with a single use case, the one that is frequent, rule-based and painful, and build it end to end. Prove the gain in weeks, not quarters, then extend. A roadmap with twelve prioritised use cases is not a mandate to build twelve things at once.
From the report to a tool that runs
To be clear about where we stand: Data Solutions Agency is not a Bpifrance-approved expert and does not deliver the funded Diag Data IA. We work downstream, as the execution partner, once the diagnosis is done. If you have not done the diagnosis yet and you are eligible, use it. It is a good, structured way to get your bearings, and it is partly paid for.
What we do is the part that comes after: taking a prioritised use case and turning it into something that works in production. For a Swiss fresh-produce wholesaler, that meant taking a painful daily process, chefs sending orders as WhatsApp voice notes that a team retyped by hand every night, and building an app that transcribes the voice note, extracts the order with AI, and matches it against a 1,600-item catalogue to produce a structured purchase order. A working proof of concept was delivered in a few days. That is the difference between a use case on a slide and a use case in production.
In short
The Diag Data IA is a good starting point, and if you are eligible, the subsidised diagnosis is worth taking. Just go in with clear eyes: you will come out with a roadmap, not a running system, and the build is on you. Plan for that phase from the start, budget it, and pick one use case to prove the value before you scale.
You have a roadmap and you are wondering what the first build actually involves? Book a 20-minute call. We will look at your prioritised use cases, tell you which one to start with, and give you a realistic timeline and a fixed price for it.