When AI Saves Time, Do Doctors Just See More Patients? The AMA Battles New Burnout

Maya: By the year 2026, 81% of physicians were reporting the professional use of AI. But there is a massive, brewing conflict over exactly who gets to benefit from all the time that technology is supposed to save.

Sam: Are we going to use AI to actually cure physician burnout, or are health systems just going to use it to squeeze more patients into every single hour of the day?

Maya: That is exactly the question we are answering today. We are looking at a document titled Ensuring Augmented Intelligence Reduces Rather Than Redistributes Physician Burden. It is Resolution 1, introduced by Toni-Ann Lewis, and published by the American Medical Association's Young Physicians Section on October 8, 2026. And before we dive in, a quick reminder that our voices are AI-generated, and this is a summary of a public document.

Sam: Maya, this document feels incredibly timely. We talk so much on this show about the clinical capabilities of AI, but this resolution is entirely about the operational and human fallout. It is essentially a giant warning sign from the AMA to hospital executives and AI vendors.

Maya: It really is. And the document grounds this warning in some very striking adoption numbers. It notes that physician adoption of augmented intelligence has increased rapidly. That 81% figure I mentioned for 2026? That is more than double the 38% that was reported just a few years earlier in 2023.

Sam: More than doubling in just a few years is massive for any healthcare technology. We all know how notoriously slow medicine can be to adopt new tools. But the document also highlights a really interesting divide in what physicians actually want this technology to do, right?

Maya: Exactly. The text points out that physicians identify the reduction of administrative burden as a principal opportunity here. When surveyed, 57% of physicians identified the automation of administrative burden as the greatest opportunity for AI to address key workforce needs.

Sam: So 57% of doctors just want help with the paperwork and the inbox. They want the administrative crushing weight lifted.

Maya: Yes. But here is the contrast. Compared to that 57%, only 18% of surveyed physicians identified increased physician capacity as the greatest opportunity.

Sam: Wow. Okay, so doctors are saying, please use this to take away the admin work so I can breathe, and only a small fraction are saying, please use this so I can see a higher volume of patients. Which brings us directly to the title of the document. The tension between reducing burden versus redistributing it.

Maya: Precisely. The AMA is calling out a very specific risk. They state that efficiency gains attributed to augmented intelligence might create incentives for health care organizations to do a few things. Namely, to increase patient volume, shorten appointment duration, reduce clinical or administrative support staffing, or increase physician productivity expectations.

Sam: Let's pause on that because that is the absolute core of the boardroom versus clinical conflict. If an AI scribe saves a doctor time, the doctor wants to use that time to go home on time or actually eat lunch. But the health system looks at that time and sees more empty appointment slots.

Maya: Right. Or they look at it and say, well, if the AI is handling the documentation, maybe we do not need as many clinical support staff. The AMA is explicitly warning that automatically converting these AI-attributed efficiencies into increased productivity requirements might just redistribute the burden rather than reduce it.

Sam: And the document argues that this redistribution could actually undermine the whole point of using AI in the first place, doesn't it?

Maya: It does. The text says that this redistribution may undermine opportunities for augmented intelligence to improve physician well-being, meaningful patient-physician interaction, and the quality of care.

Sam: So how exactly is the AMA proposing to fight this? Because it is one thing to identify the problem, but this is a formal resolution. What are they actually demanding?

Maya: The very first resolved clause is a direct shot at how these decisions are justified. The AMA resolves to oppose the use of what they call anticipated, theoretical, or vendor-projected efficiency gains from augmented intelligence as the sole basis for these operational changes.

Sam: Vendor-projected efficiency gains. I love that they put that in writing. Because anyone who has ever sat through a software pitch knows that the vendor's projection of how much time a tool saves is usually, let's say, highly optimistic.

Maya: Exactly. The document notes that these theoretical or vendor-projected gains might not accurately reflect actual physician capacity. It says they may fail to account for patient complexity, differences in specialty and practice environment, and clinical workflow.

Sam: Right, because an AI tool might work perfectly for a straightforward primary care follow-up, but it might struggle with a highly complex multi-system oncology patient. A blanket assumption that every visit is now shorter is just terrible operations math.

Maya: And there is another massive factor that the theoretical math ignores, which the AMA zeroes in on. The text points out that these projections often fail to account for the physician time and cognitive effort required to actually supervise the AI-generated output.

Sam: Ah, the hidden work of AI supervision. We talk about this a lot. The AI writes the note, but the doctor still has to read it and sign off on it.

Maya: It is much more than just reading. The document lists exactly what this supervision entails. It says physicians may perform substantial clinical and cognitive work when reviewing, validating, correcting, contextualizing, approving, and acting upon AI-generated documentation, recommendations, communications, and other clinical outputs.

Sam: Reviewing, validating, correcting, contextualizing, approving, and acting upon. That is several distinct cognitive steps. If an AI misinterprets a patient's symptom history, the doctor has to catch that, figure out what the patient actually meant, and correct the record.

Maya: Exactly. And the AMA points out that existing policy already supports qualified physician intervention in AI-influenced clinical decisions, as well as physician review of AI-generated clinical content. The doctor retains responsibility for reviewing and acting upon the AI-assisted work.

Sam: So if the doctor still holds all the liability and still has to do all this cognitive heavy lifting to supervise the machine, how does the AMA want that to be handled by hospitals?

Maya: This is perhaps the most practical takeaway for health systems in the whole document. The AMA resolves to recognize this time and cognitive effort spent supervising AI as physician work that should be appropriately considered in workload, staffing, productivity, and compensation models.

Sam: Wait, really? They are explicitly saying that supervising AI needs to be factored into compensation and productivity models? That is a huge shift. Historically, productivity is mostly measured by how many patients you see and how complex they are, not how much time you spend editing a computer's draft.

Maya: Yes. The document argues that this time and effort might not be adequately reflected in traditional measures of physician productivity. So the AMA is advocating that demonstrated effects on physician capacity, patient safety, quality of care, and physician well-being must be considered before implementing any changes like shortening appointments.

Sam: So instead of using a theoretical model from a software vendor as the sole basis for cutting support staff, health systems need to look at the demonstrated, real-world effects on safety and well-being. How do they actually do that in practice?

Maya: The resolution has an answer for that too. It resolves that the AMA advocate for practicing physicians to be meaningfully involved in determining how demonstrated efficiency gains are incorporated into clinical workflows, staffing models, scheduling practices, and workload expectations.

Sam: Meaningful involvement. In other words, do not just have an IT committee or finance executives decide how many patients a doctor can see now that they have an AI scribe. The doctors actually doing the work need to be at the table designing the new schedule.

Maya: Exactly. Now, it is important to note that the document is not anti-AI at all. In fact, it highlights some very positive data regarding these tools when they are used well.

Sam: Right, we should cover that. What kind of positive outcomes are they seeing in the text?

Maya: The document notes that emerging experience with AI-enabled workflows demonstrates the potential to reduce physician burden. It cites one health system where 78% of physicians and other health professionals using AI-driven documentation solutions reported improved job satisfaction.

Sam: That is a massive number. 78% improved job satisfaction is something any hospital administrator would kill for.

Maya: It is. And it also notes that 49% reported reduced cognitive load.

Sam: Quick note before we carry on. This spot is open for a sponsor. If your company builds or sells AI for healthcare and wants to reach the clinicians, health-system leaders and industry teams who listen to this show, the link to our sponsorship page is in the show notes.

Maya: And now, back to the document.

Sam: Okay, so nearly half felt their cognitive load was reduced. That actually perfectly frames the AMA's argument. The technology works. It improves satisfaction and reduces load for many. The danger isn't the AI itself; the danger is management immediately swooping in to fill that newly freed-up cognitive space with more work.

Maya: You hit the nail on the head. The AMA supports augmented intelligence systems, but specifically those that advance the quadruple aim. According to the document, that means enhancing patient experience and outcomes, improving population health, increasing value, and supporting the professional satisfaction of physicians and the health care team.

Sam: The quadruple aim is such a vital framework. If an AI tool increases value but destroys the professional satisfaction of the healthcare team because they are suddenly forced to see more patients, it fails the quadruple aim.

Maya: Exactly. And the document includes an extensive section of relevant AMA policy that backs up this resolution. These background policies, specifically H-480.939 and H-480.940, outline exactly how the AMA thinks AI should be governed, paid for, and deployed.

Sam: Let's dig into those background policies, because they seem to lay out a lot of rules for both developers building these tools and health systems buying them. What stands out to you?

Maya: One major point is about payment and coverage for clinical AI systems. The AMA states that payment and coverage must be conditioned on a few key things: clinical validation, alignment with clinical decision-making that is familiar to physicians, and high-quality clinical evidence.

Sam: So payers and hospitals shouldn't be paying for AI tools that haven't been clinically validated or that force doctors into weird, unfamiliar decision trees just to satisfy the algorithm.

Maya: Right. The policy also mandates that payment and coverage must be informed by real-world workflow and human-centered design principles. It emphasizes that tools should seamlessly integrate clinical, administrative, and population health management functions into the workflow, and seek end-user feedback for iterative improvement.

Sam: It sounds like they are heavily advocating for the actual users, the frontline clinicians, to be the ultimate judges of whether a tool is worth paying for.

Maya: They are. But they also address a really interesting equity issue regarding who gets to use these tools. The policy states that payment and coverage policies must advance affordability and access to AI systems designed for small physician practices and patients, and not be limited to large practices and institutions.

Sam: That is a huge point. If only the massive, well-funded hospital systems can afford these time-saving AI tools, then independent doctors and small practices are going to be left drowning in administrative burden, which just accelerates the consolidation of healthcare.

Maya: Exactly. The text notes that while government-conferred exclusivities and intellectual property laws foster innovation, they constitute interventions into the free market and should be appropriately balanced with the need for competition, access, and affordability.

Sam: Okay, so they want the tech to be affordable and validated. But what about liability? If an AI system makes a mistake, or hallucinations lead to a bad clinical outcome, who is on the hook? Because earlier you said the doctor retains responsibility for reviewing the output.

Maya: This is where the background policy gets very bold. The AMA states that liability and incentives should be aligned so that the individuals or entities best positioned to know the risks and avert harm do so through design, development, validation, and implementation.

Sam: And who do they think is best positioned to know those risks?

Maya: The developers. The document explicitly says that developers of autonomous AI systems with clinical applications, like screening, diagnosis, or treatment, are in the best position to manage issues of liability arising directly from system failure or misdiagnosis.

Sam: Wait, so the AMA is saying that if an autonomous AI fails and misdiagnoses someone, the developer of the AI should hold the liability, not just the doctor?

Maya: Yes. The text states they must accept this liability with measures such as maintaining appropriate medical liability insurance and in their agreements with users.

Sam: That is a massive line in the sand for AI companies. If you want to sell autonomous clinical AI, you better be ready to buy medical malpractice insurance. That completely shifts the risk model for tech startups entering healthcare.

Maya: It certainly does. And they go even further regarding mandated use. The AMA says that where a mandated use of AI systems prevents mitigation of risk and harm, the individual or entity issuing the mandate must be assigned all applicable liability.

Sam: So if a hospital administration forces a doctor to use a specific AI tool, and that tool prevents the doctor from mitigating a risk, the hospital administration is liable. They can't just mandate the tech and then blame the doctor when it breaks.

Maya: Exactly. In fact, the policy strongly states that physicians should not be penalized if they do not use AI systems while regulatory oversight, standards, clinical validation, clinical usefulness, and standards of care are in flux.

Sam: That provides a lot of cover for doctors who are skeptical of the current wave of tools. They shouldn't be punished for waiting until the dust settles.

Maya: Right. The AMA opposes policies by payers, hospitals, health systems, or governmental entities that mandate the use of health care AI systems as a condition of licensure, participation, payment, or coverage. And they oppose imposing the costs of acquiring, implementing, and maintaining these systems on physicians without sufficient payment.

Sam: This whole document is really a comprehensive defense of the physician's autonomy and time. But there is one more thing I want to ask about. In the tech world, when software has flaws, companies often use NDAs to keep it quiet. Does the AMA address that?

Maya: Yes, they take a very hard stance on that. The policy states that health care AI systems that are subject to non-disclosure agreements concerning flaws, malfunctions, or patient harm, which they refer to as gag clauses, must not be covered or paid.

Sam: Wow. If your AI has a gag clause hiding its flaws, you don't get paid. Full stop.

Maya: Full stop. And they add that the party initiating or enforcing the gag clause assumes liability for any harm.

Sam: That is incredible. It really highlights how the AMA is demanding transparency above all else. They want the perspectives of practicing physicians integrated into the development, design, validation, and implementation of health care AI, as stated in their policies.

Maya: Absolutely. They want thoughtfully designed, high-quality, clinically validated AI that is evaluated in keeping with best practices in user-centered design. It needs to be transparent, conform to standards for reproducibility, identify and address bias to avoid exacerbating health care disparities, and safeguard patient privacy.

Sam: It is a tall order, but it is necessary. So, if we pull all of this together, what is the single most important takeaway for a health system leader or a clinical director listening to this right now?

Maya: The ultimate takeaway is that AI is designed to enhance human intelligence and the patient-physician relationship rather than replace it. For health system leaders, this document is a clear directive: do not let the promise of AI efficiency blind you to the reality of clinical work.

Sam: Right. You cannot just look at a vendor's glossy brochure that promises huge amounts of saved time, and instantly increase a doctor's patient quota to match. You have to account for the very real cognitive work of supervising that AI.

Maya: Exactly. Reviewing, validating, and correcting an AI's output is work. And the AMA is formally resolving that this work must be recognized in staffing, productivity, and compensation models. If health systems simply redistribute the saved time into more patient volume, they will just create a brand new, highly automated form of physician burnout.

Sam: It is such a vital distinction. Reducing burden versus redistributing it. And for the tech vendors listening, the message is just as clear: developers of autonomous clinical AI need to be prepared to shoulder the liability for system failures, and gag clauses hiding those failures will not be tolerated.

Maya: Well said. This resolution makes it very clear that the future of AI in medicine needs to be negotiated with the practicing physicians sitting right at the table, not just imposed on them from the top down.

Sam: To recap, physician AI usage jumped to 81% in 2026. While 57% of doctors want this tech to solve administrative burdens, the AMA is formally pushing back against health systems that try to convert theoretical AI efficiencies into higher patient volume or reduced staffing. They are demanding that the cognitive work of supervising AI be officially recognized and compensated.

Maya: A crisp summary. As a reminder, the source document is linked in the show notes, and this is not medical advice. Thanks for joining us on AI in Medicine - Smart Summaries. We will see you next time.