Companies are focused on deploying artificial intelligence. They are paying less attention to what it is doing to the people in charge of AI transformation.
Most leaders do not need to be told that AI is changing things. They are feeling it from superiors, shareholders, competitors: demands to keep up and also show productivity benefits. The business literature is largely framed around opportunity and adaptation: which tools to adopt, how to redesign workflows, what upskilling skills are needed and what governance to put in place. These are important questions. But they largely gloss over the human impact. What does AI do to the people being asked to lead it? For many senior executives, the honest answer is: quite a lot, and not all of it welcome.
The blind spot in the AI discourse
There is no shortage of guidance on AI and leadership. The McKinsey State of AI, the DDI Global Leadership Forecast, and Deloitte’s Human Capital Trends are all tracking something real. Yet the conversation remains almost entirely focused on what to do. Rarely on the agenda is the experience of the person doing it. What happens to a leader’s authority when a tool produces in seconds what used to require effort and judgment? What does it do to their credibility when they are expected to lead on something they are still trying to understand
Henley Business School calls this the “human enablement gap.”
“The limiting factor is increasingly human readiness, not technical capability. AI value depends on whether people have the confidence, connection, and judgment to integrate it into work.”
The phrase is useful but understates the problem. This is not only about capability. It is about the psychological readiness to lead through a transition that, for many people, is also a transition in how they understand themselves.
A note from KDVI
Most research on AI and leadership asks what leaders should do. We have been asking what AI is doing to them, to their role, sense of authority, their professional identity, and their experience of leading through sustained uncertainty. This newsletter draws on what we are hearing in practice and previews a more substantial piece of work.
KDVI is conducting an in-depth research project drawing on conversations with senior leaders, HR directors and L&D professionals across global organisations, grounded in the psychodynamic tradition at the heart of our work. The findings will be published in Autumn 2026. If you would like to be among the first to receive the report or to contribute your own leader perspective as an interview participant, register your interest.
A psychological event, not only a strategic one
AI arrives in most organisations as a strategic initiative. It also arrives as something more unsettling: a challenge to what people thought they knew about their own competence. The psychological texture of leading through an AI transition is not well captured by surveys, but some numbers are suggestive.
express concern about AI at work. Leaders are rarely given space to acknowledge the same thing. (EY)
fear losing their role within two years if they fail to deliver AI gains. Those projecting confidence are carrying the same anxiety as their teams. (Dataiku)
believe an AI agent could counsel as well as a human board member. This says less about AI than about how senior leaders currently feel about themselves. (Dataiku)
The leaders being asked to project confidence are often carrying the same anxiety as everyone else, with the added expectation that they keep it to themselves. Beneath that composed exterior, four things tend to happen consistently.
Impact on identity, introspection, agency
For many professionals, expertise is not only a skill set. It is how they understand themselves: what they have earned, what gives them standing. When AI replicates their expertise, even partially, the question raised is not only practical but existential. Researchers call this the “algorithmic self”: where AI starts “mediating self-knowledge, not merely shaping what we do, but also who we become and the narratives we narrate to ourselves.” (Frontiers in Psychology, 2025)
Resistance masking grief
What reads as obstruction is often something harder to name: watching a tool reproduce their expertise, without the accumulated judgment from the years and missteps it took to earn them. That is not irrationality. It is a reasonable response to a genuine loss and closer to grief. (HBR, 2026)
Trust erosion
When people feel that AI is being introduced around them rather than with them, the response is rarely open resistance. It tends to look like adoption and feel like withdrawal. The gap between the two can be difficult to see from the top of an organisation, and expensive to discover later. (WEF, 2025)
Expertise atrophy
Judgment develops through difficulty. When AI removes difficulty too completely, development stops. A tipping point may be reached where it no longer makes sense to rely on human decision-making. Critical thinking, contextual judgment, the ability to read a room, social care and leadership itself can all gradually erode when the conditions that built them are no longer there. (Krook, 2025)
Three demands, one leader
AI does not simply add one item to the leadership agenda. It creates three simultaneous demands. The support structures around leaders have caught up with the first. The other two are largely unaddressed.
1: Operational task
Deploy AI responsibly and create value
Deploying tools, rebuilding processes, establishing accountability. Almost all the budget and external support sit here. It is also the easiest to present to a board.
2: Identity task
Help people preserve and renew professional meaning
When roles are disrupted, people need more than retraining. They need a leader who can name what has changed and help them find what remains worth doing. Without that, adaptation is performed rather than real.
3: Containment task
Hold anxiety long enough for people to think, learn, and adapt
Creating enough steadiness that people can think rather than react. The more anxious people become, the less they can reflect. This is the function that tends to fail first when leaders themselves have nowhere to process their own concerns.
The second and third tasks are where leaders most often find themselves without adequate support. They are also, as HBR has noted, precisely the dimensions of leadership that AI cannot replicate. It is worth asking why, given that, they receive so little attention.
The middle is squeezed hardest
Support is not distributed evenly. Senior leaders have visibility and resources. Front-line staff get tools and training. Middle managers get both ends at once: the strategic message coming down, the human reaction coming up, their own uncertainty running in both directions, and the work still to be done through it all. Their role is also, in many organisations, the one most threatened, not by redundancy but by the erosion of what gave it meaning: the judgment calls, the mentoring, the contextual decisions that used to be theirs. Research consistently identifies middle managers as occupying the hardest translation role in AI transformation. They are also the least supported.
A related problem goes unnamed. When junior staff are enthusiastic about AI and senior leaders defer to that enthusiasm, institutional memory starts to look like a liability. The ability to read an organisation’s real dynamics, to know what has already been tried, to anticipate what will fail, none of this shows up in an efficiency metric. Its disappearance will not be immediately visible either.
What is underneath
KDVI pays attention to what is not said in professional settings: The anxiety beneath a composed exterior, the embarrassment of not understanding something the organisation has decided is obvious, the habit of deferring to AI output when a leader has stopped trusting their own judgment. These are not peripheral experiences. They surface repeatedly in our conversations with leaders, and they are often the actual explanation for why well-resourced transformations produce performance of change rather than change itself.
McKinsey’s research on “inside-out leadership” points the same way: in a transition like this, the internal work comes before the operational. What leaders carry inside, their assumptions about their own authority, their tolerance for uncertainty, their capacity to hold anxiety without discharging it, shapes everything that happens around them, whether or not that connection is made explicit.
None of this is an argument against AI. It is an argument for a wider lens, one that takes seriously what leaders are carrying internally, and creates the conditions for honest conversations rather than performed ones.
Questions worth sitting with
— Do your AI communications name what people are actually worried about, or do they offer a confidence that rings hollow?
— Is your transformation protecting the conditions under which junior people develop real judgment or is it removing the difficulties that make development possible?
— What support are you offering the managers in the middle who are asked to hold both the strategic pressure and human distress simultaneously?
— And what, if you are honest with yourself, is this moment of AI transformation requiring of you? Not the answer you would give in a meeting, but what comes up when you genuinely try and answer the question?
These are not questions with tidy answers. The capacity to sit with them seriously, rather than route around them, is a fair measure of whether an organisation is leading its AI transition or merely administering it.
KDVI works with executive teams and individual leaders on this territory.
To start a conversation, reach out at kdvi.com
Coming Autumn 2026: our research report on AI and the human side of leadership.
Further reading
Where Senior Leaders Are Struggling with AI Adoption — HBR, 2026
The Best Leaders Can’t Be Replaced by AI — HBR, 2024
Redefining Leadership in the Age of AI — Henley Business School
Inside-Out Leadership — McKinsey, 2024
The Algorithmic Self: How AI Is Reshaping Identity and Agency — Frontiers in Psychology, 2025