Originally featured in the COBIS Supporting Associate Newsletter
The schools most at risk from AI are not the ones that have done nothing. They are the ones that have ticked the AI boxes – but for the wrong reasons. According to recent research, 2024 and 2025 were the ‘panic and pilot’ years for AI in schools – adoption driven not by clear educational purpose but by pressure: from policy, from peers, from headlines, and from the expectation of governors and parents that schools should be seen to be moving. Nearly 60% of principals now report using AI in their workflows, yet only half of schools have provided any meaningful training to support it. The tools arrived before the thinking did. MIT research suggests that 95% of AI pilots never progress beyond the trial stage. The activity is real. The transformation, in most cases, is not.
The result, in many schools, is what we might call AI washing: existing systems rebadged as ‘AI-powered’, a new policy on the website, a slot on the governors’ agenda. This is driven in part by the human tendency to prioritise visible wins today over the harder, slower work of building genuine capability. Governors want progress they can see. Parents want reassurance. So leaders deliver the appearance of both. This is not a technology problem. It is a mindset problem – and Carol Dweck identified it decades before AI arrived.
It is a challenge we have worked through directly with COBIS schools, including the senior leadership team at the British School of Amsterdam – not on AI adoption specifically, but on building the culture and mindset that makes genuine transformation of any kind possible.
This article focuses specifically on the leadership layer – the decisions, culture, and mindset of those responsible for how AI is introduced into a school – rather than how teachers use it in the classroom or how it is deployed with students.
The fixed mindset trap in AI doesn’t look the way most leaders expect. The problem isn’t that fixed mindset leaders reject AI. It’s that they adopt it superficially. They are driven by what Dweck calls ‘proving myself’ rather than ‘improving myself’: the desire to demonstrate competence rather than develop it. In a school context this means spinning up a pilot, appointing an AI lead, and presenting progress to the governing body – “Look, it’s working. Aren’t we ahead of the curve?” But nothing has been built beneath the surface: not the culture, not the skills, not the honest conversation about what specific problem AI is actually solving. The pilot cannot scale because it was never designed to. It was designed to impress.
Research on professional identity helps explain why. A 2026 study by Meltwater Consulting found that professional identity threat is the single strongest psychological driver of poor-quality AI adoption. When leaders feel that AI undermines their credibility – their standing as the person who makes sound judgements – the instinctive response is not to sit with that discomfort but to do something visible. Write the policy. Procure the tool. Put it on the agenda. What doesn’t happen is the harder question: “what are we actually trying to achieve, and is any of this getting us there?”
This shows up in a pattern most school leaders will recognise. In the SLT meeting, terms like RAG, LLM, or agentic AI get used with confidence – and nobody asks what they mean, because nobody wants to be seen as behind. There is an unspoken expectation that school leaders should know what AI means for their context. So everyone nods. The meeting moves on. And the school’s AI strategy gets built on a foundation of shared uncertainty that nobody has named. It is one of the most reliable signs that a school’s AI culture is built on anxiety rather than curiosity. And anxiety, as any school leader knows, is not a foundation for learning.
Growth mindset adoption starts with a different question entirely: not “how do we show we’re doing this?” but two harder questions: “what specific problem are we trying to solve”, and “what foundations do we need to have in place”? Those questions are harder to ask when you are under pressure to demonstrate progress. But they are the questions that lead somewhere worth going.
I nearly got this wrong myself. When the AI wave hit, my instinct was to act immediately – sign up for the courses, bring in the consultants, build a visible plan. Until I recognised the feeling for what it was: FOMO. An unconscious bias dressed up as leadership. I was about to adopt AI not because I understood what problem it was solving, but because I was afraid of being left behind. That is a fixed mindset response – proving myself rather than improving myself. And it was operating in someone who teaches this stuff for a living. If it can happen to me, it can happen to anyone.
The consequences of getting this wrong are more tangible than most schools realise. Research published in the Harvard Business Review in May 2026 identifies a phenomenon called workslop – AI-generated output that looks plausible on the surface but is hollow or riddled with errors. Forty percent of workers surveyed had received workslop from colleagues in the past month. Think of someone in the finance team who uses AI to produce a budget summary for the governors’ meeting – it looks clean and well-structured, but the assumptions are wrong, the context is missing, and someone
senior has to spend the afternoon fixing it before it can go out. The time saved upstream has been spent twice over downstream.
A 2026 survey found that 32% of employees are experiencing burnout specifically from checking AI output, and 66% say AI has not made them faster. Badly introduced AI doesn’t reduce workload. It redistributes it – invisibly, and usually to the people who can least afford it.
Whatever the official position, the reality in most schools is that AI is already being used – just not always through approved channels. Staff use tools on personal devices, outside sanctioned systems, to solve problems they can’t solve adequately with the tools the institution has provided. This is what researchers call shadow AI, and most IT departments treat it as a compliance problem. But from a leadership perspective it tells a more interesting story.
The head of department drafting reports with an unsanctioned tool. The teacher running lesson ideas through a chatbot at home. The SENCo using AI to draft communications the school hasn’t yet approved. These are not rogue actors. They are people with real problems finding real solutions, doing so despite the institution, not because of it. A fixed mindset leader shuts this down – data security, consistency, liability, all legitimate concerns. A growth mindset leader asks a different question first: “why is this happening, and what does it tell me about what my people actually need”? A school with significant shadow AI may have more genuine engagement with AI than its official figures suggest. It has simply been driven underground by a culture that didn’t make space for it. Understanding that gap – between what has been provided and what people are actually doing – is itself a form of growth mindset in action.
None of this – not shadow AI, not workslop, not the shared uncertainty in the SLT meeting – gets addressed without one thing: ‘psychological safety’. The conditions in which a senior leader can say “I don’t know what that term means” without it damaging their standing. In which a teacher can try an AI tool, produce something that doesn’t work, and treat it as useful information rather than a professional embarrassment. In which the conversation about AI is honest rather than shaped by what people think they are supposed to say.
Psychological safety is not a soft precondition for AI adoption. It is the precondition. Without it, staff say the right things in the right meetings and carry on as before. The gap between what schools report to governors and what is actually happening in classrooms widens. And the accumulated cost – what HBR’s Guy Champniss calls psychological debt – surfaces eventually as burnout, attrition, and a deep scepticism about the next initiative that comes along.
When I work with leadership teams, I often start with something simple: I ask everyone to write down one thing about AI they don’t understand and have never felt able to ask. We read them out anonymously. The relief in the room is always immediate – and almost always, the questions are the same. That moment of shared honesty is where genuine AI adoption begins.
During her tenure as Principal of the British School of Amsterdam, Ruth Sanderson began with a clear purpose: to build a senior leadership team capable of nurturing, inspiring, and empowering the whole school community. Not a vague aspiration, but a specific vision of what the school needed to become and what kind of leadership that required. That clarity of purpose was what made everything else possible.
Working with Teamworking International over an intensive twelve-week programme, the SLT built the psychological safety to be honest, creative, and genuinely collaborative. One member described it as “a common language that created closeness and understanding.” People who had dreaded leadership meetings began to look forward to them. Voices that had been quiet began to be heard. Ruth’s conclusion was unambiguous: “It has been transformational for the whole organisation. It’s the best team training I have ever done.”
That work wasn’t about AI. But it built precisely the culture that AI adoption requires: a team that begins with purpose, that is secure enough to say “I don’t understand this”, curious enough to experiment, and resilient enough to fail without it threatening anyone’s standing. Purpose first. Safety second. Tools third.
The schools that will navigate AI well are not necessarily the ones that move fastest. They are the ones whose leaders ask the right question – not “how do we show we’re doing this?” but “what are we trying to achieve, and what does our school community need us to become?”
That is growth mindset in action. Not the policy on the website. The culture in the room.
For education leaders who want to understand how their own mindset is shaping their school’s relationship with AI, a Harrison assessment offers a surprisingly revealing starting point — not a personality test, but a practical map of how you think, decide, and lead under pressure. We offer a complimentary trial assessment with a 30-minute debrief.
Get in touch and become an even better leader.
To continue the conversation with Tom, please click here or book a time directly with my PA Ilhaam, at ilhaam@team-working.com and she’ll make it happen.
Tom Flatau is CEO of Teamworking International, applying neuroscience to leadership development. To read the British School of Amsterdam case study in full, visit https://team-working.com/case-studies/the-british-school-of-amsterdam
Yes I want to receive advice and news on how neuroscience can help me succeed in business.
Once payment has completed, please click the "Return to Seller" button to continue.