Artificial Intelligence Scribes in Primary Care: UK–US Perspectives on Evidence, Adoption, and Policy 

I’m running late in clinic. I’ve just had a complex consult, and I have three patients waiting downstairs. I have to document the appointment that has just taken place and put the care plan into action, including by sending urgent referrals. I feel it’s not always appropriate to type away while speaking to patients, which can come at the cost of rapport. But when I document asynchronously, I am left playing catch up. 

Artificial intelligence (AI) scribes are ambient documentation assistants that can solve for this scenario. Their function goes beyond transcription to summarization and reorganization of clinical documentation that is produced in a format compatible with an electronic health record (EHR). As a United Kingdom–trained family and occupational physician now based in the United States, I have experimented with AI scribes and see the potential for them in both of my documentation-heavy specialties. Interestingly, AI adoption and regulatory environments in the UK and the US differ. Below I outline the current evidence base and compare adoption in both counties, drawing on the evidence alongside my own experience. 

The Evidence Base  

A growing body of literature highlights the efficiency gains and stakeholder satisfaction associated with using AI scribes. One study showed clinicians using AI scribes spent 20% less time on notes per appointment and 30% fewer after-hours work time per workday. In larger studies, the vast majority of physicians agree the technology has a positive impact on patient interaction, and over one-third of patients report less physician-computer interaction and more direct interaction with their clinician. Multi-center studies have identified improvements in clinician experience, including reduction in burnout scores. 

The UK Experience: Progress and Barriers  

In the UK, there are commercial AI scribe products in use in clinical settings, some even through National Health Service (NHS)–partner pilots, signaling institutional support. On the other hand, recent regulatory guidance from the UK Medical and Healthcare Products Regulatory Agency has classified ambient scribe technology as “software as a medical device.” NHS England and other stakeholder organizations such as the British Medical Association have followed up with guidance around information governance — ensuring standards around patient data processing are met  — and patient safety. Regulation is a necessary safeguard in deploying new products in clinical practice. On the other hand, it has added operational overhead in both cost and expertise for health care providers, such as the need cybersecurity support to ensure legal and regulatory requirements are met. 

The complex regulatory requirements present a barrier to use of AI scribes for individual clinicians and smaller health care organizations such as GP practices, which deliver most of the primary health care in the UK. While software companies that are already serving UK primary care are adding AI scribe products to their suites, the British Medical Association has been explicit that physicians working in family practice are “still ultimately responsible” for ensuring any product they use meets the regulatory standards. The key requirements include risk identification and assessment; ensuring the technology meets the appropriate regulatory compliance, as well as data compliance and security standards; a support and maintenance plan, monitoring and audit plan, as well as user training; and clear contracting arrangements to manage liability.  

That’s a lot for a small organization to take on without in-house expertise. None of the health care providers I have worked for have gone through these steps for an ambient scribing product. As such, I am yet to use an ambient scribe in my practice, though I certainly would benefit from it.  

An additional barrier is limited evidence of AI scribe effectiveness in UK settings. Although larger scale pilots are now under way in the UK, most published evidence to date comes from US systems, where documentation standards relate closely to billing. Still, the recently published 10-year plan for the UK NHS looks primarily to technology as a solution to many of the challenges around the growing demand for health care that cannot be met with current financial or human resources. 

US Experience: Evidence and Drivers  

The US health care market provides a smoother landing ground for AI scribes for several reasons. There are large integrated health care systems spanning primary and secondary care that can trial and scale innovations quickly. What’s more, US providers bear more direct financial risk and therefore have stronger incentives to adopt efficiency-enhancing technology. Vendors have easier access to markets, and AI is regulated through a mix of federal and state law, and therefore less centralized than the UK NHS. This helps foster a culture of innovation, enabling a more flexible landscape where decision-making occurs closer to the organizational level.  

A 2024 survey of 43 US health systems found that all respondents had begun deploying ambient scribing solutions, and more than half considered clinical documentation their most successful AI usecase. While the tools are more available at health system level they may not always be used by all clinicians. In a survey of over 1,000 physicians by the American Medical Association carried out last year, about 20% reported using AI tools for documentation, including creation of visit notes and progress notes. The report concluded that physician sentiment towards using AI in practice was less hostile than it was the previous year, with the most perceived opportunity around administrative tasks.  

Nevertheless, challenges include tool maturity and lack of evidence of impact on clinical outcomes and formal cost benefit analysis. Like the challenge faced by UK GP practices, there is a risk that smaller primary care practices may struggle to adopt technologies developed for and studied within larger systems.  

Future Challenges and Opportunities  

Regulation is likely the main brake on adoption of AI scribes in UK primary care, while in the US, commercial incentives for providers may also act as a catalyst. Ultimately, regulation on both sides of the Atlantic is important to protect patients’ health and their data. However, this must be balanced with the potential benefits of AI scribes in improving the quality and quantity of health care delivered. This said, two major questions still need addressing as AI scribes become more mainstream. First, if documentation time is reduced, will clinicians be expected to see more patients or offer longer consultations? Without careful workforce planning, time savings could be swallowed by increased workload, negating wellbeing benefits. Secondly, how can equity of access be established? Both the UK and US risk leaving behind small primary care practices and under-resourced clinics that cannot afford AI tools or to conduct due diligence assessments.  

In the UK, where regulation is centralized, national stakeholder organizations could provide shared procurement and evaluation support to achieve safe and equitable deployment. This would pool funding and expertise, helping smaller providers with the burden of ensuring the tools meet the standards required. In a decentralized landscape like the US, software often gains traction through informal peer learning, and stakeholder consortia, for example, could help drive progress. Collaboration between academic institutions, primary care providers, and AI scribe vendors is needed to develop the setting-specific evidence base.  

AI scribes hold promise for reducing documentation burden, improving clinician wellbeing, and enhancing patient engagement across the globe. The US experience shows that, within supportive organizational structures, these tools can be adopted at scale and deliver tangible benefits. In the UK, enthusiasm among clinicians is more tempered by regulatory uncertainty. Sharing guidance, learning from implementation in practice, and robust evaluations will be essential to realize the full potential of ambient documentation assistants in primary care in both countries. Where distribution routes differ, there is even more to learn, for example, around the impact of speed vs. regulation. Meanwhile, we can’t forget human health care workers if we want to ensure sustainable and equitable gains in meeting health care demands. As AI advances, human resource strategies need to evolve too, to recruit and retain clinicians, as well as help teams adopt AI with clear roles and thoughtful change management. That said, I look forward to the day where I look after the patients and the documentation looks after itself.