There is a compelling fable of the AI mission that goes like this: the technology arrives, people use it, productivity improves, and the organisation becomes something more capable than it was. It is a useful narrative. And in most organisations, it is not what happens.
What actually happens is more complicated and, for many leaders, results are hard to achieve despite the promises of the technology. A company AI account is set up. Some people use it enthusiastically. Others wait to be told. A few resist quietly. The results are patchy and reporting on them becomes difficult as leaders confuse the use of AI with the output from the AI.
Pockets of genuine improvement sit alongside teams that have barely changed how they work. And somewhere beneath all of it, a cultural question that nobody quite asked at the start remains unanswered: what kind of organisation do we need to be for this to work?
The data on this is striking. Deloitte's 2026 Human Capital Trends survey found that 65% of organisations believe their culture needs to change significantly in response to AI — yet only 5% report making meaningful progress on that change. A separate 2026 survey of over 2,400 executives and employees found that 79% of organisations face real challenges in AI adoption, with 54% of C-suite leaders acknowledging the transition is creating significant internal fracture (Writer, 2026). The gap between AI investment and AI transformation is, at its core, a cultural gap.
Technology, in other words, is not the hard part. Cultural change is. And cultural change is, above all, a leadership challenge.
Edgar Schein's model of organisational culture — which remains the most practically useful framework for leaders navigating change — identifies three levels at which culture operates. The visible level: the artefacts, the rituals, the way meetings are run and how decisions get communicated. The stated level: the values and beliefs an organisation claims to hold. And the deepest level: the underlying assumptions that actually govern how people behave, particularly under pressure.
Most AI adoption programmes operate at the first level. They introduce new tools, new workflows, new training programmes. These are necessary. They are not sufficient. The reason they so often fall short is that the deeper levels — what people believe about the role of technology in their work, whether they trust their organisation's intentions in introducing it, and whether they feel safe enough to experiment and fail — remain untouched.
A team that does not trust its leadership will not use AI transparently. It will use it privately, defensively, or not at all. Deloitte's research found that 80% of workers are concerned their colleagues are using AI to appear more productive than they actually are — a figure that reveals, beneath the surface of adoption, a culture of performance anxiety rather than genuine experimentation (Deloitte, 2026). That is a cultural problem. No amount of additional tooling solves it.
The World Economic Forum's 2026 research on AI readiness is direct on this point: successful AI adoption fundamentally necessitates behavioural adaptation, and leaders must prioritise cultural change over technical solutions (WEF, 2026). For people to change behaviour, they personally need to be convinced of the benefits to them and their work - in short, behaviour follows mindset and leaders must address the team's beliefs around AI use before mandating it.
Below we outline four components for leaders to address, specific to AI adoption:
1. Name the anxiety before it names you. In many organisations, leaders and employees are having very different conversations about AI. Leaders focus on the opportunities and efficiencies it offers, while employees may worry about job security, surveillance and whether their expertise will still be valued. This gap can quickly undermine trust. Leaders who acknowledge these concerns – make them public and are clear about what AI may change, are more likely to create the trust and psychological safety needed for successful adoption.
Edmondson's research on psychological safety is relevant here in a specific way: in teams where people do not feel safe to raise concerns or admit mistakes, critical information goes unshared. In the context of AI adoption, this means teams where safety is low will not surface the failures, the biases, or the quality concerns that good AI use depends on catching early. The leader who creates safety is not being soft. They are managing a real operational risk (Edmondson, 1999).
2. Model the behaviour you are asking for. Schein's central insight about cultural change is that leaders do not change culture by announcing new values. They change it by visibly, consistently, and imperfectly modelling new behaviours themselves. If a leader wants their organisation to use AI experimentally — to try things, share what worked and what did not, and build collective learning — they need to be seen doing exactly that. The leader who admits in a team meeting that an AI-generated output was wrong, and explains how they caught it, is doing more for cultural change than any communication campaign.
3. Distinguish between contexts that suit AI and those that do not. One of the most important things a leader can do is help their team develop genuine judgment about where AI adds value and where it introduces risk. This is not a skills question — it is a cultural one. Organisations that treat AI as a universal productivity solution, to be applied indiscriminately, produce teams that use it poorly and lose confidence when it fails. Organisations that develop shared norms about where AI is appropriate — which decisions benefit from it, which require human judgement, which carry risks that make automation unwise — are building the cultural foundations for sustainable adoption. Involving everyone in this exercise naturally increases buy-in.
4. Measure learning, not just output. The metrics most organisations apply to AI adoption — time saved, tasks automated, efficiency gains — measure the wrong thing at the wrong stage. In the early period of cultural change, the more important question is whether teams are learning: experimenting, reflecting, sharing what they discover, and adjusting. In other words, is the team's mindset supportive of these new behaviours and what detail does the data reveal that may need more conversations with the team to support their beliefs. These are cultural practices. Leaders who measure and reward them are shaping the culture that makes long-term transformation possible.
Beneath the practical questions about AI tools and workflows, there is a more fundamental question that cultural change in the context of AI requires leaders to sit with: what is the purpose of our work and do the team(s) believe in it.
This is not a philosophical indulgence. It is the question that shapes everything else — which tasks get automated, which human capabilities get developed, how performance gets measured, and what kind of organisation people are being asked to be part of. Organisations that answer it thoughtfully, and that bring their people into that conversation rather than announcing the answer from above, tend to be the ones where cultural change takes hold rather than fades.
References
Deloitte. (2026). AI and cultural debt. Deloitte Insights Human Capital Trends 2026.
Edmondson, A. C. (1999). Psychological safety and learning behavior in work teams. Administrative Science Quarterly, 44(2), 350–383.