AI today is a bit like Excel thirty years ago. Back then, adding up three numbers in cells wasn’t enough to become more productive: the real gain came when everyone adapted the tool to their job, their data, their tasks. It is the same with AI. Typing a few queries into a consumer tool isn’t “doing AI” yet.

Yet there is a whole part of the economy we could call AI’s forgotten: the traditional fabric of small and mid-size companies, from a handful of employees to two hundred, often far from the big cities. The local factory, the nonprofit, the notary’s office, the family business: everything that is neither a digital services firm nor a startup. These organizations aren’t behind for lack of interest or skills. Nobody has simply shown them the first step.

Adopting AI in a small business: beware the “super Google” trap

On the ground, the same scene keeps repeating itself. AI is used as an improved search engine: you ask a question, you get an answer, you move on. Everyone tinkers in their own corner, with no shared framework: a kind of shadow AI, invisible and unmanaged. People stick to what the media popularized, “prompting”. And they miss what AI can really do when it takes over an entire process.

That is the heart of the misunderstanding: confusing “asking a question” with “delegating a task”. Asking an AI to rephrase an email is useful, but it is a one-off, and invisible in the numbers. Productivity doesn’t show up there. It shows up when AI takes over a recurring task: sorting customer requests every morning, preparing a quote template, summarizing meeting notes, screening applications. Many business owners hear about “AI agents” and think: “that’s not for us”. That is a mistake. As soon as someone spends their day in front of a computer, it concerns them. To understand what these terms really cover, we tell them apart in our article on the difference between an AI skill, an AI agent and an AI assistant.

Finally, there is a quieter obstacle: companies that tested AI two years ago often came away convinced that “it makes too many mistakes”. This lack of trust, inherited from less mature tools, still weighs today.

Why do smaller businesses use AI less? What INSEE says

The gap isn’t an impression: it is measured, and it is widening. INSEE, the French national statistics institute, attributes it first to the fixed costs of adoption. Putting AI in place requires an upfront investment (time, skills, scoping) that weighs proportionally much more on a small organization. The result: French companies with fewer than 50 employees use AI almost four times less than large ones, 15% versus 58%. And the gap isn’t closing: between 2023 and 2025, it grew from 16 to 43 points (INSEE study published in 2026, 2025 survey of ICT use in companies). The further AI advances, the further ahead the large companies get.

Behind this overall finding, the study identifies specific obstacles. And the first one is neither price nor technology.

“We don’t see the point”: the first obstacle to adoption

Among the companies that don’t use AI, the most cited reason is a lack of perceived usefulness, given by 71% of them, especially the smallest ones (INSEE, 2026). Here is the initial misunderstanding again: never having gone beyond a one-off trial, they don’t see what AI would change in their daily work. The obstacle isn’t the tool: it is the lack of a concrete use case, tied to a real task of the job.

Lack of AI expertise: a blocker that persists after the start

The skills gap isn’t specific to smaller businesses, but it hits them hard: in a small organization, nobody has “AI” in their job description. And it doesn’t only block the start, it persists afterwards, whatever the size of the company: 53% of the companies already using AI say a lack of expertise holds back their use of it (INSEE, 2026). Without an identified owner, use stays makeshift and isolated, and never scales up.

“Where does my data go? Am I allowed to do this?” These questions weigh heavily. Among the companies that use AI, 43% say they are held back by concerns about data protection and privacy, and 42% by a lack of legal clarity; cost only comes next (33%) (INSEE, 2026). The fear isn’t irrational. But without a simple framework, set once and for all, it often becomes an excuse for doing nothing.

It is not a technology problem, it is an adoption problem

The real work isn’t IT. The tools are there, accessible and often inexpensive. What is missing is adoption: bringing AI into actual work routines. That is why, in a small business, AI is as much an HR topic as a technical one. Jobs will evolve, tasks will be reshaped. Supporting that change is, in the end, the real work.

What your business doesn’t need

No consulting firm, no data team, no complete AI strategy before getting started. A small organization that waits until everything is planned will never launch anything. Its strength is agility: test a use on Monday, see the result on Friday. A big transformation plan isn’t a prerequisite, it is often an excuse for launching nothing.

Finding the right use cases: start from daily work

The right use cases aren’t decreed from above, they are spotted on the ground. In ideation workshops with the teams, one question is enough: which tasks come back several times a month and take a lot of time? Those are the best candidates. And AI isn’t there to replace the employee on an entire assignment: most often, you break a task down into several steps, and AI takes on only one or two of them, the most repetitive ones. The human keeps the rest.

Adapt to the job, measure, spread

A good use is tailored to a specific job: that is what we call “creating a skill”. You set it up, you measure the time saved (the return on investment of AI is known to be hard to quantify, all the more reason to start from concrete, measurable tasks), then you spread what works instead of letting it die in a corner. This approach, which starts from frontline pain points rather than from the top, is the one we detail in our article “Rolling out AI: why starting on the front line is the only method that works”.

Conclusion

AI’s forgotten aren’t latecomers. They are organizations that nobody has shown the first step, and that wrongly believe a big transformation plan is needed to take it. The gap isn’t closed with a budget or with a consultant’s strategy: it is closed with one delegated task, a clear framework and a use that gets shared.

You don’t need a big plan. You need a first step.

FAQ: your questions about AI in small and mid-size businesses

Do small businesses use AI?

Far less than large companies: French companies with fewer than 50 employees use AI almost four times less than large ones, 15% versus 58%, and the gap grew from 16 to 43 points between 2023 and 2025 (INSEE, 2025 survey of ICT use in companies). INSEE attributes it first to the fixed costs of adoption (time, skills, scoping), which weigh much more on a small organization.

What holds small businesses back from AI?

The first barrier is a lack of perceived usefulness, cited by 71% of the companies that don’t use AI (INSEE, 2026): without a concrete use case, they don’t see what it would change. Next come a lack of expertise, which holds back 53% of the companies already using it, then concerns about data protection (43%) and a lack of legal clarity (42%). Cost only comes after that (33%).

Where should a small business start with AI?

No need for a consulting firm, a data team or a complete AI strategy. With the teams, spot the tasks that come back several times a month and take a lot of time, give AI one or two steps, the most repetitive ones, measure the time saved, then spread what works. Productivity shows up when AI takes over a recurring task, not when you ask it a question now and then.