AI may change what healthcare can do. It will not, by itself, make healthcare work better. That depends on the decisions we make about people.
You cannot close a hospital for six months while you work out the new operating model. Reform has to happen with the lights on.
We have become rather good at announcing change.
We have programmes, milestones, dashboards and, if we are very lucky, a new acronym. But changing a structure is not the same as improving a service.
The test is not whether a new system went live on Tuesday. It is whether something is better on Wednesday: whether a patient is seen sooner, whether a clinician has more time to care, whether a manager can spot a problem before it becomes a crisis.
Performance is not something you extract from people. It is something you build with them.
That means taking management seriously. It means designing jobs properly. It means giving people the skills, the confidence and the space to exercise judgement.
We should be ambitious about the technology. We should also be demanding about the evidence.
There is a great deal of excitement about AI, and some of it is justified. It may help people find information faster, reduce repetitive administration, support planning and give professionals better tools for difficult decisions.
But there is a difference between being leading edge and bleeding edge. And in healthcare, the consequences of confusing the two can be serious.
What is it good at? Where does it make mistakes? Who notices those mistakes? And who is accountable when a recommendation turns out to be wrong?
If AI saves someone ten minutes on a task, that is potentially useful. But if those ten minutes simply disappear into another queue, another report or another demand on their time, we have not necessarily improved the service.
I do not say that to argue against AI. Quite the opposite. I want us to get the benefit. But the benefit does not arrive in the box with the software licence. It arrives when we redesign the work around what the technology can genuinely do, and around what human beings still need to do better.
It feels rather obvious, but managers have to manage.
And we have to give them the capability to do it. It is not reasonable to ask a manager to introduce unfamiliar technology, support anxious colleagues, maintain safe services and improve performance, and then give them neither time nor training to do any of those things well.
We need to be honest about skills. Not just whether someone can operate a system, but whether they understand its limitations, know when to challenge it and have the confidence to say, “That does not look right.”
Professional judgement is not an inconvenient obstacle to efficiency. It is one of the things we are trying to protect and strengthen.
And we need to involve staff early. Not as a final consultation after the big decisions have been made, but while there is still a genuine choice to make.
The seat is not the point.
For years we have talked about wanting a seat at the table. I think we should stop talking about the seat. The point is whether we make a difference to the decisions being taken around it.
If an organisation is considering a new AI system, I do not want its people leaders brought in at the end to write the communications plan and deal with the fallout. I want you there when the organisation decides what it is trying to achieve, which work will change, what capability it will need, and how it will know whether the decision has succeeded.
Compliance matters. Of course it does. But compliance is the floor. It is not the ambition.
None of these challenges can be solved by the people profession alone. But very few can be solved without it.