Shier Ziser Dawood is a GPST2 training in North West London, currently undertaking an integrated training post on the use of AI in general practice
The NHS 10 year plan was recently released¹ and as expected the benefits of ambient scribe and other AI support for clinicians featured prominently. One figure that stood out for me was that “AI scribes will end the need for clinical notetaking … (and) can save the same time as adding 2,000 more doctors into general practice.”²
The first somewhat problematic part of this was the suggestion of AI scribes ‘ending’ the need for clinician note-taking entirely. In the future this may well be the case, but current technology requires detailed reviews of all scribe outputs which is made clear in all implementation guidelines published by the NHS and scribes themselves.³ In practice I’ve found this review and editing process often takes almost as long as writing an old-school note from scratch. The scribes rely on large language models to operate, which are AI models trained on large amounts of text and often function in ways even their programmers don’t fully understand.⁴ This is a phenomenon often referred to as ‘black box’ operating. This can lead to unexpected outputs that can’t be easily mitigated by the scribe creators, for example a recent consultation of mine where the scribe was adamant I told the patient they are to ‘continue their Prozac’ that I had prescribed, a medication neither I or any of my UK-based colleagues would likely have ever authorised and that had not been mentioned in the consultation or featured in the scribe-formulated transcript. Errors such as this are often referred to as ‘hallucinations,’ essentially made-up results created because a LLM aims to please and would like to give you an output at any cost in order to fulfil its duty. Similarly to a junior clinician scribing, who may write something nonsensical rather than questioning their senior when they didn’t hear or understand part of the consult, the AI scribe will almost always give you a consultation summary even if it only picked up a few lines of a consult or even none at all.
There are murmurs of this theoretical time saving being utilised in primary care to facilitate 2 extra patients per day being seen…However, considering over 80% of doctors currently work unpaid overtime hours…
As technology improves errors like this will likely become less and less commonplace, but as things stand there is a concern that doctors will be facilitated with the scribes, expected to deliver a certain amount of time-saving and therefore given more work to fill these time gaps.
The largest and most recent studies into the topic have demonstrated a range of time-saving gained through use of scribes, roughly between 0.7 minutes per note/patient⁵ or just under 20 minutes in the EHR (electronic health record) per day.⁶
There are murmurs of this theoretical time saving being utilised in primary care to facilitate 2 extra patients per day being seen, something I’ve heard being discussed in relation to how ‘productivity benefits being realised’. However, considering over 80% of doctors currently work unpaid overtime hours,⁷ UK primary care consultations are already some of the shortest in the world⁸ and that these factors partially account for the high levels of stress and burnout in GPs in particular,⁹ an abundance of caution should probably be utilised when converting minutes saved to consultations gained.
Arguably one would be hard-pressed to find a GP who wouldn’t take 2000 extra GPs versus access to AI scribes, but one is clearly easier to source than the other, and particularly in public services such as the NHS it is true that there’s no such thing as a free meal. Indeed, the scribes cost around £40-100 per month per user,10,11 money that has to come from somewhere and with primary care budgets being tighter than ever12 ICBs will not be able to justify funding these without costs being cut elsewhere or objective benefits demonstrated, which are more likely to be extra patients being seen rather than an extra pep in a GPs step due to less onerous admin time.
Indeed, the scribes cost around £40-100 per month per user…
A legitimate fear is that because scribes are an already established and relatively ‘low risk’ AI tool, when compared to other tools that are used for clinical diagnosis for example, they will be first and fastest to be rolled out en mass in the NHS. Often this is done without proper on-boarding or clinical governance, an issue which was recently picked up by the NHS’s chief clinical information officer,¹³ but on the ground has not really led to tangible changes in how they are used. The scribes may be a double-edged sword in this regard, offering clinicians a plethora of benefits but with considerable risk if their implementation leads to piling on work onto an already overstretched workforce without properly establishing how much time will actually be saved and whether they are safe for use within this shorter timeframe and with minimal training.
So far, there have not been any major studies demonstrating safety and efficacy of scribe-enabled notes, they may be saving time and helping clinicians feel less hindered by admin, but as it stands there is nothing convincing that quicker most definitely equals better. The personal cost of this ‘free meal’ might end up leaving quite a bitter aftertaste for some clinicians and their wider teams in future, the ones that haven’t been replaced by an AI tool at least.
References
1. UK Government. 10-year health plan for England: fit for the future. Available from: https://www.gov.uk/government/publications/10-year-health-plan-for-england-fit-for-the-future [accessed 18/7/25]
2. North West London Integrated Care Board. Launch of “Fit for the Future”: 10-year health plan for England. Available from: https://www.nwlondonicb.nhs.uk/news/news/launch-fit-future-10-year-health-plan-england [accessed 18/7/25]
3. NHS England. Guidance on the use of AI-enabled ambient scribing products in health and care settings. Available from: https://www.england.nhs.uk/long-read/guidance-on-the-use-of-ai-enabled-ambient-scribing-products-in-health-and-care-settings/ [accessed 18/7/25]
4. Masood A. Is it true that no one actually knows how LLMs work? Towards an epistemology of artificial thought. Medium. Available from: https://medium.com/@adnanmasood/is-it-true-that-no-one-actually-knows-how-llms-work-towards-an-epistemology-of-artificial-thought-fc5a04177f83 [accessed 18/7/25]
5. Tierney, A. A. et al. Ambient artificial intelligence scribes: learnings after 1 year and over 2.5million uses. NEJM Catal. Innov. Care Deliv. 6, Available from: https://catalyst.nejm.org/doi/full/10.1056/CAT.25.0040 [accessed 18/7/25]
6. Ma SP, Liang AS, Shah SJ, Smith M, Jeong Y, Devon-Sand A, Crowell T, Delahaie C, Hsia C, Lin S, Shanafelt T, Pfeffer MA, Sharp C, Garcia P. Ambient artificial intelligence scribes: utilization and impact on documentation time. J Am Med Inform Assoc. 2025 Feb 1;32(2):381-385. doi: 10.1093/jamia/ocae304. PMID: 39688515; PMCID: PMC11756633.. Available from: https://pubmed.ncbi.nlm.nih.gov/39688515/ [accessed 18/7/25]
7. Gulliford M, Cowie L, Morgan M. Relational continuity of care and prescribing in general practice. BMJ. 2015;350:h1086.
8. Pulse Today. GPs across the world: if 15-minute appointments are a 20th century relic, is the UK prehistoric? Available from: https://www.pulsetoday.co.uk/analysis/general-practice-across-the-world/gps-across-the-world-if-15-minute-appointments-are-a-20th-century-relic-is-the-uk-prehistoric/ [accessed 18/7/25]
9. The Health Foundation. Stressed and overworked. Available from: https://www.health.org.uk/reports-and-analysis/reports/stressed-and-overworked [accessed 18/7/25]
10. Heidi Health. UK pricing. Available from: https://www.heidihealth.com/uk/pricing [accessed 18/7/25]
11. Digital Marketplace. Pricing document. Available from: https://assets.applytosupply.digitalmarketplace.service.gov.uk/g-cloud-14/documents/722631/287098019864081-pricing-document-2025-03-13-0635.pdf [accessed 18/7/25]
12. The King’s Fund. ICB cuts: what does it mean? Available from: https://www.kingsfund.org.uk/insight-and-analysis/blogs/icb-cuts-what-does-it-mean [accessed 18/7/25]
13. Mind the Product. NHS clamps down on unregistered AI scribing tools. Available from: https://www.mindtheproduct.com/nhs-clamps-down-on-unregistered-ai-scribing-tools/ [accessed 18/7/25]
Featured Photo by Igor Omilaev on Unsplash
