A question came up in a recent workshop on delegating with AI that I haven’t been able to stop thinking about since.
One of the managers asked what happens if somebody in their team simply doesn’t want to use AI.
My first instinct, because I spend a lot of my working life helping people use AI, was to think about how you might help that person move forward. Do they need more confidence? Better examples? A chance to practise somewhere that feels safe?
But the more we talked about it, and the more conversations I’ve had with managers since, the less convinced I am that “How do I get them to use AI?” is the right place to start.
Because what does “I don’t want to use AI” actually mean?
Perhaps they haven’t got a clue how to use it properly and they’re worried about making a mess of something important. Perhaps they’ve tried it for this particular job and honestly found it more trouble than it was worth. Or maybe they’re perfectly capable of using AI but uncomfortable about what it means for their job, the environment or a piece of work they don’t believe should be handed over to a machine.
Those are very different conversations.
Yet I think we’re in danger of bundling them together under one very convenient label: resistance.
If someone in your team doesn’t want to use AI, I wouldn’t start by trying to persuade them. I’d start by finding out what sits behind the hesitation, because the right response might be training, clearer guidance, a conversation about their role, an organisational question that needs answering or even a perfectly sensible decision not to use AI for that particular task.
In this article
You Can Be Excited About AI and Worried About It at the Same Time
We have a tendency to make AI adoption quite binary. You’re embracing it or resisting it. You’re an enthusiast or a sceptic. A user or a refuser. Except that doesn’t seem to be how people actually feel.
A September 2026 Pew Research Center study across 37 countries found that a median 41% of people were equally concerned and excited about increased AI use in daily life. In fact, that mixed position was more common across the countries surveyed than being predominantly concerned.
I love that finding because it feels much closer to reality. You can absolutely be both.
I use AI a lot. I spend a frankly ridiculous amount of time experimenting with it, finding new ways to use it and thinking about what it could make possible. I’ve seen it help people get unstuck, do things they couldn’t previously do and get through work that would otherwise have swallowed hours. And I still have concerns.
I question whether we’re using too much AI for too little value. I care about the environmental impact. I get twitchy when people accept an AI answer because it arrived confidently and beautifully formatted. I wonder what happens to particular jobs and skills as more work can be automated, and I’m interested in where we need to deliberately keep humans in the loop.
Those things don’t cancel each other out.
It’s one of the reasons we developed our Right-size AI thinking. I want the value AI creates, but I don’t think that means using as much of it as possible for absolutely everything. Our approach is essentially: don’t use less AI, waste less AI.
So when somebody raises a concern about AI at work, I don’t think we should automatically read that as a signal that they’re on the wrong side of adoption. They may simply be thinking carefully about it.
I’m Also Learning Not to Turn Every Concern into a Debate
This came up for me again in another AI workshop recently.
Because I’m broadly pro-AI, I’ve realised my natural response when somebody challenges it can sometimes be to start building the counterargument. If they raise the environmental cost, I want to explain the nuance in the data. If they’re worried about jobs, I want to talk about jobs being transformed and created as well as lost. And, obviously, I usually think my argument is pretty good.
But proving somebody wrong isn’t necessarily a brilliant opening move in a management conversation.
I’ve caught myself doing exactly this outside work too. Someone close to me has very strong concerns about the environmental impact of AI and, because I thought some of the argument needed challenging, I kept pushing back with evidence and examples. All I really achieved was making both of us defend our positions more strongly. That experience came straight back to me during a recent workshop when we were talking about how managers respond to concerns about AI.
If I immediately start explaining why somebody’s concern isn’t quite right, we stop exploring what they actually think and start having a debate.
That doesn’t mean managers have to agree with every concern. There may absolutely be assumptions to challenge, information to share and expectations to clarify. But curiosity is usually a better first response than persuasion.
If someone says they don’t want to use AI because they think it will eventually cost jobs, I’d now be much more inclined to ask what specifically worries them about that before launching into my TED Talk on the future of work. I might still disagree with them, but at least I’ll know what we’re actually disagreeing about.
AI Resistance is Often a Diagnosis Problem Before It is an Adoption Problem
The more conversations we have with managers, the more I think the word “resistance” can get us into trouble.
In one recent workshop, a manager talked about someone in their team who was reluctant to use AI. Almost immediately, the wider conversation moved into completely different possible explanations: fear about what AI might mean for the role, lack of confidence, whether the use case actually made sense and whether the organisation itself had been clear enough about what was expected.
That’s the problem with a label like resistance. It sounds like the diagnosis has already been made. If someone is resisting, the implied next step is to overcome it.
Seven Places to Look Before Deciding What Somebody Needs
We didn’t start with seven neat categories. They emerged as I went back over recent leadership workshops and client conversations and noticed the same themes cropping up in different forms. Sometimes the issue was capability or confidence. Sometimes it was uncertainty about what people were actually allowed or expected to do. Again and again, the conversations also moved into value, trust, workload, jobs, ethics and what people didn’t want to lose from their work.
Across three separate client sessions and conversations over the past couple of weeks, we’ve heard variations of all of these. We’ve grouped those recurring themes into seven places for managers to look before deciding what somebody needs.
| What might be underneath it? | What it might sound like | What might help |
| Capability | “I don’t really know how to use this properly.” | Relevant training, practice and feedback |
| Confidence | “I know roughly how, but I’m worried I’ll get it wrong.” | Safe experimentation, support and lower-risk practice |
| Clarity | “Are we actually allowed to use AI for this?” | Clear guidance, boundaries and expectations |
| Value | “It takes longer to fix the AI output than to do it myself.” | Revisit the use case or workflow |
| Trust | “I don’t trust what it gives me.” | Verification, critical thinking and better checking processes |
| Impact | “What happens to my job if this works?” | An honest conversation about workload, roles and what is known or unknown |
| Values and identity | “I don’t think we should use AI for this.” or “This is the part of my job I actually enjoy.” | Explore the concern rather than immediately trying to remove it |
I wouldn’t sit somebody down in a one-to-one and clinically diagnose them against seven boxes. That would be a fairly quick way to make a normal management conversation extremely weird.
But it gives us somewhere useful to look, and it makes it much less likely that we solve the wrong problem.
Training Matters Enormously, but It Needs to Answer the Question People are Actually Asking
I want to be particularly careful here because I don’t want this article to drift into “AI training doesn’t work”. VTT delivers AI training. I have no intention of writing myself out of a job.
Good training matters enormously.
In Deloitte’s 2026 survey of 25,000 UK workers, half of people using GenAI said they had received no formal training at all, while only 28% said their organisation had a GenAI policy they knew how to find, understood and regarded as clear and current. That’s quite a combination.
We have people being encouraged to use rapidly changing technology on real work while huge numbers are teaching themselves and many aren’t particularly clear on the rules.
So yes, sometimes what looks like reluctance is absolutely a capability or confidence issue. If everybody around you is merrily discussing agents, models and prompts while you’re still wondering which box you’re meant to type into, that can feel pretty exposing.
Giving somebody an opportunity to practise with an approved tool on work that matters to them, understand where it goes wrong and learn how to judge the result can make an enormous difference. That’s exactly the sort of practical capability we build in our AI for Productivity workshop.
But training works when the problem is something training can solve.
If the real question is “What happens to my role if I become really good at this?”, another prompting session isn’t the answer. If somebody doesn’t know whether they can put client information into the tool, they need clarity. If they have a genuine ethical objection, they need a conversation. And if the tool has repeatedly produced poor work for this particular task, it may be the workflow rather than the person that needs fixing.
The point isn’t less training. It’s better diagnosis before we decide what support people need.
And Then There’s the Productivity Bargain
This is one I keep coming back to because we throw around the phrase “AI will save you time” incredibly casually.
AI absolutely can make particular tasks much quicker. There are things I now do in minutes that would previously have taken considerably longer, although give me a new AI feature and I can still quite happily lose three hours “saving time” while I investigate what it does, where it works and seventeen other things I hadn’t planned to explore that afternoon.
But the more interesting question is what happens at organisational level.
In one recent workshop, we asked managers what they would want to do with capacity released by AI. They talked about more thinking, better coaching, less rushed conversations and more time for genuinely human work.
Then someone challenged the premise. A report that once took two weeks to produce might now arrive in half an hour. That sounds wonderfully efficient until the leader who previously had one report to review suddenly has ten of them. The work hasn’t necessarily disappeared. The pace and volume have changed.
Another recent conversation added a different wrinkle. If AI removes more of the routine or straightforward work, what gets left behind can be the cognitively demanding stuff: judgement, analysis, complex decisions and difficult conversations. Someone may technically have fewer repetitive tasks while spending a greater proportion of the day doing the work that takes the most out of them.
Deloitte’s research found that half of employees using GenAI reported no time saving at all, while around half of those who did save time used at least some of that capacity to do more of the same work.
What do we want to do with the capacity AI creates?
If employees hear “AI will save you time” and interpret it as “great, now we can give you more work”, their caution isn’t particularly mysterious. They may not be questioning the technology. They may be questioning the bargain.
Some People are Worried About What Happens If AI Actually Works
Job concerns are probably one of the hardest parts of this conversation because I don’t think managers should give reassurance they can’t honestly guarantee.
In another recent client conversation, an HR leader told me that even somebody working in a highly specialist, hard-to-recruit role had asked whether AI could eventually put their job at risk. On paper, you might look at that role and think, AI replacing you is hardly the immediate problem here. That doesn’t stop the concern feeling real to the person having it.
Deloitte found that 64% of weekly GenAI users in the UK were worried managers would think AI could do their jobs.
There’s a comforting line that gets repeated a lot: AI won’t take your job. Someone using AI will.
I understand the intention. It encourages people to develop capability rather than hide from a technology that is changing work. But it is also a bit glib.
The International Labour Organization estimates that around one in four workers globally are in occupations with some degree of GenAI exposure. Its conclusion is not that all those jobs will disappear. Because most occupations still contain tasks requiring human input, transformation is considered more likely than wholesale replacement.
That’s considerably more nuanced than either “AI is coming for everyone’s job” or “don’t worry, everything will be fine”.
So if somebody says, “I’m worried that if I get really good at automating this work, I’m proving we don’t need as many people doing it”, I don’t think their manager should automatically reply, “Don’t worry, your job is safe.” They may not know that.
A more useful answer might be: “I understand why you’re asking. Here’s what I know about how we’re using AI in our team at the moment, and here’s what I don’t know yet.” That may sound less reassuring. I suspect it builds more trust.
People Can Be Worried About Falling Behind as Well as Worried About Moving Too Fast
There’s another tension that makes the whole idea of “AI resistance” feel increasingly inadequate.
Some people are nervous because AI is changing too much. Others are nervous because they can see the change happening and don’t feel they’re keeping up. And, again, the same person can feel both.
Microsoft’s 2026 Work Trend Index, based on a survey of 20,000 AI users across ten countries, found that 65% feared falling behind if they didn’t adapt quickly enough, while 45% simultaneously said it felt safer to focus on existing goals rather than redesign their work around AI.
That tension makes enormous sense to me. People are being told to experiment, rethink how they work and learn a technology that changes underneath them, while still being accountable for everything they were already supposed to deliver.
Leaders aren’t immune to this either.
When computers first became commonplace at work, senior people could sometimes delegate the technology itself. John and I worked in the same organisation years ago where it genuinely wasn’t unusual for a senior leader to dictate emails to a PA rather than use the computer themselves. That feels almost comically old-fashioned now.
AI is different because a leader doesn’t necessarily need to become the organisation’s best prompt engineer, but they increasingly need to understand enough about AI to ask intelligent questions about productivity, risk, people, capability and how work is changing.
You can delegate some of the doing. It’s increasingly difficult to delegate the understanding.
Sometimes the Cautious Person is Exactly Who You Need
This might be the point I care about most.
We don’t need everybody in an organisation to become equally enthusiastic about AI, and I’m not convinced that would be healthy anyway. We need enough curiosity and confidence for people to explore what AI can do, enough capability to use it well and enough scepticism in the room to question it when something doesn’t look right.
The leadership conversations we’ve been having bear that out. Managers have raised concerns about people losing opportunities to develop because AI does the early-career work for them, about independent thinking becoming weaker, about convincing but incorrect outputs and about keeping human judgement and a “moral compass” in decisions that affect other people.
Interestingly, Microsoft’s research found that its most advanced AI users were more deliberate about sometimes not using AI. Forty-three per cent said they intentionally did some work without AI to keep their skills sharp, compared with 30% of other users, and 53% said they deliberately paused before starting a task to decide whether AI or a human should do it.
That’s a very different picture of AI maturity from “use it as much as possible”. Sophisticated use includes knowing when not to use it.
We need somebody asking where the answer came from, whether the evidence holds up and whether a human should still be making the decision. We need people who notice when the impressive-looking AI output is actually worse than the original, and people who ask whether we’re automating something because it creates genuine value or simply because we can.
That kind of scrutiny sits at the heart of our Critical Thinking for AI work.
Healthy caution isn’t the opposite of AI capability. It can be part of it.
So What Should a Manager Actually Do?
All of that thinking is interesting, but if you’ve got a one-to-one tomorrow with somebody who has told you they don’t want to use AI, you need something slightly more useful than “human reactions to emerging technology are complex”.
We’ve started using a simple four-step approach at VTT.
1. Get Specific
Start with the use, not AI in general.
“I don’t want to use AI” is enormous. Ask what it is about this particular tool, task or use that concerns them.
Someone who doesn’t want AI influencing a decision about a customer may be perfectly happy using it to organise their meeting notes. Somebody who has had a terrible experience generating a report may already use AI very effectively for research.
Don’t turn a concern about one use into a position on the entire technology.
2. Get Curious
This is where those seven areas help.
Are you dealing with capability, confidence, clarity, value, trust, impact, values and identity, or perhaps a combination?
- “What is it about this use that concerns you most?”
- “What happened when you tried it?”
- “Are there uses of AI you’re comfortable with and others you aren’t?”
- “What would you need to feel clearer about what’s expected?”
And before responding, I’d ask myself one more question:
Am I trying to understand this person, or am I already trying to persuade them?
That one is as much a reminder for me as anyone else.
3. Work Out Whose Problem It is
Some things genuinely sit with the manager. You may be able to provide more practice, show someone a better use, clarify a team process or create somewhere safer to experiment.
Others belong somewhere else.
If the question is about approved tools, sensitive data, organisational policy, job design or what AI use is actually expected across the business, the employee shouldn’t have to rely on their manager making up an answer.
And neither should the manager.
4. Agree One Useful Next Step
The conversation should end somewhere.
Depending on what you’ve heard, the next step could be training, more practice, clearer guidance, a small experiment, revisiting the workflow, getting an organisational answer or having a more substantial conversation about how the role is changing.
Sometimes you may conclude that AI simply isn’t adding enough value to this particular task. That is also a useful decision.
Our Before You Call It Resistance conversation guide takes those four steps further, with the seven areas to explore and practical questions managers can use before and during the conversation.
Managers Can’t Compensate for Organisational Ambiguity
Eventually, though, you may reach the harder version of the original question.
What if somebody understands the tool, understands why it is being used and still says: “No. I’m not using it.”
That’s when the manager needs something more solid than an instruction to “drive adoption”.
This question has surfaced repeatedly in our recent client work, and what’s striking is that managers aren’t always the only people unsure of the answer. L&D and HR leaders are wrestling with it too.
Before Managers Can Deal with Refusal or Reluctance, Organisations Need to Know Which of These Applies:
AVAILABLE → ENCOURAGED → EXPECTED → REQUIRED
- Available means people can use AI when they find it helpful.
- Encouraged means the organisation actively wants people to explore useful applications.
- Expected means AI capability or use forms part of how particular work should now be done.
- Required means using an approved AI tool is an explicit part of performing a particular task or role.
Those are very different positions, and the answer may legitimately change depending on the work.
Deloitte’s research suggests many organisations still have a gap here. Sixty-five per cent of GenAI users surveyed said they did not hear leaders talking about the technology with a clear understanding of it, while only 28% felt they had a clear, current policy they understood and knew how to find.
Before we ask managers to handle difficult AI conversations brilliantly, we need to make sure the organisation has given them some answers.
That doesn’t mean producing a 47-page AI policy and considering the job done. Managers need practical clarity about the tools that are approved, the uses that are encouraged, where human judgement remains essential, what capability genuinely forms part of somebody’s role and where to take a concern they cannot resolve themselves.
The Goal Isn’t Enthusiasm. It’s Thoughtful Capability
If someone in your team doesn’t want to use AI, I wouldn’t start by asking how you can get them on board.
I’d start by finding out what is behind the response.
Sometimes the answer will be a skills or confidence gap and good training will make an enormous difference. Sometimes someone needs clearer organisational guidance, a chance to experiment or better information. Sometimes they’re raising a concern that deserves a proper conversation rather than an attempt to reassure it away.
And yes, sometimes AI use may genuinely be becoming part of what is expected in their role, which means a firmer conversation eventually has to happen.
But you can’t know which conversation you’re having until you understand the concern.
For me, the goal isn’t unquestioning adoption and it certainly isn’t trying to turn every sceptic into an AI enthusiast. It’s thoughtful capability: people who can use AI well where it adds value, question it when they need to, understand the responsibilities that come with it and recognise when another approach would serve the work better.
Because the colleague asking the awkward question isn’t necessarily the person holding your AI adoption back. They might be helping you do it better.
Take the Conversation Further
If you’re dealing with this conversation right now, our Before You Call It Resistance guide gives managers a practical way to understand what may sit behind somebody’s response to AI and decide what to do next.
And if these questions are becoming a broader leadership challenge across your organisation, our Leading in the Age of AI workshop helps managers navigate the people and judgement challenges that come with AI, from mixed reactions and changing expectations to productivity, delegation and keeping the right work human.
Frequently Asked Questions
What Should I Do If Someone in My Team Doesn’t Want to Use AI?
Start by getting specific about the use they are uncomfortable with and understanding what sits behind their response before deciding what support they need. The next step might be training, clearer guidance, experimentation, an organisational answer or a conversation about how their role is changing.
Is AI Hesitation at Work the Same as AI Resistance?
Not necessarily. Someone can use AI regularly and still have concerns about a particular task, quality, privacy, workload, jobs, environmental impact or ethics. Treating every concern as AI resistance can lead managers to solve the wrong problem.
How Should Managers Respond to AI Anxiety at Work?
Start with curiosity rather than immediate reassurance. Find out whether the concern is about capability, confidence, trust, job impact, workload, organisational expectations or something else, then respond to the issue the person is actually raising.
When Will AI Training Help?
AI training is particularly useful when the issue is capability or confidence, especially when people need practical experience with an approved tool and a better understanding of how to judge AI output. If the underlying issue is policy, role impact or a values-based concern, training may still be useful but it won’t answer the whole question.
What If AI Use is Expected in Someone’s Role?
Make the expectation specific and make sure you understand the organisation’s position before having a firmer conversation. Managers should not have to invent policy themselves, particularly where the concern touches data, employment expectations, ethics or role design.
