AI adoption is a people issue, not a tech project

In a nutshell

The real barrier to AI adoption is confidence

Most AI projects don’t fail because of bad tools. They fail because people aren’t ready to use them.

Every few months, someone posts a diagram on LinkedIn showing how AI is “transforming everything”. Boxes and arrows buzz with potential. But reality is far less tidy. AI tools are licensed and rolled out, yet adoption is patchy. Some teams embrace them; others quietly avoid them. Value remains… debatable.

This isn’t just theory. MIT’s report, The GenAI Divide: State of AI in Business 2025, found 95% of generative AI pilots fail to deliver measurable business impact. Only the 5% that targeted one clear problem, integrated well, and often used specialist partners saw tangible results, such as rapid revenue growth*. (See Fortune, 18 August 2025)

This isn’t a minor glitch. It’s a leadership problem wearing confusing technology as a disguise.

As the Fortune article goes on to say, the AI models themselves aren’t the stumbling block; it’s the ‘learning gap.’ Organisations rarely adapt tools into workflows, fall short on governance, and expect transformation from generic pilots. That explains the scale of the failure to date.

Change management first

Paul Roetzer, founder of the Marketing AI Institute and host of The Artificial Intelligence Show podcast, is blunt: “Change management needs to be thought of first.”

Too many organisations treat AI like an IT procurement exercise. But AI isn’t another system implementation. It’s a capability shift that touches everyone, including those who’ve perfected the art of sitting at the back of training sessions and hoping nobody calls on them.

AI Adoption doesn’t need a new AI-flavoured methodology. It’s still a change project: define the goal, understand the impact, engage people, support learning, track what works.

Overcomplicating it just adds to the fear and the noise.

Mike Kaput, Roetzer’s podcast co-host, sums it up: “100% of leaders know it’s critical. But only a fraction actually do something about it.” Many wait for the perfect tool or bulletproof strategy while more agile competitors simply get on with it.

Cross-functional or cross-your-fingers?

If AI lives solely in the IT department, every other functional leader breathes a sigh of relief, delighted to see it’s now someone else’s problem. In reality, it demands input from all core functions: HR (roles and behaviours), L&D (skills), Comms (messaging), Legal and risk (safe use).

Managers are pivotal – and vulnerable. A Harvard Business Review article (“How AI is Redefining Managerial Roles”, July 2025) highlighted how middle managers are under growing pressure as AI automates many of the coordination and monitoring tasks that once defined their roles. Some organisations are using this as an excuse to remove layers of management, while others are reducing junior roles instead.

Both moves may save money in the short term, but risk hollowing out your future leadership pipeline. AI can fill gaps in spreadsheets, but it can’t mentor, coach, or stop bright graduates from leaving after 18 months.

To support meaningful AI adoption, HR might expand its focus to include roles and behaviours; not just what people do, but how work actually gets done, how that’s changing, and what individuals need to do differently to make AI useful.

It could also mean rethinking how we measure contribution. Less about time in the office, more about what gets done, and potentially a move, formal or informal, towards results-only work environments (ROWE).

Unlike rolling out a finance system, you can’t just flick a switch and expect new behaviours to follow. People don’t change because they’re told to; they change when there’s support, clarity, and something in it for them.

That’s why behaviour and change management need to be built into any AI adoption strategy from day one, not bolted on at the end.

Fear isn’t resistance

AI triggers real anxieties: “Will I lose my job?”, “Do I look stupid if I don’t get this?”. Senior leaders are often the most exposed. They’re expected to lead the AI conversation but may feel out of their depth with the tools or terminology. That uncertainty can lead to avoidance, defensiveness, or an over-reliance on consultants.

Middle managers face a different problem. They’re told to drive adoption while shielding their teams from disruption, usually without support, training or a clear strategy. It’s little wonder they risk becoming the frozen layer, stuck between strategy and execution.

Psychological safety matters. People need permission to say, “I don’t understand this,” without fearing they’ll be marked down for it. Managers have to create space for questions, experimentation, and failure. Fear doesn’t vanish because a memo from the sixth floor tells people to “embrace innovation”.

Credibility concerns hold back adoption

A recent Harvard Business Review study (Research: The Hidden Penalty of Using AI at Work, August 2025) found that workers who used AI were judged around 9% less competent than those who didn’t, even when their output was identical.

That reputational hit may help explain why some people hold back. In the study, only 31% of female engineers and 39% of those over 40 years old adopted a new AI tool, compared with 41% overall. The authors suggest that reputational concerns may weigh more heavily on some groups, especially where credibility is already hard-won.

That’s not a usability issue. It’s a social one. If people believe using AI makes them look less capable, they’ll avoid it or use it quietly. 

Practical entry points

The fastest way to build AI confidence is by starting small. Everyday tasks work best:

• Drafting emails or briefing notes
• Summarising documents or meeting notes
• Generating ideas for campaigns or presentations
• Improving the tone and clarity of communications

These aren’t cutting-edge use cases, and that’s the point. They’re safe, repeatable, and genuinely useful. Pair them with some basic skills, such as writing clear prompts, editing AI outputs, and knowing when not to use AI, and you give people some quick wins and a sense of control.

Practical workshops help accelerate this shift. Create safe spaces where people can experiment without fear of looking foolish. For senior leaders, these workshops are particularly valuable. They allow them to ask questions they might otherwise swallow and build the baseline understanding they need to lead AI conversations credibly.

Workshops also send a cultural signal: you don’t need to be an expert yet, but you do need to engage.

Culture trumps Copilot

Culture makes or breaks adoption. A culture built on control and perfectionism sees AI as a threat. A culture comfortable with ambiguity and learning treats it as a tool.

Culture isn’t something you can “roll out”. You shape it through leadership behaviour, feedback loops, and consistent reinforcement. People copy what they see at the top. If leaders use AI sensibly, ask questions, and admit what they don’t know, employees will follow. If leaders hand AI over to IT and disappear, so will everyone else.

Even if you’re “not adopting AI”

Some organisations are holding back from big AI initiatives. Fair enough, that is an option. But here’s the awkward truth: employees are probably using AI already, often without guidance or safeguards. It’s better to set clear boundaries than let things happen in the shadows.

At the same time, staff are bombarded with media headlines about AI-linked job cuts. If you don’t explain your position, rumours will fill the vacuum. Even without a major rollout planned, an AI policy and an open conversation about where you stand are essential.

What good looks like

A mature approach to AI adoption focuses less on the toolkit and more on the people.

It includes:

• Clear goals and impact mapping
• Role-specific training tied to real work
• Honest dialogue about risks and limits
• Feedback loops that track usage and impact, not just licences
• Safe spaces to test and learn without judgement
• Recognition that fear of “looking stupid” can be a bigger barrier than the tech itself
• Culture audits to surface hidden fears or blockers
• Leadership visibly engaging with AI
• Rethink roles and responsibilities to match the capability shift

This isn’t particularly flashy, but it works and it’s sustainable.

Recommendations for moving from experiment to impact

To give AI adoption a fighting chance:

1. Start simple: Treat AI like any other change project. Define outcomes, map the impact, and avoid overcomplicating the process.
2. Address the fear factor: Particularly for senior leaders, offer safe, judgment-free ways to explore AI tools. Don’t assume digital confidence.
3. Select practical use cases: Start with tasks people already perform, such as drafting, summarising, and improving tone. Show how AI helps without replacing.
4. Build prompt literacy: Teach people how to write clear prompts, review AI outputs, and know when not to use AI. Keep it pragmatic.
5. Support cross-functional engagement: Involve HR, L&D, marketing, and legal from the start. AI is a whole-business issue.
6. Align leadership behaviour: Leaders need to be seen using AI, asking questions, and encouraging curiosity. It sets the tone.
7. Track adoption, not just deployment: Look beyond licences and logins. Focus on real usage and measurable value.
8. Ask why some people aren’t using the tools: And whether the blockers are about personal credibility, confidence, or pride, rather than technical capability.
9. Create safe spaces to test and learn: Workshops and pilots build confidence, surface issues, and prevent “quiet avoidance”.
10. Audit your culture: Identify blockers to experimentation, psychological safety, and curiosity. These matter more than toolkits.
11. Rethink role design: Use AI as a prompt to reconsider responsibilities and development pathways, especially for managers and early-career roles.
12. Create an AI policy: Clarify what employees can and can’t do with AI, even if a major rollout isn’t on the cards.
13. Reorient investment: Most AI budgets chase sales and marketing pilots, but the strongest ROI often comes from back-office automation, reducing outsourcing and streamlining admin.
14. Don’t do it all yourself: Firms that work with external vendors succeed at about twice the rate of those attempting purely internal builds. (Fortune, 18 Aug 2025)

The leadership test

Rolling out AI tools without a change strategy isn’t transformation. It’s just deployment.

And that’s not a technology problem.

It’s a leadership one.

If you’d value some support in rolling out AI in your organisation, please get in touch.

* This MIT report has been strongly and very effectively challenged by AI expert and podcaster Paul Roetzer. We believe that while the figures may have been overestimated, there is a trend of many companies realising fewer gains than anticipated.