AI can improve healthcare, but realising its potential depends on shared learning across adopters and strong coordination between policy, innovation, and practice. That was the clear message of an independent evaluation that my University of Edinburgh colleagues and I, with partners NHS Arden & GEM, conducted of the £143.5 million UK programme to support the safe and effective use of artificial intelligence across the National Health Service, the NHS AI Lab.
Our evaluation assessed how the AI Lab was implemented, its outcomes, impacts and value, and lessons learned. The assessment included reviewing over 80 funded AI development projects to understand what enabled real-world impact and what held it back. Some projects showed significant improvements in effectiveness and efficiency in care delivery. In stroke care, one tool helped hospitals increase the use of mechanical thrombectomy, a procedure to remove blood clots from the brain. Sites using the AI reached an average of 5.7%, compared to a national rate of 3.6%. Some exceeded the NHS Long Term Plan’s 10% target, up from a baseline of just 1% in 2019, by enabling faster decision-making and better coordination.
In radiology, another system changed clinical decisions in nearly 8% of lung nodule cases. In most of those, clinicians would likely have dismissed the findings without the tool’s input. It helped less experienced staff follow national guidelines more confidently, improving consistency and quality of care.
However, despite clear benefits and significant investment, scaling AI solutions within the health and care sector presents unique challenges beyond the pilot stage. The barriers weren’t technical. They were systemic, requiring coordination across policy, practice and innovation. Success depends not just on the tools but on the infrastructure, leadership, and learning surrounding them.
As part of the review team, I focused on two areas that often receive too little attention: learning ecosystems and early-stage evaluation. Learning ecosystems are networks that support continuous learning and knowledge exchange across organisations to facilitate organisations' and sectors’ digital transformation. They connect adopters, suppliers, clinicians and system leaders. Without such ecosystems, innovation often remains siloed – confined to individual projects or local settings, with limited opportunities for shared learning or scaling and spreading.
We also saw the importance of ongoing formative evaluations from the early stages of innovation and digital transformation. Traditional evaluation methods are too rigid and slow to capture the emergent, long-term, and context-dependent nature of digital change. Instead, ongoing, adaptive approaches are essential for surfacing risks and unexpected outcomes, informing real-time decision-making, and ensuring that evolving technologies deliver sustained value across complex systems.
These aren’t new ideas, but they’re often underused. That’s why as a core member of the steering group for the newly established NHS AI Ambassador Network, I am committed to fostering cross-organisational learning and facilitating meaningful connections among stakeholders across the health and care sector. This collaborative approach is essential to effectively bridging the gap between the immense promise of AI and its practical, real-world implementation, ensuring that AI innovations are deployed to achieve improvements in healthcare outcomes.
So, what can leaders take from this?
If your organisation is investing in AI, particularly in a public or complex setting, you must treat implementation as a whole-system challenge. That means:
- Focus on digital transformation, not just automation: AI can improve efficiency, but the most impactful projects were those embedded within broader digital transformation strategies. Leaders must look beyond automating tasks to redesigning systems and processes that support better care, coordination, and decision-making.
- Facilitate innovation through cross-organisational collaboration: Scaling AI is not just a technical challenge – it requires alignment across policy, clinical practice, and technology development. Cross-organisational collaboration, including partnerships between adopters, suppliers, and system leaders, is essential to overcome fragmentation and move beyond pilot projects.
- Evaluate as you go: Make evaluation part of implementation, not an afterthought. Use it to steer decisions and support continuous improvement.
- Ensure sustained national support and strategic Leadership: AI success in healthcare depends on more than good tools – it requires long-term national investment, strong strategic leadership, and evidence-based decision-making. Consistent support at the system level helps bridge policy, practice, and innovation, enabling scalable and sustainable impact.
Artificial intelligence can support better care and more intelligent decisions but won’t transform anything independently. Technology matters, but success depends on long-term national support and strong coordination between technology developers, healthcare professionals and the wider system they work in.
Dr Hajar Mozaffar is a Senior Lecturer in Innovation and Programme Director for MSc Entrepreneurship & Innovation at the University of Edinburgh Business School, and the Director of Digital and AI Transformation (DAIT) Lab.
She also leads the School’s DAIT- Lab, which brings together researchers, industry and policymakers to explore the responsible design, use and governance of digital and AI technologies in complex organisational settings.