Independent expert in healthcare AI, Alex Fairweather explains why leadership is key when building a modular AI architecture for healthcare.
“You cannot build a great building on a weak foundation,” said US religious leader Gordon B. Hinckley. A pattern is emerging across healthcare organisations: a growing collection of AI tools, each delivering a degree of value in isolation, but failing to add up to anything approaching enterprise-wide transformation. Each is a point solution to a local problem. Together, they create a new layer of fragmentation on top of an already fragmented system.
McKinsey identified this dynamic explicitly in its 2025 healthcare AI outlook. “The rush toward point solutions has created a fragmented AI environment and new operational friction,” it said. The consequence? “Unless leaders course correct now, they will merely automate today’s inefficiencies.”
This is a fundamental strategic challenge calling for strategy and leadership as much as technology expertise.
What a modular architecture looks like
McKinsey’s alternative is a modular, connected AI architecture. A framework in which AI point solutions, data infrastructure and intelligent agents are brought together into a coherent whole. The model rests on three components: domain-specific AI models trained on clinical data rather than general internet text; a connected data layer that enables these models to share information in real time; and intelligent agentic AI that can coordinate workflows across the system.
Critically, this architecture is about designing the connective tissue between them. The US Centers for Medicare and Medicaid Services’ interoperability mandates, alongside the formation of new data-sharing frameworks, are reinforcing this shift. In the UK, the NHS interoperability mandate and the establishment of secure data environments create a comparable regulatory and structural context. The goal is open, interoperable architecture that enables AI to operate at the level of the whole system, not just the individual application.
The modular architecture McKinsey describes treats patient data as a strategic asset, which they call a clinical data foundry. When patient records and multiple personal clinical data points are integrated into a single, governed data layer, the AI applications built on top of that layer become exponentially more capable.
This connects directly to themes I have explored in previous articles in this series. The NHS initiatives such as Nightingale AI and Foresight – the generative AI model trained on de-identified data from 57 million patients – are early examples of what is possible when health data is treated as a collective resource for innovation.
For private healthcare providers, the challenge is to build the data foundations now that will make this insight available at the level of the individual organisation.
Leadership as the decisive factor
Technology architecture alone does not deliver transformation. The organisations that successfully navigate from pilots to platforms will do so because their leaders made deliberate, enterprise-wide strategic choices, not because they adopted the best individual tools. McKinsey’s 2025 State of AI survey found that nearly two-thirds of organisations have not yet begun scaling AI across the enterprise. The gap between experimenting and transforming is real, and it is primarily a leadership gap.
Capgemini’s 2025 research confirms that gen AI adoption has surged from 6% of organisations in 2023 to 30% in 2025 but that scaling sustainably requires investment in governance, training and change management, not just technology procurement. For healthcare leaders in the UK, the message is consistent with everything this series has argued from the beginning: the technology is ready.
The question is whether the organisational conditions of leadership, strategy, governance and training for deploying it responsibly are in place.
