Why AI adoption is far more than a technology problem
Q5 has partnered with the Pracademic Lab at the Department of Psychological and Behavioural Science (PBS), London School of Economics and Political Science (LSE), to host a dissertation project exploring the human experience of AI adoption. The research was conducted by Devon Litherland, a recent graduate from the LSE MSc in Organisational and Social Psychology, with Q5 participating at key moments across the process to support rigour, relevance and impact. The following article shares key insights from that research, translated for business leaders.
The research explored what working with AI actually feels like from the inside, drawing on latest research and conversations with professionals who see AI adoption up close across many organisations.
Reading time: 6 minutes
Almost all of the conversations started in the same place: AI makes everything faster. But a few minutes in, something else surfaced. Someone wasn’t sure they were still needed. The part of the job they were proud of was disappearing. On the surface, AI was about freeing people up for better work. Underneath, it sounded a lot closer to a form of professional ‘grief’.
One of the most striking findings was how little of this ‘grief’ looked like forms of resistance. Instead, the human side of this shift is playing out where leaders can’t see it, and that may be shaping how much value organisations get from AI.
This problem is deeper than simply job security
Most conversations about AI and work stop at one question: will I be replaced? But beneath that sits a harder question. When a tool can do in seconds the thing that used to mark you out as good at your job, what makes me valuable now?
Three things kept surfacing. First, Identity; expertise is part of who people are at work, and AI can blur the line that used to make them feel skilled. Second, Mastery; we build pride and confidence by learning to do hard things well, and when the hard parts get automated, the sense of getting better goes with it. Finally, Meaning; keeping your job is not the same as keeping the parts of it you care about.
The research suggests meaning comes less from the output itself than from the work of getting there: the judgement, the effort, the pride in an outcome you earned. When AI takes over that work, it can hollow out parts of a role even where the job itself remains.
And this is where a skills-and-tools approach to AI adoption starts to fall short. Training can teach someone how to use the tool. It can’t tell them what makes them valuable now, where their judgement still matters, or what will make their work feel meaningful. Those are the real questions people are working through, and in doing so, they’re renegotiating who they are at work, often without realising it.
Fear doesn’t always look like resistance
Uncertainty does not always show up as resistance. In many cases, it goes quiet. Fear doesn’t always create resistance. Sometimes it creates concealment.
A recurring pattern was the hidden use of AI at work. Not because people were doing anything wrong, but because being open about it didn’t feel safe. For some, it’s about credit: if AI shaped a piece of work, will it still be seen as their work? For others, it’s about workload: share an efficiency, and the reward might end up as more work. But underneath both sits a bigger fear: if I show how much faster I am with AI, am I making my own role easier to replace? Three different worries, but they lead to the same behaviour: concealment. Being seen to use AI has itself become the risk.
This points to a form of ‘Shadow AI’: experimentation that happens in private. People are building AI into their own workflows and learning what it is useful for, but doing this quietly. This creates a productivity paradox. Organisations want people to share the efficiencies and learning they’re discovering, yet the people discovering them have good reasons to keep them to themselves. The result is learning that often stays with one person, when it’s exactly the kind of learning that needs to travel from colleague to colleague. This may help explain why some organisations are investing heavily in AI and still struggling to see the return.
Where will tomorrow’s experts come from?
There is a second cost associated with concealment that is even harder to see. It hides value from leaders today, but there is a longer-term issue too: the erosion of foundational skills. Early-career employees can now produce impressive-looking work without doing the groundwork that used to teach people what ‘good’ looks like. Writing endless reports, building the analysis, structuring the argument. This is traditionally where people build judgement, or what one person called ‘taste’, an instinct for what good looks like.
The risk is easy to picture through the analogy of pilots who rely so heavily on autopilot that their manual flying skills fade, until the day the system fails.
This remains a concern rather than a proven outcome. But it raises a question every business should be asking: if AI takes over the work that built today’s experts, where will tomorrow’s come from? This matters even more as some organisations cut entry-level roles while still needing a pipeline of people who can one day manage and lead.
Leading a change that you are still navigating
A clear leadership pattern emerged in how organisations are approaching AI adoption. AI was often treated as a technology rollout rather than a human transition, and in some cases, adoption seemed driven more by the need to prove AI was being used than by any clear sense of what it should change. Leaders also tended to say little about what AI meant for people’s roles, and in that gap, the story was written elsewhere, by headlines about layoffs and jobs being automated.
But there was an added dimension here. Some of the leaders pushing hardest for adoption were among the least confident users themselves. People noticed, and it created a credibility gap that made everything else the leader said about AI harder to believe.
As with any organisational issue, it would be easy to read this as a story about poor leadership, where executives and managers just need to do better by their people. But AI is quite democratic in its impact. Leaders are often facing the same questions themselves: the same uncertainty about their own expertise, their relevance and what their role even becomes.
One participant put it quite starkly. “The change curve is the same as the grief curve, and as a leader, you have to process that for yourself before you can help someone else go through it.” The research terms this the ‘Leader grief gap’. Leaders are expected to hold the intense emotional weight of this change for their teams before they’ve had the chance to process their own. That’s a lot to ask of anyone.
This may help explain the patterns described above. If leaders haven’t yet worked out what AI means for their own identity and role, the human side of the change becomes the hardest part to lead. So, they may retreat to what feels safer, like tools, targets and adoption for show, or just saying nothing at all. Seen this way, capable leaders are being asked to guide people through something they haven’t yet come to terms with themselves.
One of the most important shifts in the research was around psychological safety. The study began with the expectation that it would be the key to helping people adapt to AI. The conversations told a much more complicated story. Psychological safety turned out to be necessary, but not sufficient. Yes, it matters because it makes disclosure, experimentation and shared learning possible, which is exactly what concealment prevents. But it doesn’t reach the deeper layer of threatened identity, lost meaning and grief. People don’t just need permission to use the tools. They need to see leaders acknowledge what this change is really costing: the uncertainty and the parts of the job people are losing, not just the gains.
What can we do?
The implication for leaders is clear: adoption is hard to force, but the conditions for it can be built.
And the order matters. The research points to a sequence that starts with leaders making sense of AI for themselves, moves through organisational permission and trust, and then allows learning to spread through teams.
- First, make sense of it. Leaders must work out what AI means for their own role and value before they ask anyone else to. In practice, that means facing the same questions their teams are working through.
- Second, legitimise. Create a clear moment where the organisation openly says AI use is welcome. It sounds simple, but its impact goes well beyond policy. It turns AI from something people use privately, even guiltily, into something they can talk about, share and learn from together. And it’s far more believable when it comes from leaders who’ve already made sense of AI for themselves.
- Third, socialise. Peer influence is one of the most underestimated drivers of adoption. Having someone in your own team who has worked out how to use AI well, and shows others, can do more than a company-wide training programme on its own. Training still matters, but it works best alongside trusted local champions who make experimenting with AI feel normal and practical.
- Fourth, redesign. Use what you’re learning to rethink how work gets done, rather than simply bolting AI onto old ways of working. This redesign includes how people build expertise. If AI is taking over the groundwork people used to learn from, they’re less likely to develop those skills along the way, so organisations will need to think deliberately about how judgement and ‘taste’ get built. All of this might start with a simple question: ‘What work should people still do themselves, and at what stage in their career?’
The opposite dynamic is possible too. In one example from the research, a technology company was so open about its workforce plans that, over time, people had to justify why they weren’t using AI, rather than hiding that they were. This creates the exact opposite dynamic to the concealment described earlier.
It is worth holding on to this central point: AI adoption is not the same as AI value. An organisation can have everyone using the tools and still achieve suboptimal value. Part of that value is hidden right now, in the experimentation and learning people keep to themselves. Part of it is at risk over time, in the judgement and expertise people have traditionally built by doing the work. So, the challenge for leadership goes beyond getting more people to use AI. It means turning individual experimentation into shared capability, without trading tomorrow’s expertise for today’s efficiency.
The bigger lesson is this: AI was never only a technology challenge. It’s a human one too. How well any organisation comes through it will depend on whether its leaders are willing to turn toward their people, and toward themselves. The organisations that draw real value from AI may well be the ones that keep investing in the judgement, mastery and expertise that still make their people valuable.
About the research partnership
Q5 has partnered with the Pracademic Lab at the Department of Psychological and Behavioural Science (PBS), London School of Economics and Political Science (LSE). This partnership involved hosting a dissertation project for a student on the MSc in Organisational and Social Psychology at the LSE. In line with the approach of the Pracademic Lab, Q5 participated at key moments across the process. This ensured the highest standards of rigour as well as relevance and impact within the business.
By combining ‘Practitioner’ and ‘Academic’, Pracademia seeks to capture where two words, and worlds, meet. The LSE Pracademic Lab makes the intersection between the worlds of abstract ideas and concrete practices visible. It creates value by facilitating the translation and travel of ideas, under messy, real-world conditions. The Lab is believed to be the first of its kind in the world.
Researcher: Devon Litherland — Recent Graduate from the LSE MSc in Organisational and Social Psychology
Academic supervisor: Barry Rogers — Professor in the Practice of Organisational and Social Psychology (LSE) and Founder of the Pracademic Lab
Q5 sponsor: Pavlina Kouvela – Partner and Head of Q5’s People Performance Practice
We are all about organisational health, which separates good organisations from the great. Whether our clients are at the top of their game (and want to remain there) or are in ‘turnaround’ mode, we all need to work on our organisational health.
Whatever the situation, be it a strategic conundrum, a market opportunity, or an operational gripe, we combine the art and science of organisational health to help our clients improve and excel.
Please visit the firm link to site
