AI grants for SMEs do exist in France in 2026, but they mostly fund advice: Bpifrance's Diag Data IA is 40% subsidised, the Accélérateur IA 46%. The tool itself, a subscription or software, is rarely covered. For a food and drink SME, the right scheme depends on its size and on whether it already knows what it wants to automate.
This article sorts through national schemes, the sector training fund (OPCO) and regional programmes, with verified amounts and what you actually pay. It is written for managers of French food and drink SMEs who want to move forward on AI without filing an application for nothing.
Key takeaway
the Diag Data IA costs €10,000 excluding VAT, of which €6,000 remains payable by you since funding rose to 40% on 17 June 2026. The Accélérateur IA, reserved for companies with more than 50 employees and €8m in revenue, is 46% funded over 18 months. One year after its launch, the Osez l'IA plan had reached 35,000 companies but started only 70 diagnostics, against a target of 1,000. No national scheme directly funds a subscription to an AI agent.
AI grants for SMEs: the schemes available in late 2026
Most funding goes through the national Osez l'IA plan, run by the French Directorate-General for Enterprise (DGE) and operated by Bpifrance (entreprises.gouv.fr). It targets 80% of SMEs and mid-sized companies using AI by 2030. Around it, the sector's training fund and some regions complete the picture.
| Scheme | Who for | What is funded | Funding rate | Indicative cost to you |
|---|---|---|---|---|
| Diag Data IA (Bpifrance) | SMEs and mid-caps with 10 to 2,000 employees | A diagnostic of about 8 expert days, over 3 months at most | 40% | €6,000 out of €10,000 excl. VAT |
| Accélérateur IA (Bpifrance) | Over 50 employees and over €8m revenue | An 18-month programme: diagnostic, modules, group training | 46% | About €39,000 out of €71,800 excl. VAT |
| IA Booster France 2030 | All SMEs and mid-caps | Self-assessment, training and implementation advice | Co-funded | Depends on the support |
| Boost IA loan (Bpifrance) | Small businesses up to 49 employees | A loan of €5,000 to €100,000, unsecured | Loan, not a grant | Repayable |
| OPCO (OCAPIAT for food) | Employees in the sector | Training | Depends on scheme and caps | Variable |
| Regional aid | Depends on the region | Variable, sometimes adoption costs | Example: 50 to 70% in Bourgogne-Franche-Comté | Variable |
Diag Data IA Bpifrance and Accélérateur IA Bpifrance
The Diag Data IA amounts come from the scheme's listing, those for the Accélérateur IA from the Infogreffe listing, which announces applications opening on 30 November 2026. The IA Booster France 2030 programme plans 2,000 supported companies over 2026 and 2027 (DGE). The Diag Data IA funds an assessment and a roadmap, not the building of solutions. The Accélérateur IA goes further, but over a long period and for companies that are already well structured.
Boost IA loan: a loan, not a grant
The terms of the Boost IA loan differ between published listings: check them on Bpifrance's website before relying on them.
OCAPIAT: AI training through the sector fund
OCAPIAT, the training fund (OPCO) for the French food industry, has already supported introductory AI courses, for example in Bourgogne-Franche-Comté. Caps depend on your size and on the scheme used: only your OCAPIAT adviser can confirm them.
Regional aid
The clearest example is the Impulse Intelligence Artificielle programme in Bourgogne-Franche-Comté, which explicitly targets the food industry and covers 50 to 70% of eligible costs. Other regions run broader digital programmes, whose names and rates change often. Be wary of lists that claim dozens of schemes: many are out of date.
Osez l'IA plan: what AI funding does not cover
The plan's review, published on 3 September 2026, is telling. 615 AI ambassadors mobilised, 35,000 companies reached, 88 solution providers listed, but 70 Diag Data IA started against a target of 1,000 (Bpifrance). Managers are interested. They do not take the step of the diagnostic.
One explanation comes down to a calculation. For a 40-employee charcuterie maker, €6,000 for a roadmap is real money, spent before any gain. And at the end of the diagnostic, nothing is running yet: implementation still has to be funded.
Three costs almost always fall outside public funding:
- The tool itself. No national scheme covers a software subscription. Only a few regional schemes cover adoption costs, under their own eligibility rules.
- Your team's time. The sales administration manager or management controller who attends the workshops is not reimbursed. This is often the heaviest cost for an SME.
- Cleaning up the data. If your bills of materials are incomplete or your rebates live in a spreadsheet outside the ERP, they will need fixing. We cover this work in our article on ERP data quality in the food industry.
Funding lowers the cost of a step. It does not tell you whether that step is the right one.
Which AI support fits the size of your food and drink SME
1. Under 50 employees: training and a first use case first
A 25-employee cheese maker is not eligible for the Accélérateur IA, and the Diag Data IA weighs heavily on its budget. For this company, the logical order is reversed: train the team in the basics through the OPCO, then put a first, precise use case into service on a process that costs time every week (order entry, customer follow-ups, purchase price tracking).
If financing is needed, the Boost IA loan and regional aid are the first options to explore. Their advantage is that they can cover adoption, not only the study.
Starting small is not giving up. It is the surest way to find out what AI earns you.
2. Over 50 employees and €8m revenue: the Accélérateur IA makes sense
For a mid-sized company or large SME that wants to structure AI across several departments, the Accélérateur IA is a real programme: 18 months, a 360-degree diagnostic, group training and exchanges with other managers. The remaining cost, around €39,000 excluding VAT, is justified if the company can run several projects in parallel.
The risk is spending 18 months preparing for AI without any use case running. From day one, require a first use case in production within the first three months.
A long programme is worth it when it speeds up real projects, not when it replaces them.
3. You already know what you want to automate: go straight to testing
Many managers do not need a diagnostic to know where they lose time. The sales administration manager who re-keys 80 orders a day, the sales rep who finds out too late that a customer has stopped ordering, the management controller who rebuilds margins by hand every month: the use case is already there.
In this situation, a subsidised diagnostic delays the first result by several months. It is better to test an agent on that process, with your real data, and judge on the evidence.
The best diagnostic is still a use case running on your own data.
Subsidised diagnostic or direct test: do the maths
| Criterion | Subsidised Diag Data IA | Direct test of a specialised AI agent |
|---|---|---|
| Cost to you | €6,000 excl. VAT | One-off set-up fee, then €149 excl. VAT per month per user at Agrolytics |
| Time to a first result | Up to 3 months of diagnostic, then implementation | About one week with our customers |
| What you get | A roadmap and prioritised use cases | An agent in service on one process |
| What remains to be done | Fund and run the implementation | Extend or stop depending on results |
| Suited if | You do not know where to start | You have identified the process that costs you |
The two approaches are not opposed. A diagnostic can identify the right use case, then a specialised vendor can put it into service. But if the process to tackle is already obvious, the diagnostic money is often better spent on the first month of use. At a €12m charcuterie maker using our sales administration agent, set-up took one week, on the company's real data.
To put a figure on what the use case earns, apply the method in our article on calculating the ROI of sales AI.
What an AI agent does not do in place of a support programme
It does not define your AI strategy. An agent handles one specific process: order entry, follow-ups, margin tracking. If you want to rethink how several departments work, a broader programme such as the Accélérateur IA has its place.
It does not fix missing data on its own. If costs are not allocated by customer or if prices live outside the ERP, the agent will make it visible, but your teams still have to do that work.
It does not prepare your funding application. Dealing with Bpifrance, the region or the OPCO remains your job. On the other hand, a first use case with figures makes any application stronger.
AI funding for food SMEs: the traps to avoid
Signing before applying. The general rule for public aid is the incentive effect: the application must come before any commitment. A quote signed before you file can be enough to make you ineligible.
Trusting a rate read on an unofficial website. The Diag Data IA went from 25% to 40% in June 2026, and many pages still show the old rate, or figures for programmes that no longer exist. Only Bpifrance, the DGE, your region and your OPCO are authoritative.
Confusing the provider of the study with the provider of the solution. Whoever carries out the diagnostic may have an interest in recommending the project they will then bill for. Ask for use cases described independently of any solution, then compare offers, including those of specialised vendors in your sector. Our article on choosing an AI vendor as a food manufacturer sets out the criteria.
Letting the funding choose the project. A scheme funded at 46% is still expensive if it tackles a secondary problem. Start from the process that costs you most, then look for the funding that applies to it.
Where to start when funding your AI project
List the three processes that cost you the most time. Order entry, follow-ups, margin calculation, traceability: put a number on the hours spent each week. That figure will guide the choice between a diagnostic and a test, and it will be useful in any funding application.
Call your OCAPIAT adviser and your region before any quote. They are the two contacts who know the conditions that actually apply to your company. An Osez l'IA ambassador in your area can also point you in the right direction.
Test a use case before funding a study. If one process clearly stands out, see what an agent does with your own data. Within a few weeks you will know whether AI pays off, and any funding application will rest on a result rather than an assumption.
Frequently asked questions about AI grants for SMEs
Can a subscription to an AI agent be subsidised?
Not through the national schemes of the Osez l'IA plan, which fund diagnostics and support programmes. Some regions cover AI adoption costs under their own eligibility rules, and Bpifrance's Boost IA loan can finance an adoption project without being a grant. Check with your region before committing the expense.
Can the Diag Data IA be combined with regional aid?
Combining public aid is generally possible, within the European caps that apply to SMEs and provided the two schemes do not fund the same expense. Rules vary between schemes. Ask Bpifrance and your region at the same time, describing precisely what each scheme is meant to cover.
Does OCAPIAT fund artificial intelligence training?
OCAPIAT, the training fund for French agricultural and food companies, funds training for employees in the sector and has already supported introductory AI courses. Amounts depend on company size, the scheme used and the training provider. Your OCAPIAT adviser will give you the applicable caps before you enrol.
How long does it take to obtain AI funding?
Allow several weeks between first contact and approval, longer for regional aid that goes before a committee. The Diag Data IA then lasts up to three months, the Accélérateur IA eighteen. If your timetable is tight, for example before a sales season, a direct test gives a first result sooner.
Do you need a diagnostic before setting up an AI agent?
No, not if the process to tackle is clearly identified and the data needed is in your ERP. A diagnostic is useful when you do not know where to start or when several departments are involved. In other cases, a first use case in service teaches you more than a study.
Going further
To find out which AI use cases pay off most in a food and drink SME, read our guide to AI agents in the food industry. It covers six concrete cases, from sales to production.
Unsure whether to file an application or test directly? In 30 minutes, we look together at the process that costs you most and what an agent could change there, before you spend a single euro. Book a call.
Ready to take control of your data?
Book a 30-minute demo and see what Agrolytics can do for you.
Book a demo

