Dr Richard Ovens and Philip Brocklehurst from Deloitte navigate the future of AI and the NHS
The recent Darzi report highlighted key NHS challenges across a wide range of areas, from an NHS in ‘serious trouble’ to ‘worrying health inequalities’ with growing demands placed on the system.
Advancements in digital health technologies, including artificial intelligence (AI), offer opportunities to reimagine and build health services that are fit for the future. Deloitte’s recent prediction report, Accelerating the Future, predicts what 2030 may look like – from intelligent healthcare delivery and democratisation of health data to consumers empowered as ‘CEOs’ of their own health.
AI as a technology
AI is not enough as a technology to deliver outcomes and value. Here, we can look at the ‘Gutenberg Myth’. The printing press, invented by Gutenberg in 1440, is widely considered the driver of educational improvement, coming at a time when literacy was 10% in Europe.
In reality, it was not until over 400 years later, with improvements in education availability and reduction in the cost of paper and materials, that we saw the true value of the technology take shape, a key interdependent piece of required change – with literacy rapidly climbing to 90%. We are, of course, in the modern world, but the principle still stands equally as strong. Technology in itself will not determine improved outcomes but rather how we navigate and build new ways of working.
Health literacy is a key barrier to tackling health inequalities. While useable AI technology has been around a while (although not 400 years), the major advances in technology and connectivity provide previously unrealisable opportunities to effectively embed within healthcare workflows and transform how services are designed – supporting staff and patients to navigate and engage with the health system.
AI in healthcare: Focus on value
A technology-first approach is certainly enticing, particularly in light of the recent advancements we’ve seen in AI and the advancing performance, availability, and incorporation of large language models (LLMs) into novel use cases and products. Agentic AI, a type of AI with a degree of task autonomy, is a further step towards rapid automation of tasks within healthcare – for example, electronic health record (EHR) workflow optimisation and support for a technology-agnostic approach to overcome the current mixed technology stack and operational environments we see today.
Understanding where AI fits in, and the value we seek to achieve is key. A technology-first approach can often miss the target objective and deliver no value – “If the only tool you have is a hammer, you tend to see every problem as a nail”.
The alternative, a value-first approach, supports the development of clear goals and an in-depth understanding of the problems and challenges to overcome. A persistent focus on value and an iterative process of discovery can deliver a critical understanding of the desirability (is there demand?), feasibility (can we do it?), and viability (is it worth it?) of a concept or idea.
Delivering AI for the NHS
Our work with health organisations, integrated care systems, and shared care records are good examples of success and how we use this approach to develop and implement AI to solve the NHS’s most challenging problems. Focus is consistently centred on the value we are seeking to achieve, working with multidisciplinary teams to refine the challenges and pain points – including clinicians and users, operational, and technical teams throughout.
We have built AI products, including Software as a Medical Device (SaMD), alongside a wide range of NHS and partner health organisations – from clinical triage to administrative and operational support. Expert in-house clinical, regulatory, and technical development teams deliver responsible build and governance in AI.
Our Quartz AI suite is already driving cutting-edge developments in drug discovery through Atlas AI and supporting patient-centred care through Frontline AI Teammate – a multilingual digital avatar in healthcare supporting improved communication.
Developing AI for NHS shared care records highlighted the criticality of a value-focused approach with the development of Generative AI search, document identification, and classification across unstructured health records – reducing clinician search time for key information and providing more time for patient care.
It is not always the latest AI developments that will deliver key value. Our work on NHS bookings optimisation showcased the power of iterative discovery with multidisciplinary teams to define clear definitions of value outcome, the correct technology to be used, and a pathway to success with machine learning – delivering the capability to identify under/overutilised sessions in advance and the drivers behind these for actionable change.
AI for good starts with healthy AI
Ensuring the safe and responsible development of AI in health is imperative. Early review of AI system risk in areas including privacy, security, and bias are essential hygiene factors – for example, the use of unbiased representative data in training. Indeed, being able to demonstrate responsible use of AI is essential and can be supported through validated frameworks.
As AI becomes increasingly embedded, inherent risks related to hallucination and reliance, exacerbated through the attribution of human characteristics to AI, need to be mitigated. The regulatory landscape is evolving rapidly, with compliance and an understanding of intended purpose critical concerns. Upskilling healthcare teams supports teams in improving their confidence while increasing awareness of limitations as part of risk management.
AI and the NHS: Summary
The NHS is under pressure, with rising demand and widening health inequalities. The potential use cases for AI in healthcare continue to grow and the emerging evidence suggests these can and will help tackle the current challenges and help usher in much needed improvements in efficiency, productivity, and patient outcomes.
Implementing technology alone won’t solve these complex issues. A value-driven, evidence-based approach is required. By prioritising value, collaborating with multidisciplinary teams, and directly addressing risk and ethical considerations, we can realise AI’s potential to build a more sustainable and equitable NHS for the future.
This work is licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International.
