Scaling what works: the next phase of AI in the NHS

Guest blog by Meena Sebastian, Healthcare Lead at NEC Software Solutions

AI has become one of the NHS's biggest opportunities, and for good reason.

From helping detect signs of disease in clinical imaging to recording and analysing case notes, its potential is becoming increasingly clear.

Much of the conversation over the past year has focused on Ambient Voice Technologies (AVTs), given the productivity benefits in helping clinicians spend less time on administration and more time with patients. However, as the push for AI adoption gathers pace – in accordance with the necessary governance that is evolving - it’s important for the true scale of its potential in the NHS not to be overlooked.

Diabetic eye screening is a perfect example of the broader benefits offered by AI. NHS organisations are exploring AI-enabled grading to help clinicians make “disease/no-disease” decisions more efficiently and consistently. Used appropriately, it could reduce some of the administrative burden surrounding screening programmes while supporting clinical judgement and speeding up diagnosis for patients. But, like many promising AI use cases, these deployments often begin on a relatively small scale, as pilots.

Creating the right foundations

Positive steps are being made to address the issues of ‘pilotitis’. Recent publications, such as the MHRA’s reports on the use of AI in healthcare - which will inform the new AI Commission’s recommendations later in the summer - show broad support for the continued use of AI in the NHS. Rightly so, there was also a strong emphasis on the need for a balance between robust regulation and the ability to adopt innovation quickly enough for it to benefit patients.

In tackling this dichotomy, it’s important that the Commission addresses the risk of embedding existing inequalities into the algorithms that are being built. AI models are only as reliable as the data they are trained on, and healthcare data is not always representative across every population group or condition.

In practical terms, this means ensuring diversity in training datasets, transparency in how algorithms are validated, and representation from across the health ecosystem, including minorities in leadership roles, the workforce and patient groups, in decision-making. Otherwise, we’ll end up automating the inequalities we’ve spent decades trying to fix.

User centred design

Another ingredient for success is user centred design. Clinicians don’t have time to learn another platform or log into another system. Therefore, solutions that succeed at scale will be the ones that slot quietly into the process, improving it rather than adding friction.

For example, the planned use of AI in the “disease/no-disease” grading system in diabetic eye screening services has been designed with the users’ needs in mind – to support clinicians rather than replacing their role in a safe and effective way. It can give them time back, reduce human error, and help bake-in evidence-led decision making.

This is where partnerships and user-centred design matter. By co-designing with the NHS – to understand the patient journey, work within existing pathways, and evolve solutions based on real-world use. It’s what makes the difference between a clever tool and a sustainable service.

Scaling what works

The diabetic eye screening programme is one example of many, but it shows what’s possible when AI is used to augment, not replace, clinical expertise.

These use cases are “anchors in the sea” that prove how AI (not just AVT) can adhere to the necessary regulation, integrate into workflows and be used by clinicians, and improve patient outcomes.

The next phase must be about scaling what works, sharing data responsibly, and keeping ethics at the heart of every algorithm.

The AI Commission could be a powerful catalyst for that change if it listens to those on the ground and acts on what the data tells us, alongside what the workforce needs and vendor partners can do to help.

We don’t need a revolution - we need a consistent, evidence-based evolution.

Health and Social Care Programme activities

techUK is helping its members navigate the complex space of digital health in the UK to ensure our NHS and social care sector is prepared for the challenges of the future. We help validate new ideas and build impactful strategies, ultimately ensuring that members are market-ready. Visit the programme page here.