> ## Content Index > Fetch the complete content index at: https://www.readtangle.com/llms.txt > Use this file to discover other available public pages before exploring further. # The AI assistant your family doctor needs? - URL: https://www.readtangle.com/otherposts/the-ai-assistant-your-family-doctor-needs/ - Published: 2026-09-18T18:00:19.000Z - Updated: 2026-09-18T18:00:19.000Z - Author: Tangle Staff - Tags: reader-essay, #hide *By Lucas Seuren* --- In just a few years, artificial intelligence (AI) has come to pervade every aspect of our daily lives — whether you’re just searching for something on Google or ordering a self-driving taxi. People are now even using Large Language Models (LLMs) to help with daily interactions like making a recipe, asking someone out, or even ordering coffee from a barista. The impact of the AI explosion is wide-reaching: Entry-level and apprentice jobs are being displaced, the stock market (and with it, the global economy) has become seriously reliant on AI companies and hyperscaling with limited monetization, and we are losing the ability to tell whether anything we see and read is even real. The widespread anxiety around AI comes as no surprise, especially with changes happening at such a high pace. But not everything is doom and gloom. AI has many potential applications to provide a huge boon to society. Healthcare in particular stands to benefit immensely. AI can already read scans, spot tumours and suggest treatments, thereby improving human judgment. One promising area is frontline care. By assisting our healthcare professionals, AI can help prevent countless medical errors, alleviate workload pressures, and allow clinicians to do what they do best: caring for patients. This potential of AI for healthcare has become the focal point of my career. I have been researching technological innovation in healthcare since 2018 — first at Oxford, studying video consultations, and now at the University of Edinburgh, where I study AI tools such as automated documentation and decision-support systems. And I can already see the benefits. When I was 15, I started to have eczema. My case was not particularly bad, occurring mainly on my elbows and only when the weather was hot and humid. I could easily control it with a mild hydrocortisone cream. Then in my early 20s, it started getting worse. I would use my cream twice daily just to keep the eczema from spreading. Yet my family doctor did not want to escalate my treatment as long as a lower dose worked. It was not until a few months later, when my family doctor had talked to a dermatologist, that she was willing to prescribe a stronger treatment that would actually clear the eczema. This status quo worked for about 15 years; but then my eczema got so bad that my cream barely slowed down the progression. During this time, I had moved to a new country, and my new family doctor was even more hesitant to escalate treatment. Again, I had to wait months as my eczema got considerably worse before she finally decided to prescribe a stronger treatment. But my family doctor missed that my flare-up had an environmental trigger: The apartment I had moved into had mold. She had not questioned what could be causing the sudden severity of my eczema, and we ended up treating the symptoms rather than the cause. If you’re keeping track, that’s two family doctors, two timepoints, and two cases where I had to tolerate an irritating illness that was easily treatable. If only my first family doctor had been better aware of the latest guidelines, and if only my second had asked if anything had changed in my life recently — I might have gotten relief sooner. But the reality is that family doctors face extreme time pressures. They are our first points of contact with the health system; they see everyone and need to know something about everything. That’s asking a lot. Combine that with an ever-growing administrative burden, and it is no surprise that [an estimated 6.3% of all family doctor visits](https://qualitysafety.bmj.com/content/23/9/727?ref=readtangle.com) in the U.S. have some diagnostic error. In comes artificial intelligence. If you’ve visited a clinician in the last 2–3 years, you may have already been asked whether you are okay with your clinician using AI to make their notes. Nearly half of physicians now use Ambient AI Scribes or Ambient Voice Technology (I’ll use AI Scribes for short). The tech is simple: A microphone records your consultation, speech-to-text software produces a transcript, and then an LLM (such as ChatGPT) processes that transcript to generate a summary. Many clinicians, especially those who write long notes, find that AI Scribes save them considerable time both in- and outside the clinic. And taking the keyboard out of the consultation means that clinicians can have more meaningful conversations with their patients — something everyone prefers. Because of the perceived benefits, politicians are quick to embrace the technology: The UK government [envisions](https://www.england.nhs.uk/long-term-plan/?ref=readtangle.com) that by 2035, every healthcare practitioner will be using an AI Scribe. Although there is no long-term evidence to bear this out, the potential of a technological quick fix to healthcare is too good to pass up. Of course, the accuracy of my medical record was not the reason why my family doctors initially did not adequately address my problem. While AI Scribes can help prevent errors from slipping into the patient’s record (although, of course, LLMs can introduce their own errors), diagnostic errors sometimes happen before a note is even generated. Take my case as an example: My first family doctor was not aware of a change in prescribing guidelines, and my second family doctor did not ask if a recent change in my life could explain my symptoms. To prevent medical errors before they happen, we need an omniscient family doctor who never forgets anything. To prevent them from being classified as a medical device, all AI Scribes do — and, for the moment, are allowed to do — is provide a summary of the consultation. Regulation gets stricter as their functionality increases, and because they currently only summarize, regulators in the U.S., UK and the EU let developers self-certify that they are safe. But developers are already building newer AI Scribes that are full-fledged physician assistants. These could be extraordinarily helpful new tools. With access to the internet and the latest local guidelines, AI Assistants could suggest questions for the family doctor to ask, suggest a diagnosis, provide treatment options, and generate handouts for patients — all in real-time. And because they would be listening to the consultation anyway, clinicians could just ask them questions verbally, instead of having to type, which would take their focus away from the patient. I spoke to a charge nurse who works at a palliative care ward, and she reported that AI Scribes made it possible to do what she was trained to do: care for patients. This vision of healthcare, where technology makes it possible for the clinician to focus on caring for the patient, is not new. Tech enthusiasts have been preaching it for decades. Except we don’t know if any of this is a good idea. There is a reason that healthcare is slow to adopt innovation: We need high-quality, unambiguous evidence of safety and effectiveness before adopting anything new. For decades, randomized controlled trials have been the gold standard: Divide patients into two groups, give one the drug and one a placebo, and see which group does better. These trials work well for drugs, because we know how to measure their effect. Technology, however, is what we call a *complex intervention*. Its effects depend not only on whether it is used but how it is used, and how it gets embedded in existing routines and infrastructures. Its impact is often unpredictable and it always comes with unintended consequences. For example, Electronic Patient Records were expected to improve patient safety, but they also led to burnout and impersonal care. Over the past year, I’ve talked to many clinicians and patients, and I’ve worked with some of the UK’s leading research teams on AI Scribes. The findings are consistently more nuanced than politicians and technological leaders would have us believe. Yes, in some situations AI Scribes can lower the administrative burden, mitigate burnout, and facilitate patient-centered care. But in many cases, time savings are minimal. Two recent studies, one a [clinical trial](https://ai.nejm.org/doi/10.1056/AIoa2501000?ref=readtangle.com) with 238 physicians and one a large [real-world trial](https://ai.nejm.org/doi/10.1056/AIoa2500945?ref=readtangle.com) of thousands of physicians’ notes, found that at best AI Scribes reduce time spent on note-taking work by 9.5%, and in many cases there was no difference. This does not look like technology that will free up clinician time to focus on patient care. Again: AI Scribes are a complex intervention. They change how clinicians work, and this comes with unintended consequences. As clinicians remain responsible, they still have to review and edit the AI note, which can take more time than writing them. Some clinicians fear that by offloading their note-taking to an AI, they lose or never even acquire clinical reasoning skills. And worse, some report that when an AI takes notes, clinicians do not remember what is in the notes — or who the patient even was. More and more, we hear that clinicians who participate in pilot studies do not want to continue using AI Scribes after the study ends. For many clinicians, AI Scribes do not solve a problem; they just create new ones. None of this is to say that AI will not make healthcare better and safer. There are many cases of LLM chatbots successfully diagnosing patients where clinicians failed — although, of course, there are plenty of reports of patients being misdiagnosed. Technology, when implemented appropriately, allows clinicians to focus on what they are best at: providing care. But instead of viewing AI as a replacement for clinical work and skills, we need a vision of AI as augmenting those skills. AI Scribes are a classic case of “move fast and break things.” Tech companies recognized a business opportunity, and the medical workforce was desperate, especially post-Covid. As one family doctor told me, it is a gold rush with AI companies treating hospitals and clinicians as mineral-rich veins. Once clinicians depend on AI Scribes, companies can start raising prices. This all may sound deeply frustrating, and it is at times — but let me end on a positive note. Technology is always a learning process. Video consultations promised salvation during the Covid pandemic and social distancing. Many clinicians quickly found video did not work for them, and we learned where it was beneficial. AI Scribes will no doubt go through the same hype curve. I am still an optimist: I anticipate that in 3–5 years, AI will be part of any consultation. It will advise where appropriate, with clinicians remaining the expert on how to use that information. It won’t be perfect, it will make mistakes, and it will have unintended consequences. But for the family doctor who missed my mouldy apartment and the family doctor who was unaware of the latest guidelines, AI will help to better attend to the full picture of their patients’ problems. --- *Lucas Seuren is a Research Fellow at the School of Public Health Sciences at the University of Edinburgh, where he studies technology in healthcare. He currently works on* [*AI Scribes*](https://www.bbc.co.uk/news/articles/cy0zylpdveqo?ref=readtangle.com) *and precision medicine* [*for glioblastoma*](https://gliomatch.eu/?ref=readtangle.com)*.*