Adoption is accelerating
The France Num 2026 barometer, published on 28 September by the French Directorate General for Enterprise, records a spectacular leap: 40% of very small, small and medium-sized businesses now use at least one artificial intelligence solution, 14 points more than in 2025. Among SMEs, the share reaches 53%.
The most common uses are text, voice or image generation (34%) and conversational assistants (24%). And the effect is felt: 68% of companies using AI believe it has a positive impact on their business, a share that rises to 85% among those using a paid solution.
The figure that stands out: 19%
Only 19% of these businesses pay for an AI solution (32% of SMEs). In other words, a large share of uses rely on free tools, opened individually by employees, with no framework and no control over the data entrusted to them. That is precisely the territory of shadow AI.
The barometer also highlights a marked gap between sectors: 72% of businesses use AI in the digital sector, against 30% in construction and 18% in agriculture.
Uses still very focused on text
The breakdown of uses is telling. Content generation and conversational assistants dominate, while data analysis, forecasting or resource optimisation still concern only 9% of businesses, even if progress is real (+4 points). In other words, most small businesses use AI as a writing tool, rarely as a tool connected to their own data.
That is where the next step lies: an AI that knows your procedures, contracts, product sheets and client history answers questions no general-purpose tool can handle. But that is also where the data question becomes critical: you do not connect your document base to just any service.
Business leaders’ enthusiasm is fading
The barometer picks up another, more discreet signal: the share of business leaders who see a real benefit in digital tools is falling. Tools multiply, subscriptions pile up, and the value is not always there. For AI, this argues for tighter projects: few uses, well chosen and measured, and one tool that covers all needs rather than a collection of applications.
- Choose two or three use cases with high volume rather than ten experiments.
- Measure the time saved from the first weeks.
- Train the teams, because the gap between spontaneous and effective use comes down to method first.
Our reading
- AI is already in the company: the question is no longer whether to go, but how to govern what already exists.
- Moving to paid tools improves perception: a tool that is chosen, deployed and supported creates more value than improvised use.
- AI remains little planned: it is a project for only 18% of these businesses, far behind e-invoicing. Support makes the difference.
From spontaneous to controlled use
As an Activateur France Num, Artemia supports small and medium-sized businesses through this step: identifying useful uses, securing them with AI deployed on their premises, and training teams. An approach that often starts with an AI audit.
Source: France Num 2026 barometer, French Directorate General for Enterprise, published on 28 September 2026.
