AI doesn't create leadership problems - it exposes them
Five patterns from conversations with product and technology leaders - and practical questions to help you spot them in your organisation
I’ve been talking with senior leaders about what AI is doing to the human side of their organisations. Not adoption rates or tooling decisions, but what is happening to trust, psychological safety, status, identity, expertise and the way work gets done.
These conversations have mainly been with product and technology leaders, which makes this transformation particularly interesting. They are not just watching this happen from the sidelines - their organisations are building AI into products and ways of working, while their own teams are among those feeling its impact. Researchers, designers, engineers and product people are already seeing roles reshaped, reduced or removed. These leaders are being asked to lead a transformation that is happening to them.
The leaders come from scaling mid-sized companies and FTSE 100 businesses across financial services, beauty and wellness, media, renewable energy and retail, in the US, Europe and Asia Pacific.
How far their organisations have integrated AI varies wildly. One CPO has an executive team building its own agents. In another organisation, most people are using AI to summarise their growing deluge of emails and documents - which, of course, their colleagues created using AI.
Despite those differences, I keep hearing versions of the same five patterns:
Adoption travels through trust, not hierarchy. People are more likely to experiment after seeing someone they respect use AI on real work than because they have been told to use it.
Leaders set the ambition ceiling. Teams take their cues from what leaders visibly do, not simply from what they approve or announce.
Resistance is often about identity, not technology. When AI can do work that has long defined a person’s expertise, they need to reconnect with the contribution they want to make—and have a meaningful role in shaping how their work evolves.
Psychological safety shapes what people disclose. People need to know what is expected—and to feel able to be honest about how they are experimenting, where they are struggling and whether they are using AI at all.
AI is changing who carries the work. AI does not always remove work; often, it simply moves it. Leaders need to look beyond the immediate efficiency and ask what new work will be created, who will inherit it and whether they are equipped to carry it.
From my experience working with organisations going through transformation of one form or another over the past 15 years, these patterns are not unique to AI - but the speed of development, ease of access, competitive pressure to respond quickly and a direct challenge to professional expertise are making them harder for leaders to ignore.
Below, I explore each pattern in more detail, the leadership challenge it reveals and a prompt to help you consider how it may be playing out in your organisation.
1. Adoption travels through trust, not hierarchy
Many of the leaders I spoke with described some version of a centralised approach to introducing AI. Training was offered, tools were approved, incentives were put in place and messages came from the top telling people they should be using them.
A great deal of effort went in. But that was rarely what persuaded people to start.
What moved them was watching someone they knew and respected use AI on real work and come out fine—maybe even a bit excited.
In one organisation, a group of technical experts had resisted AI for months. Then a respected member of the group demonstrated how she was using it on a live project. Within days, one of her colleagues had built something himself and was showing it to his team, and usage began to spread from there.
Adoption was social. Seeing a trusted colleague use AI made experimenting with it feel relevant, legitimate and a little less risky.
What this asks of leaders: This makes the informal relationships inside an organisation increasingly important. Leaders cannot manufacture trust by putting “AI champion” in someone’s title. They can, however, pay attention to whom people already turn to, give colleagues time to experiment together and make it acceptable to share work before it is polished.
A question to explore: Who are people actually learning about AI from, and what would make it safer to experiment and learn in a practical and open way?
2. Leaders set the ambition ceiling
Teams do not act on what leaders authorise so much as what they see them do as individuals.
In organisations where executives were building with AI, asking questions and learning in public, more ambitious uses began to spread. Where leaders mainly used it to summarise their inboxes, that was roughly what everyone below them did too.
The ceiling tended to sit where the leader's curiosity stopped.
What this asks of leaders. Leaders do not need to become AI experts. But they cannot entirely outsource their curiosity either. What they use AI for, the questions they ask and how they respond when an experiment fails are all being watched and interpreted. They tell people what the organisation really values, how ambitious it wants them to be and how safe it is to be a beginner.
A question to explore: What is your own behaviour working with AI signalling about the ambition, experimentation and learning you expect from others?
3. Resistance is often about identity, not technology
No one I spoke with described resistance to adopting AI simply because they didn't believe the tools worked. The promise of specific functionality wasn't the problem.
Instead, they talked about people protecting what they were known for and pushing back when AI began to blur familiar roles. Specialists and subject-matter experts dismissed AI as something for “software people,” while engineers were unsettled when product managers began producing working code.
It is tempting to treat this as an argument about responsibilities or the boundaries between roles. But underneath it is a question about identity: What happens to the thing I am good at?
Expertise still matters. But where that expertise creates value may be changing—from being the only person who can produce something to being the person who can frame the problem, tell good work from bad and take responsibility for the result.
What this asks of leaders. Telling people that AI will free them to do “more valuable work” does not answer the question. More valuable to whom? And who gets to decide?
People may not be able to control everything AI changes, but they need some meaningful influence over how their work evolves. Too often, they are invited to help shape the future after the important decisions have already been made.
Leaders also need to be honest when roles, status or opportunities really are disappearing. Purpose and agency matter, but they cannot be used to put an optimistic gloss on genuine loss.
A question to explore: What might people believe they are losing as AI changes their work - and how much genuine influence do they have over what their contribution becomes?
4. Psychological safety shapes what people disclose
I found this pattern especially interesting. Some people are not sure whether using AI counts as cheating, so they do not say when or how they have used it. Others are reluctant to admit that they are struggling, sceptical or not using it at all.
The result is a culture of guessing: about who is using AI, what they are using it for and whether they should disclose it.
Organisational choices can make that ambiguity worse. When only some people have access to approved tools, others may find alternatives. When training bears little relation to their actual work, people are left to decide for themselves what acceptable use looks like.
AI use continues. It simply becomes harder for the organisation to see, learn from or govern.
What this asks of leaders. Psychological safety on its own is not enough. Team members also need clear agreements about where AI use is appropriate, when it should be disclosed and where individual responsibility still sits.
Leaders have a role here too. If they want people to talk openly about their use of AI, they need to be prepared to talk about their own—including where it produced something poor, misleading or simply rather odd.
A question to explore: What might someone on your team hesitate to tell you about how they are - or are not - using AI? What makes that hesitation reasonable?
5. AI is changing who carries the work
The first four patterns are largely about how leaders show up. This one is about how far they look before deciding.
As we're all learning, AI does not always remove work - it often simply moves it. Several leaders described a sharp rise in documents and analysis. Producing content had become easier, but other people had inherited more to read, assess and reconcile before they could decide what mattered.
In another organisation, leaders removed a support function entirely after deciding that AI could cover its visible tasks. But the function had also coordinated work, interpreted analytics and helped people navigate what happened next. The roles disappeared, but much of the work remained and landed elsewhere, damaging morale and trust.
What this asks of leaders. “Can AI perform this task?” is too narrow a question. Leaders also need to understand what new checking, interpretation or coordination will be required. What knowledge could disappear? Who will inherit the less visible parts of the work? Do they have the capacity, context and authority to carry them?
A decision can look efficient when viewed one task at a time and create a considerable mess when viewed across the whole system.
A question to explore: Take one current AI decision: what work will disappear, what new work will be created, who will absorb it, and what might become harder to see?
What this reveals about transformation
As I said at the beginning of this article, strip away the AI language and little here is new.
Trust has always shaped whose example people follow
Change and transformation has always unsettled identity and status
People have always concealed behaviour when honesty feels risky
Leaders have always set what feels possible by what they do, not what they authorise
What feels like efficient decisions has often created work elsewhere.
What has changed is the pace and reach.
AI asks the same leadership work of more people at once, with less time to get it right. And it reaches directly into what many people believe they are good at and why their contribution matters.
The danger is that organisations become so preoccupied with what AI makes possible that they forget what decades of transformation have taught us about how people experience change.
AI may be the transformation everyone is talking about. But the leadership fundamentals it exposes have been hiding in plain sight for years.
This is increasingly part of my work with senior leaders, groups and executive teams: helping them make these dynamics discussable, understand what AI is changing beneath the surface and create clearer agreements about how roles, responsibility, authority and work need to evolve.
These five patterns are still early. I’m continuing the conversations through the autumn to test where they hold, understand what I may be missing and see what else is emerging.
If you’re leading through this change and recognise any of them - or are seeing something different - I’d be interested to compare notes. You can get in touch here.
