Artificial intelligence has become healthcare’s latest “whizz-bang” technology, dominating boardroom conversations, filling conference agendas and prompting increasingly ambitious predictions about how services could be transformed. Yet behind that excitement, the question of AI readiness in healthcare is becoming just as important as the technology itself.
The excitement is understandable. Applied well, AI could improve screening programmes, reduce administrative pressure, and give overstretched healthcare professionals more time with patients.
However, throughout my career, I have learned that successful transformation is rarely determined by the sophistication of the technology introduced. The visible innovation attracts the attention, but it is the less glamorous work beneath the surface around data, infrastructure, governance, processes and people, that decides whether the transformation succeeds.
I have worked in enterprise architecture across government, financial services, fraud prevention and healthcare, and what has always fascinated me is never what a technology can do. It is how organisations bring together their people, processes, information and systems to create something that works as a whole. Technology might perform brilliantly in isolation, but if it didn’t connect with the wider service, support the people using it or solve a clearly understood problem, it rarely delivered what had been promised.
That is why, amid all the enthusiasm surrounding AI, I believe healthcare needs to pause and ask a more fundamental question. Are the foundations ready to support it? In other words, what does AI readiness in healthcare really require?
A Ferrari still needs a road
AI is often presented as though introducing the technology is synonymous with transforming a service. It is an appealing idea because everybody wants the Ferrari, something sophisticated, powerful and capable of moving faster than what came before. But there is little value in putting a Ferrari in a stable. It cannot do its job without a road to travel on, a clear destination and somebody equipped to drive it.
The exact same principle applies to AI. Before an organisation can expect it to deliver meaningful value, the relevant data must be available, accessible and trustworthy. The infrastructure must be secure and resilient, the governance arrangements must be clear, and those responsible need to understand what is entering the system, what is coming back out and how that output should be used.
The biggest barrier is the fragmented nature of healthcare data. Information is stored in different places, captured in different formats and managed under different governance arrangements. The data may exist, but that doesn’t mean it can be accessed, shared or interpreted consistently, yet here we are asking AI to draw reliable conclusions from that fragmented landscape and no model can compensate for information that is unsuitable for the job it is being asked to do.
So, before asking what AI can do for healthcare, organisations need to ask whether they have done the difficult work of making their data usable. That means addressing standards, interoperability, information sharing, quality and accountability, while developing a shared understanding of how information moves across organisational and geographical boundaries.
This will never generate the excitement of a new AI tool, but it is far more consequential. Without it, healthcare risks building increasingly impressive capabilities on foundations that cannot take the weight.
Start with the problem and bring people with you
The fixation on technology also leads organisations to begin in the wrong place. Too often, the starting point is an impressive AI product followed by a search for somewhere to deploy it. The more useful approach is to begin with the healthcare challenge, understand the service surrounding it and only then decide whether AI has a role to play.
This matters because every activity in healthcare forms part of a wider pathway, and decisions made in one place create consequences elsewhere. An AI tool might generate an excellent clinical recommendation, but if it cannot reach the right professional or fit the existing workflow, its technical performance will translate into very little practical value.
Some of the least successful programmes I have seen involved technology that worked exactly as designed. They failed because the ecosystem around it was never redesigned.
For all the discussion about models, platforms and technical capability, I always return to the same conclusion. AI adoption is primarily a people challenge. A system will not improve care simply because it has been installed. Healthcare professionals need to understand when its output can be trusted and when human judgement must take precedence. They need training, time to adapt and ongoing support, rather than being expected to absorb another technology into an already pressured working day.
Trust also depends on visible governance. Organisations must be able to explain who is accountable for an AI system, how its performance is monitored, and what happens when its output is challenged.
We should also be honest that our tolerance for risk is not always consistent. An organisation might reject an AI system because it is “only” 87 per cent accurate while accepting a manual process that delivers far lower consistency. That is not an argument for adopting the AI. It is an argument for a fairer comparison with what it would replace.
Patients must be part of this conversation too. AI is still associated with scams, privacy breaches and the erosion of human contact. Most do not care about AI for its own sake, nor should they. They care whether it will get them an appointment sooner or an earlier diagnosis. If organisations cannot clearly explain what AI is being used for and what difference it will make, they will struggle to build the trust adoption depends on.
Building the conditions for meaningful change
What gives me confidence is that these conversations across healthcare are beginning to mature. Events like AireFest are valuable because they create space to ask the harder questions about use cases, data readiness, governance and workforce capability. The most useful conversations are rarely sales pitches about silver bullets. They happen when clinicians, architects, researchers, suppliers, policymakers and frontline professionals share what is and is not working and challenge one another’s assumptions.
I am optimistic about the role AI can play in healthcare, but the organisations that succeed will not be those that adopt every new capability first. They will be the ones that get their data into better shape, strengthen their governance, and prepare their workforce to use new tools confidently while staying focused on the problems they are trying to solve and the people whose lives should be improved as a result. That is the real test of AI readiness in healthcare. The Ferrari may attract the attention, but it is the road, the driver and the destination that determine whether it gets anywhere.
By Vicky Rothwell, Lead Enterprise Architect at Aire Logic
