Why Employees Aren't Using AI, Even After the Training

Key takeaways

• Employees often aren't using AI even after the training because they still don't feel comfortable with it. AI educator Fey Pelleg says most workplace training is technical and done alone, so people rarely get to try something small and see that they can do it.
• The gap is large. In PwC's 2026 survey of nearly 50,000 workers, 11% of the largest group use generative AI every day, against 51% of the group furthest ahead.
• You can test a different approach with one team and one hour a week: the tool you already have, a task with no wrong answer, and everyone showing what they made.

When I talk to companies in Europe, I keep seeing the same picture. They have bought the AI licences, the team has done the AI literacy training, and some even have their first agents running. Then I look at the team, and using AI every day in the actual work is another story.

So why aren't employees using AI, even after the training? Is it the tool? Is it the course? Or are people worried about what happens once they get faster? My short answer, after talking to Fey Pelleg, is that training hands people the material and many of them still don't feel comfortable using it.

Fey is an AI educator and the founder of AI Goddesses, a global community where women learn AI through play and connection. Before that she was a software architect and team lead. I asked her about this in episode 4 of intelligence.md, the Ninja Partners podcast and YouTube series, recorded on 5 November 2026. We called the session "Feeling behind on AI? You might be learning it alone", and I feel behind too, even though I work with AI every day. You can watch it on YouTube or listen on Spotify.

How many employees actually use AI at work?#

Daily use is still a minority habit, and it's unevenly spread. PwC's Global Workforce Hopes and Fears Survey 2026, published on 29 September, asked nearly 50,000 workers worldwide. Daily use of generative AI rose from 14% to 22%. Among the 14% of workers PwC calls front-runners, 51% use it daily. In what it calls the core workforce, 56% of all workers, it's 11%.

Fewer than 40% of that core workforce also say they have access to learning and development, against nearly 80% of the front-runners. So the people who are behind are getting the least help. These are self-reported, global figures, and I wouldn't read them as a number about your company, but they do match what I see.

Why employees aren't using AI after the training#

AI adoption challenges usually get discussed as a data or integration problem. Fey looks at the people. "Even the teams that were trained in AI probably still don't feel comfortable with AI," she told me, "and still don't know how to approach new tasks with new tools."

She names three things that get in the way of most AI training for employees.

The training is technical. As far as she remembers, most work training covered a set of material and wasn't emotionally oriented. "There wouldn't be a playful moment."

People do it alone. It's something you complete on your own when you have time, or a Zoom call where one person talks. "Usually people are turning off their cameras, cooking their dinner in the meantime." Alone is also where asking gets hard. In a Harris Poll for INTOO, 42% of 1,158 US workers said it would be embarrassing to ask a coworker for help with new technology.

The first step is too big. The word Fey hears most at her meetups is overwhelm, and she says it comes from not knowing where to start. "Like anything new, you go too big, you fail fast."

I'd add a fourth from my own conversations. I've heard from people who saved real time with AI and then held back, because they are paid for 40 hours and want to show there is enough work for them.

A licence or a course won't solve any of this, because it's change management. Tools nobody uses are also one version of a wider problem: why companies spend more on AI than it returns.

What gets people to start using AI?#

Fey's answer is small, personal projects done with other people. "If I had to say it in one sentence, it's come play with me."

AI Goddesses has been running for about a year and a half, and hundreds of women have come to learn and play, some of them against AI. Every week she brings one activity, tested in advance and written up as a guide of up to five steps. One step is signing in, which she says takes the longest. The topics are personal, like an infographic in a funny style or a song for a friend who lives far away. There's no wrong friend and no wrong topic, so there is nothing to fail at, and at the end everyone shares what they made.

In the session she ran one live in Gemini. The whole prompt is "Create an image of educational poster explaining [topic]. style: [style]". She typed "quilt" as the style and got a poster that looks embroidered. "It might seem like a silly infographic," Fey said, "but it's actually a gateway to how we use AI." And once you have made something, "it's hard to ignore that you can do it."

How to get employees to use AI: what a team can copy#

Fey thinks her approach carries over: "I believe it's actually very, very similar." Steps 1 to 4 and step 6 come from what she described, step 5 from Gallup. The order and the wording are mine.

1. Use the tool you have already approved. Fey starts from whatever your team already uses. This matters in Europe, where GDPR and AI compliance mean not every tool is usable, especially the free ones. If people go looking on their own, you get shadow AI.
2. Pick a task with no wrong answer. An infographic for an internal document, or Fey's example: "take your latest task and make it a comic book."
3. Do it at the same time, in a group. People learn from seeing what others make. Three Microsoft researchers wrote in Harvard Business Review in March that AI adoption stalls because learning stays invisible.
4. Take away the result. "Keeping it light and playful will go much longer in the training than trying to actually make them perform or do a project while they're still overwhelmed by the tool."
5. Say out loud that trying is allowed. Gallup's February study of 23,717 US employees looked at organizations that make AI available. Of the people who strongly agree that their organization supports experimenting with AI tools, 72% use AI frequently. Of everyone else, 44% do. That's a correlation, and US only, but it's a cheap thing to change.
6. Give it a slot in the working week. When Fey was a team lead, she ran a weekly training during work hours instead of the stand-up.

I can't tell you whether this works the same way in a company. People choose to come to a meetup, and at work your manager is in the room. Fey's community is also for women, and I don't know how much changes in a mixed team. I have no data on it, but trying it only costs an hour.

How much time does AI training take?#

Fey recommends about two to four months if you meet once a week, adjusted to how much your team works with AI. This was my first objection in the session, because people don't have time right now. Her answer was to write down what you want to achieve and what it would save. She mentioned a student whose week-long task took two minutes live in class, and she was clear that this happens "in some cases".

For a team of ten, one hour a week for three months is about 130 hours. That's my rough sum, not hers. Put it next to what you pay for licences your team hardly opens, and you have the calculation Fey says everybody needs to do.

What if AI makes people's work run out?#

Fey doesn't soften this one. "Your work is going to run out anyway, because AI is here," she said, and the people working in your field later will be the ones who know how to use it. Then she explains why the fear might still be proved wrong. From her experience, "work doesn't seem to run out in companies." Her development team of about 10 people always had a backlog of at least a year.

If someone does finish early, she suggests adding projects that still benefit the company, and she doesn't think anyone should feel guilty about it. "You are paid for your talents."

For an owner, I think that leaves one job: answer this question before someone has to ask it. Tell your team what happens with the time they save.

Where to start#

Pick one team and one hour next week. Use the tool you have already approved, pick something nobody is graded on, and let people show each other what they made. Then see who asks for a second round.

Playing won't fix everything. In the same Gallup study, the widest gap was workflow fit: 88% frequent use among people who strongly agree that AI integrates well with their systems and processes, 55% among everyone else. If you're not sure whether the hold-up is the people, the tools or the way work is set up around them, that's what the AI Readiness Audit at Ninja Partners is for: a senior-led look at where AI pays back in your business, where it doesn't, and what it will actually cost. The full episode with Fey and the other conversations are on the intelligence.md series page.

Sources#

• The session: Fey's approach, her quotes and the live demo - intelligence.md Ep. 04, "Feeling Behind on AI? How to Learn AI Without Doing It Alone", with Fey Pelleg, hosted by Kerstin Dallinger, recorded 5 November 2026, on YouTube and Spotify.
• Daily use and access to learning, global, self-reported - PwC, Global Workforce Hopes and Fears Survey 2026, nearly 50,000 workers, published 29 September 2026.
• Support for experimenting, workflow fit and frequent use, US - Gallup, AI in the Workplace: What Separates Adopters and Holdouts, 23,717 employed US adults, 4 to 19 February 2026, published 12 April 2026.
• Embarrassed to ask for help, US, commissioned by a vendor - The Harris Poll for INTOO, 1,158 employed US adults, 31 March to 2 April 2026.
• Learning stays invisible - Nancy Baym, Eleanor Dillon and Sonia Jaffe, Harvard Business Review, 3 March 2026. I read the summary on the page, the article itself is behind a paywall.
• Fey Pelleg - AI Goddesses.