AI, Leadership and Creating the Space to Think

I have been thinking a great deal about AI and leadership recently. Not so much about what the technology can do, because that seems to be changing almost daily, but about what it might mean for the people who are trying to lead our health and care services now.

I have seen courses offered for specific platforms and aimed at healthcare leaders, such as healthcare professionals learning how to use Claude or AI for nurses. This may miss the point. With AI in leadership and healthcare moving so fast, perhaps it is not a course in one platform that we need, but the knowledge and tools to reduce fear and increase confident, informed use through a better awareness of what AI is and what it can do for us.

In particular, I keep coming back to our middle managers. They are often the people sitting at the most difficult point in an organisation, receiving strategy and direction from above while simultaneously trying to make it work amid the reality of services, teams, workforce shortages, budgets, quality, patient expectations and constant operational pressures that do not conveniently stop while we create space to think.

Our In the Middle of Everything course identifies this as an area where managers feel nervous and unsure about how AI can support their daily delivery and personal and professional development. This is where I see the need for knowledge.

When I talk to managers and leaders, I rarely hear that they do not know how to lead. More often, they simply do not have enough time to step back from what is immediately in front of them and really interrogate what is happening. We ask them to be strategic while their day is consumed by operational decisions. We want them to lead culture while they manage vacancies and sickness. We want them to develop people while delivering performance. Increasingly, we are also asking them to understand a new technological landscape in which AI is becoming part of how people work.

Perhaps, therefore, we are asking the wrong first question about AI. Rather than beginning with what work can AI do for us?, perhaps we should be asking: can AI help our leaders create the space to think differently about the work they already have to do?

I think it can, but only if we are very clear about what it is there to do and, equally importantly, what it must never be allowed to replace.

Using AI to challenge thinking rather than simply produce answers

One model I have found useful is the CRIT framework developed by Geoff Woods: Context, Role, Interview and Task. Woods describes CRIT as a way of using AI as a thought partner rather than simply giving it an instruction and waiting for an output. The element that particularly interests me for leadership development is the “Interview” stage, where the AI asks the leader questions before attempting to complete the task. Woods argues for using AI to deepen thinking rather than simply accelerate activity.

So, let us go back to Sarah.

Instead of telling the AI that she has a workforce problem, she begins by giving it context. She describes what she knows: sickness has risen, vacancies remain high, agency expenditure has increased, flow is deteriorating and some staff are describing themselves as exhausted. Recruitment remains difficult, and there is limited additional money available.

Notice the difference. She has not told the AI what the problem is; she has told it what is happening.

She then asks it to look at the situation through the lens of an experienced health service leader who understands workforce, operations, patient safety, organisational culture and finance. This does not turn an AI system into a workforce director, nurse leader or finance professional, and that distinction matters. What it can do is encourage consideration of the problem from more than one perspective before Sarah settles on an answer.

Then she reaches the part of CRIT that I think has real potential for managers and leaders. She asks the AI not to give her a solution at all. Instead, she says:

The conversation that follows may begin with something very simple: What evidence tells you that staffing is the primary cause of the deterioration you are seeing?

Sarah has to think about that. Yes, she has vacancies, but when she looks more closely, not every team is experiencing the same problems. Sickness is particularly high in two areas and comparatively stable in others.

The next question might explore what changed before the sickness increased. Sarah remembers a restructure six months earlier. She is then asked what the better-performing teams are doing differently, and she realises she has never really explored that because her attention has quite naturally been concentrated on the teams where performance is deteriorating.

As the questioning continues, another issue emerges. Staff have raised concerns about duplication, unclear accountability and how long it takes to make relatively straightforward decisions. Some frontline managers no longer feel confident making decisions because they are uncertain about where authority now sits following the restructure. People are escalating more, work is slowing down and frustration is growing.

Sarah may still have a workforce problem, but she can now see that this is unlikely to be simply a recruitment problem. Job design, decision-making, leadership confidence, workload, culture and process may be sitting alongside the vacancies. Her original interpretation was not necessarily wrong; it was incomplete.

AI has not solved Sarah's problem. What it has done is create enough challenge for her to look at it differently.

For me, that is where its potential becomes much more useful for leadership.

Putting productive friction back into leadership thinking

Experienced leaders are good at recognising patterns. Indeed, much of what we describe as professional judgement develops because we have seen situations before, understand context and can draw on experience quickly. That is an enormous strength, particularly in complex health environments where decisions cannot always wait for perfect information.

However, the same experience can occasionally make us move very quickly from this is what I can see to therefore this is what it means. Under pressure, the space between those two things becomes even smaller.

Used intelligently and confidently, AI can put some productive friction back into that space. It can ask what evidence supports an assumption, whether another explanation exists, whose perspective is missing, what has changed, what would disprove the current hypothesis and what could happen if the leader's first interpretation turns out to be wrong. The first step is exploring the possibilities through the use of a tool.

Those questions are not unique to AI. A good coach, mentor, supervisor, colleague or board member may ask exactly the same things. The potential advantage is that the manager can begin this process at the point when they are actually wrestling with the issue, perhaps late in the afternoon before a meeting or when trying to bring several competing pieces of information together. It becomes an additional source of challenge, not a replacement for the people around them.

There is, however, an important line we must not cross

As soon as we begin discussing AI in health and care, we have to discuss governance alongside it. I do not think the ethical and legal discussion can be something we add once people have become confident users. It has to develop at the same time as the capability.

Sarah cannot simply copy staff records, identifiable sickness information, patient details, safeguarding information or confidential organisational material into whatever generative AI application happens to be open on her computer.

NHS England's current guidance on AI and information governance states that an organisation's information governance lead, Data Protection Officer and Caldicott Guardian should be involved in decisions to implement AI or share data to develop AI technology. It also advises healthcare professionals to raise concerns about false, inconsistent or potentially biased outputs rather than assuming the technology is correct.

The same NHS England guidance makes another point that I think is fundamental to this discussion. Where AI is supporting clinical decision-making, the final decision about a person's care should be made using professional judgement and in consultation with the patient or service user. AI may support the process; it does not remove professional responsibility.

Although Sarah's scenario is principally about leadership rather than clinical decision-making, the principle travels well. If AI contributes to our analysis, we still need to be able to explain the decision that followed, understand the evidence on which it was based and be clear about who owns it.

This is also consistent with the UK Government's Data and AI Ethics Framework, updated in December 2025, which places transparency, accountability, fairness, privacy and safety at the centre of responsible public-sector use of AI and stresses the importance of appropriate human oversight. In higher-risk situations, the framework specifically describes the need for people who can identify risks, intervene and validate AI-supported outputs rather than simply accepting them.

The ICO makes a similar distinction between genuine human oversight and someone simply accepting an automated recommendation. Its guidance describes meaningful human involvement as active and critical: the person reviewing an AI-supported recommendation needs sufficient authority and competence to challenge it and, where appropriate, reach a different conclusion. Some ICO AI guidance is currently under review following the Data (Use and Access) Act 2025, so organisations need to keep their governance arrangements under review rather than treating current guidance as static.

A Healthcare Governance Gate

This is why, at Axis Culture Group, I would add something to CRIT when we bring it into health and care leadership: a Healthcare Governance Gate.

Not another complicated framework, and certainly not another checklist managers complete and forget about, but a deliberate pause in thinking. Before Sarah moves from AI-supported exploration to action, she should be able to ask herself whether what she is proposing is lawful, ethical, safe and fair, whether she can explain how she reached the decision, and where accountability sits.

• If the analysis affects staff, have we considered fairness and equality rather than simply efficiency?
• If it influences patient services, have we considered safety and the consequences for different communities?
• If information has been used, were we entitled to use it in that way?
• If AI has generated something that sounds convincing, what have we done to verify it?
• If we were challenged six months later, could somebody explain how and why the decision was reached?

Returning to Sarah

By the time Sarah eventually reaches the “Task” stage of CRIT, the task she gives AI is very different from the one she would have given at the beginning.

Rather than asking for a staffing plan, she might ask it to help her organise three plausible explanations for what she is seeing and identify the evidence that would support or challenge each. She could ask it to explore the possible workforce, quality, financial and patient implications of different responses, identify gaps in the information she currently has and suggest which people she needs to involve before reaching a conclusion.

Importantly, she can also ask it to clearly distinguish between information she has provided, assumptions that have emerged during the conversation and suggestions generated by the AI. That alone helps prevent an AI-generated idea from quietly becoming treated as an organisational fact.

Sarah still has to lead. She still needs to speak to her staff, compare the teams, look properly at the data and involve workforce, finance, HR, quality or clinical colleagues where appropriate. She may discover that the original workforce diagnosis was largely correct, or she may find that the restructure and the way decision-making changed afterwards are central to what is happening. Most likely, she will find a combination of factors.

What has changed is that she enters those conversations with a more developed understanding of the problem and, importantly, better questions.

That is what interests me.

I am less interested in whether AI can produce another action plan for an already overloaded middle manager. I am much more interested in whether it can help that manager stop for long enough to recognise that the obvious answer might not be the whole answer.

We need to develop the human capability alongside the technology

There is a danger that organisations see AI adoption principally as a technical implementation project: procure a platform, write a policy, train people to prompt it and measure whether productivity improves. I think that would be a missed opportunity.

If AI is going to become part of leadership practice, then our leaders need more than technical competence. They need curiosity, critical thinking and the confidence to challenge what appears on the screen. They need to understand bias and information governance, but they also need to recognise their own biases. They need to know when AI is useful and when the issue requires a human conversation, professional advice or simply going to the service and listening to people.

Regulatory considerations also go significantly beyond everyday generative AI use. For example, if software or AI has a medical purpose, it may fall within the UK's medical-device regulatory framework; the MHRA confirms that many software and AI products used in health and social care are regulated as medical devices. That is another reason why leaders need to distinguish between using a general AI tool to support thinking and procuring or deploying an AI-enabled healthcare technology. They are not the same governance proposition.

At Axis Culture Group, we focus on the space between technology and leadership. We should not be frightened of AI, but nor should we be seduced by its ability to produce something polished and plausible very quickly. A well-written answer is not necessarily an accurate answer, and an efficient decision is not necessarily a good decision.

CRIT gives us one way of structuring that relationship. Context stops us from jumping immediately to the solution. Role encourages us to consider the problem through a different lens. Interview lets us question our assumptions before we settle on an answer. Task helps us turn that thinking into something useful. The Healthcare Governance Gate then asks us to pause again before thought becomes action.

Perhaps that is what AI can offer our middle managers at its best. Not another voice telling them what to do, and certainly not something to which they hand over their professional judgement, but a mechanism that occasionally asks: What have you not considered yet?

In a health and care system where so many leaders are moving from one pressure to the next, creating that small amount of space to think may prove to be one of the more human uses of artificial intelligence.