AI Productivity in SMEs: 2026 Key Figures
Official figures say two things at once: AI is already in most SMEs with 20 or more employees (67% use generative AI in early 2026, according to the Banque de France), yet only 31% of those users report a productivity gain. In other words, adoption is moving faster than measured results.
This changes the question to ask. It is no longer whether your SME should get started, but why some companies see a visible gain and others do not. This article draws on the most reliable public data (Insee, Banque de France), explains why the numbers do not contradict each other, and offers a simple method to measure your own gain. For the broader picture, see our complete guide to AI in business for SMEs.
What is the AI adoption rate among French companies?
Between 18% and 67% depending on the source, because the two major studies do not measure the same thing.
Insee (Insee Première no. 2120, July 2026) counts companies with 10 or more employees that use at least one AI technology: 18% in 2025, up from 10% in 2024 and 6% in 2023. The rate has tripled in two years. The European Union average was 20% in 2025.
The Banque de France (Bulletin no. 266, 3 September 2026) surveyed about 7,000 companies with 20 or more employees between 25 February and 3 April 2026, and focuses on generative AI: 67% say they use it. The scope is narrower (20 or more employees; manufacturing, construction and market services excluding retail), the period is more recent, and the definition of use is looser, since opening ChatGPT to reword an email counts.
The two figures are therefore compatible. The first measures AI technologies reported by the company in 2025, the second generative AI use in early 2026, often at individual level. If you can only quote one in a funding application or a management meeting, quote it with its exact scope.
Does adoption depend on company size?
Yes, clearly: the bigger the company, the more it uses AI, but the gap is narrower for generative AI.
Here is what the two sources say, by size band:
| Size band (employees) | Insee 2025: at least one AI technology | Banque de France 2026: generative AI |
|---|---|---|
| 10 to 49 | 15% | not covered |
| 20 to 49 | not covered | 64% |
| 50 to 249 (Insee) / 50 to 99 then 100 to 249 (BdF) | 31% | 68% then 75% |
| 250 and over | 58% | 73% to 84% depending on the band |
Two useful takeaways for an SME. First, generative AI has spread even to small structures: 64% of companies with 20 to 49 employees use it, according to the Banque de France. Second, more structured AI technologies (text analysis, speech recognition, automation) are growing much more slowly in small companies: 15% among those with 10 to 49 employees according to Insee, versus 58% for 250 and over. The gap has less to do with access to the tool than with how it is built into the work.
Does AI really improve productivity?
So far, a little under one in three user companies sees it: 31% of generative AI users report a productivity gain over the past three years, and 41% of non-generative AI users.
The Banque de France makes a point that summary articles often leave out: measured against all surveyed companies rather than users only, the generative AI figure falls to 22%. The same study notes that use remains "mostly limited or experimental".
Reported effects over the past three years:
- Productivity: a gain for 31% of generative AI users and 41% of non-generative AI users.
- Employment: a slightly negative net effect, 2% for generative AI and 3% for non-generative AI.
- Investment: a positive net effect of 15% (generative AI) and 22% (non-generative AI).
Looking ahead, 60% of companies that use or plan to use AI in 2026 expect a positive effect on productivity within three years, but only 17% expect a "significant" one. The optimism is real, and it is measured.
These figures do not show that AI saves no time. They show that an individual time saving (writing faster, summarizing a document) does not automatically turn into a gain that is visible at company level. We cover this in our article on calculating AI ROI in SMEs.
Why do gains remain limited?
Because AI is mostly used for support tasks, and the barriers companies report are about skills and use cases, not technology.
The Banque de France figures are telling. 74% of user companies report that their main use is in support functions, versus 14% for production. The top two areas of use are sales and customer relations (28.5%) and internal administration (27.5%). Among generative AI tools, text generation dominates: 85% of users.
On barriers, the difference between companies that do not use AI and those that do is instructive:
- Non-user companies: 49% cite the lack of use cases, 24% insufficient technical skills, 16% unreliable results, 9% cost.
- User companies: 33% cite insufficient technical skills, 29% the lack of use cases, 27% regulatory obstacles, 24% unreliable results, 17% cost.
Cost comes last in both groups. The real issue is skills and the choice of use cases. A tool bought without a precise usage scenario is unlikely to change anything in daily work. We listed the recurring pitfalls in our article on AI deployment mistakes in SMEs, and building team skills is covered in our AI change management guide.
How do you measure the productivity gain in your SME?
Measure one specific process before and after, with three indicators: time spent, error or rework rate, and volume handled.
Most SMEs answer "we save time" without being able to put a number on it. Here is a method that fits on one page:
- Pick a single repetitive process (for example writing quotes, sorting incoming requests, meeting minutes).
- Record the baseline over two weeks: average time per task, number of tasks, rework rate.
- Roll out the tool with a shared prompt or working template, not free-form use.
- Record the same indicators for two to four weeks.
- Calculate the net gain: time saved minus the time spent reviewing and correcting.
A purely illustrative calculation, with assumptions to replace with your own: 10 people saving 2 hours a week for 46 weeks, at a loaded hourly cost of 35 EUR, free up 32,200 EUR of time per year. If the tool, training and setup time cost 12,000 EUR, the theoretical net gain is 20,200 EUR. That figure only means something if the freed time is really reassigned to a useful task, and if review does not absorb half the gain.
To choose which indicators to track over time, see our article on AI KPIs in business. The term AI ROI is defined in our glossary.
Which uses deliver the clearest gains in SMEs?
Those that are repetitive, well bounded and easy to check: standard documents, summaries, sorting and filing, and automation of administrative tasks.
According to Banque de France data, the dominant uses are already there: text, customer relations, administration. Gains materialize when the use is framed. Three conditions come up repeatedly:
- A single, measurable use case rather than a general rollout with no objective.
- Generative AI backed by shared instructions, so quality does not depend on who is using it.
- Process automation when the task is stable, to get out of manual copy-pasting between tools.
Watch the flip side: without usage rules, your teams use AI on their own, with sensitive data. Our article on shadow AI covers this risk. And to go from a few tests to a plan, the 90-day AI roadmap offers a realistic framework.
Frequently asked questions
What share of French SMEs use AI in 2026?
According to the Banque de France, 67% of companies with 20 or more employees use generative AI in early 2026, although use is most often limited or experimental. According to Insee, 18% of companies with 10 or more employees used at least one AI technology in 2025.
Why do adoption figures vary so much?
The scopes differ: minimum company size (10 or 20 employees), year of measurement, and the definition of AI (AI technologies broadly for Insee, generative AI for the Banque de France). Always quote the scope with the figure.
What productivity gain can you expect from AI?
Official data give no average percentage gain. They show that 31% of generative AI users report a productivity gain. Yours will depend on the targeted process, the quality of use and review time. Measure it on one specific case.
Does AI destroy jobs in SMEs?
Over the past three years, the Banque de France reports a slightly negative net effect on employment: 2% for generative AI users and 3% for non-generative AI users. These are reported effects and should be read with caution.
What is the main barrier to adoption?
For companies that do not use AI, it is the lack of use cases (49%). For those that do, it is insufficient technical skills (33%). Cost is cited by 9% and 17% respectively.
Should you train teams to gain productivity?
The reported barriers are first about skills and use cases, which argues for training built around real tasks. It also answers the AI literacy obligation in Article 4 of the European AI regulation (EU Regulation 2024/1689).
How GrowthPerf helps
GrowthPerf is a Qualiopi-certified training organization specializing in AI, no-code and automation for SMEs and nonprofits. We start from your real processes: one use case, a baseline measurement, training for your teams on that specific case, then a measurement point a few weeks later. Our approach is described on the operational AI training page.
If you want to size the possible gain on one of your processes before investing, we can do it with you. Contact GrowthPerf for a first conversation, or go back to the complete guide to AI in business to place your priorities.