AI adoption in SMBs: team resistance is not what leadership imagines
22% of business owners name team resistance as a barrier to AI. Yet training, when it exists, gets 89% satisfaction. The real problem lies elsewhere.
22% of SMB leaders cite their teams' resistance as a barrier to adopting artificial intelligence, according to the Bpifrance Le Lab study on French SMBs and mid-sized companies. The figure is real, but it deserves a careful reading. Ranked behind fear of confidential data misuse, at 33%, and lack of skills or training, at 26%, it is not the top obstacle encountered on the ground, contrary to what many leadership teams assume before launching their first project.
What I regularly see in the companies I work with looks less like resistance on principle than a poorly prepared rollout. An employee shown a new tool in a general meeting, with no demonstration on their own work, no dedicated time to try it, and no clear answer to "what does this actually change for me," has no reason to engage. Their silence or reluctance then gets read as resistance to change, when it is really a lack of support.
Training figures back up this reading. France Num programs dedicated to digital and AI uses show 89% satisfaction, including 46% of beneficiaries reporting they are very satisfied. Few public support schemes reach that level. It suggests that supposed resistance largely disappears once structured support exists, which reframes the question, it is no longer how to overcome team resistance, but how to organize training that produces this result.
A sector study on SMB digital transformation adds a useful nuance, in 73% of cases, it is the leader personally who drives the transformation. This figure, often cited to underline the importance of leadership, hides something useful, a leader can drive the decision to adopt AI without driving its actual rollout with each team. A convinced leadership team that fully outsources training to an external provider, with no internal relay or individual follow-up, often gets disappointing results despite a sincere initial conviction.
The format of training matters as much as its existence. A generalist group session on "AI in the workplace," given to the whole staff on the same day, rarely produces lasting adoption. What works better, in field experience, is targeted support by job function, with a use case immediately applicable to the trained person's role, followed by supervised practice time rather than a simple presentation. A salesperson who leaves with an automated call-summary template they will use the next day retains and adopts more than an employee who sat through a general conference on AI uses two weeks earlier.
Tool choice also affects adoption speed, especially for less technical roles. A tool like Comet, designed to act directly on the web without writing a line of code, filling out a form, posting on social media, replying to a customer comment, lets a non-technical employee get a visible result on their very first use. That early win often matters more, for lasting adoption, than the technical sophistication of the tool chosen.
There is also a rarer but genuine form of resistance that has nothing to do with lack of training. Some employees, often the most experienced in their field, fear that AI will undermine the value of expertise built over years. That worry is not addressed with another technical demonstration, it calls for an honest conversation about what the tool actually replaces, and what it does not. An SMB that avoids this topic, implicitly promising that nothing changes for anyone, builds a fragile trust that cracks at the first role adjustment.
Telling real resistance apart from a lack of support means watching the questions asked, not just the objections voiced. An employee who questions result reliability, actual time saved, or the tool's limits is trying to understand, not being hostile. Silence, on the other hand, or minimal use with no questions at all, often signals deeper disengagement than a spoken objection.
This is exactly what change management addresses, at the end of the process: planning job-specific training and individual follow-up on first uses, rather than treating resistance as a general communication problem to solve in a single meeting. A five-business-day diagnostic identifies, department by department, what is actually slowing adoption in your company.
Frequently asked questions
How many SMB leaders cite team resistance as a barrier to AI?
22% of leaders surveyed by Bpifrance Le Lab cite team resistance among the main barriers to AI adoption, behind fear of confidential data misuse (33%) and lack of skills (26%).
Is team resistance really the top obstacle to AI adoption?
No. Available data shows that lack of training and unclear use cases weigh more heavily than resistance itself. What looks like resistance is often the symptom of a rollout with no support, not a rejection on principle.
Does AI training actually work to remove these barriers?
Results are clearly positive where it exists. France Num training programs on digital and AI uses show 89% satisfaction, including 46% very satisfied, a level rarely reached by other public support schemes.
Should an entire company be trained on AI at the same time?
No, a generalist training session spread across all departments rarely produces good results. Targeting first the employees whose job will be most transformed, with a concrete, immediately applicable use case, delivers better adoption rates than a theoretical group session.
Why is the leader's role so decisive in AI adoption at SMBs?
In 73% of cases, the business leader personally drives the company's digital transformation, according to Bpifrance Le Lab. Their personal involvement, beyond stated intent, largely determines how fast teams actually adopt AI.
How do you tell real resistance apart from a simple lack of support?
An employee who asks precise questions about their own use case, result reliability, or time saved is not resistant, they are trying to understand. Silence, or minimal use with no questions at all, is often a more worrying signal than a spoken objection.
Do no-code tools make adoption easier for non-technical teams?
Yes, a tool like Comet, built to browse and act on the web without writing code, lets non-technical employees take ownership of a first concrete use case, filling in a form, publishing content, replying to a comment, without depending on an IT department.
When should change management be addressed in an AI rollout?
The Change Management step, at the end of the process, plans job-specific training and individual follow-up on first uses. This step, more than the choice of tool, determines whether adoption holds over time.