This article is part of our series, The Clinician’s Perspective, where we explore the intersection of AI and healthcare through the eyes of our team – former clinicians who understand the realities of patient care firsthand.
AI can't replace our clinical judgment (yet) – but it can still augment us as clinicians
Many of us clinicians have grown weary of digitalisation’s empty, out-of-touch promises. Complex digital systems, unintegrated solutions, and increasing administrative burdens have often worsened our workload. But what if digitalisation could actually liberate us to do what we do best – being clinicians for our patients?
Medicine is about managing the "caveats"
In my dissertation, Visits in the Digital Era of Swedish Primary Care, I highlighted the limitations of using automated medical histories for AI-driven triage. Our recent study published in the BMJ found human doctors still outperform AI (GPT-4) when cases aren’t simple multiple choice questions. At the same time, a recent study in JAMA found that AI scribe use was associated with greater efficiency, lower mental burden and a greater sense of engagement with patients. Value in health care emerges during the encounter between the patient and the clinician – through relationship-building and through our ability to act on nuances and adapt to individual patient needs.
Sure, evidence suggests that simple cases, such as uncomplicated urinary tract infections, can be “automated” using simple questionnaires. But for most clinical situations, our professional judgment is essential to ensure the right diagnosis and treatment.
At the core of general practice – and medicine as a whole – is handling all the "caveats" that arise in nearly every clinical encounter. There are always unique factors that prevent us from following a rigid, predetermined protocol. AI can be a valuable tool before and after we apply our clinical judgment, but it cannot replace it. When circumstances change, we need to reassess and adapt care to the patient’s specific needs.
The PETRA research group: Keeping AI in check for primary care
As many disappointing digital solutions claim to “use AI”, it’s easy to develop a cynicism towards all AI, but here we should be aware that the new-generation of AI – large language models such as ChatGPT – have taken the world by surprise with their unexpected capabilities. To ensure that we, as general practitioners, have a seat at the table in AI’s integration into healthcare, I co-founded the PETRA research group (Primary care Emergent Technology Research and Advancement) alongside fellow GPs.
Soon, we will publish two systematic review articles and a study where we stress-test large language models on their ability to assess complex primary care cases. So far, we have not found an AI system capable of fully replacing clinical judgment. However, we are seeing significant progress in AI’s ability to support decision-making.
Judgment has many layers
It’s important to note that “clinical judgment” is not a single skill – it has multiple components, and AI is only beginning to navigate them. How long it will take for AI to match human judgment is uncertain. So far, AI has demonstrated strengths in what Daniel Kahneman describes as System 1 thinking – fast, instinctive, and imprecise. However, the latest AI models, such as GPT-4 and the newly released GPT-o1, are beginning to show capabilities in System 2 thinking – slow, deliberate, and analytical reasoning.
Where AI still struggles is in contextual understanding – knowing when additional information is needed to make a fully informed decision. Until AI can reliably interpret clinical context, it should not replace tasks that require judgment and holistic thinking.
And even as the technology evolves, there will always be a demand for human connection. Patients may have access to all the medical information they need online, but they still want to hear a verdict from a human they trust. No one wants to receive a cancer diagnosis from a chatbot.
Reclaiming time with patients
As clinicians, we all recognise the moment at the end of a consultation when a patient brings up “one more thing.” Thee thought of extra documentation simply becomes a an additional stressor on our already time pressed schedule, and this ultimately is felt by our patients. If AI could reduce our administrative burden, we could be fully present with patients – making sure they understand their treatment plans and strengthening the trust that is fundamental to good care.
Digitalisation should support, not replace
Alongside my work as a GP, I realised that helping to build digital solutions solving real clinical problems in primary care is a purpose worth committing to. One of the tools we’ve developed transcribes clinical conversations and automatically generates drafts of consultation notes, referrals, and medical certificates – like a digital personal assistant for doctors.
By automating these documentation-heavy tasks, we can free up time for what matters most: the clinical encounter. Here, the AI assists after our professional judgment has been applied, and once reasoning capabilities have matured, we can soon let AI help us before and during the patient encounter.
The future of healthcare: Protecting the human element
AI cannot yet replace clinical judgment. And more importantly – there will always be a genuine demand for human interaction in healthcare. As we deploy AI in health care, let’s make sure we do it in a way that enhances, rather than diminishes, our presence with patients.
At the end of a consultation, when I look a patient in the eyes, summarise our plan, and ask hear them repeat it back to me – all while seeing the medical note seamlessly generated – I finally feel a sense of optimism about the future of primary care.
This post is an adapted version of my original article in SFAM's journal Allmänmedicin.
About Dr. Artin Entezarjou
Dr. Artin Entezarjou, Medical Operations at Tandem Health, is a specialist physician with over a decade of experience across emergency medicine, primary care, and preventative medicine. He holds a PhD in applied artificial intelligence and telemedicine and has co-founded initiatives advancing evidence-based training and AI in healthcare. Now focused on building scalable AI tools with Tandem Health, Artin continues to practise clinically to ensure technology solves the right problems.
If you’d like to discuss AI’s role in patient care, feel free to connect with Artin on LinkedIn.
