Duncan Light, MBBS MSc FRCS
DOI: 10.5281/zenodo.20213750
Artificial intelligence is increasingly presented as an important part of the NHS's future. Its potential benefits are now widely recognised across both clinical care and service delivery. Successful adoption depends on whether the NHS is prepared to implement these tools effectively when many organisations still rely on ageing computers, limited interoperability and fragile digital infrastructure.¹⁻³
Digital health products must face regulatory approval, including validation and evidence generation. This creates delay between development and adoption, along with the need for significant investment. These conditions favour larger companies over smaller startups, with a risk that more disruptive forms of innovation are lost. As a large public healthcare provider, the NHS is well placed to support startup innovation in the healthcare space from internal developers, entrepreneurs and external parties.
A snapshot survey conducted by the Royal College of Physicians found that 70% of physicians were either very or somewhat supportive of the widespread implementation of AI tools in the NHS. At the same time, 68% reported that the NHS lacked the digital infrastructure required to introduce AI appropriately, while 48% strongly disagreed that the service was ready to integrate the technology. The same survey found that 70% considered poor integration with existing systems, including the electronic patient record, to be the leading barrier to implementation.² The British Medical Association has estimated that more than 13.5 million working hours are lost each year in England alone because of inadequate IT systems and equipment.³ This represents a substantial loss of clinical time within a service already under considerable pressure. If routine care continues to be hindered by outdated hardware, it is difficult to assume that advanced AI applications will integrate smoothly into practice or deliver their intended benefits.
In healthcare, AI is not simply a software product. Its value depends on integration, governance and staff confidence. Mistry has argued that implementation often fails because investment in technology is not matched by investment in the conditions required for effective use.¹ From this perspective, the responsibilities of suppliers cannot reasonably be regarded as ending at the point of procurement.
In terms of training, the Royal College of Physicians reported that 79% of doctors believed they required training in clinical AI tools, while 66% had no access to such support.² This is not a marginal issue, given its direct implications for safety, judgement and adoption. There is limited value in introducing AI into clinical environments if staff are left to develop familiarity through informal experimentation alone. A related issue arises in relation to patients, who must also be supported through clear information and inclusive digital engagement.¹ Otherwise, adoption may reinforce existing inequalities. The introduction of Copilot into the NHS has been welcomed in principle, but there has been criticism that staff have not been given sufficient training or support for its use.⁴
Clinician involvement is vital for medtech innovation generally, but it is particularly important in relation to AI. Tools developed without meaningful clinical input frequently fail at the point of use because workflow fit is poor and the original problem has been inadequately defined. The Royal College of Physicians has called for clinicians and patients to be involved from the outset so that digital and AI tools address real clinical challenges and improve workflows.² This should be regarded as a basic standard of responsible innovation.
Companies such as Microsoft, OpenAI, Anthropic and major hardware providers are likely to benefit substantially from the growth of AI in healthcare. There remains significant development and validation to be performed for large language models (LLMs) to become ubiquitous in healthcare. An important question is whether innovation in healthcare AI will remain open to smaller companies, or whether it will become concentrated within a small number of well-funded firms. This may be particularly relevant where there is platform dominance, for example through EPR vendors or enterprise AI tools. This makes it important that stakeholders in the NHS and other health systems provide feedback and insight on real-world needs.
Large technology companies possess the technical and financial resources to do more than market products into a publicly funded system operating under sustained pressure. In this context, corporate responsibility should extend beyond pilot schemes and product access. It should include support for digital readiness. This position is also consistent with policy commentary suggesting that the priority for UK healthcare is implementation support rather than further AI regulation.⁵ If large technology firms wish to be regarded as genuine partners in healthcare transformation, their responsibilities must extend beyond sales alone. In the NHS context, this means contributing not only to the technologies of the future but also to the foundations required to make those technologies usable.
Conflict of interest statement: The author declares no conflict of interest.
Corresponding author: Duncan Light, duncanmlight@gmail.com
References
- Mistry P. Infrastructure for innovation: getting the NHS and social care ready for AI. The King's Fund. 25 June 2025. Available from: https://www.kingsfund.org.uk/insight-and-analysis/long-reads/infrastructure-nhs-social-care-ai
- Royal College of Physicians. The NHS is fundamentally unprepared for AI: 7 in 10 doctors say the NHS is not digitally fit to deploy it. RCP. 15 January 2026. Available from: https://www.rcp.ac.uk/news-and-media/news-and-opinion/the-nhs-is-fundamentally-unprepared-for-ai-7-in-10-doctors-say-the-nhs-is-not-digitally-fit-to-deploy-it/
- British Medical Association. Building the Future: Getting IT Right. London: BMA; 2022. Available from: https://www.bma.org.uk/media/6578/bma-infrastructure-2-report-getting-it-right-dec-2022.pdf
- Flashman C. Copilot has arrived in the NHS — But no one told us how to fly it! Nov 2025. Available from: https://www.pslhub.org/learn/digital-health-and-care-service-provision/288_artificial-intelligence/380_large-language-models-llms-and-generative-ai/copilot-has-arrived-in-the-nhs
- Curia. Regulating AI for the NHS: consultation response. PoliticsUK. 2026. Available from: https://politicsuk.com/news/regulating-ai-curia-mhra-consultation-response/