What Happens When All Medical Expertise Becomes Essentially Free?

I was reflecting on two studies recently published – one showing AI achieving 4x the diagnostic accuracy of physicians while reducing costs by 20%, and another demonstrating AI systems outperforming doctors on challenging clinical cases by significant margins (references at the end of this piece). This got me asking a fundamental question: What happens when all medical expertise becomes essentially free?

The implications are staggering, and most healthcare leaders are dramatically underestimating what’s coming.

The False Promise of “Administrative Efficiency”

When I speak to healthcare companies (and frankly any other company), the conversation typically centers around AI as an opportunity for “administrative efficiency” or modest cost savings. This perspective misses the profound transformation ahead. The real question isn’t how AI can optimize existing workflows – it’s how we can redesign healthcare delivery when diagnostic expertise becomes abundant rather than scarce.

The research from Microsoft and Google demonstrates this isn’t a distant possibility. Microsoft’s Sequential Diagnosis Benchmark showed their MAI-DxO system achieving 80% diagnostic accuracy – four times higher than the 20% average of generalist physicians on challenging medical cases. When paired with OpenAI’s o3 model, the system also reduced diagnostic costs by 20% compared to physicians and 70% compared to off-the-shelf AI models.

Similarly, Google’s study of 302 New England Journal of Medicine cases found their specialized language model achieved 59.1% top 10 accuracy versus 33.6% for unassisted clinicians. More importantly, when used as an assistive tool, it improved clinician diagnostic accuracy to 51.7% while making differential diagnoses more comprehensive.

These aren’t incremental improvements – they point toward a fundamental shift in the availability and quality of medical reasoning.

The Real Disruption is Coming from Outside Healthcare

Here’s what should keep healthcare executives awake at night: these breakthrough studies aren’t coming from healthcare companies, academic medical centers, or pharmaceutical giants. They’re emerging from Microsoft, Google, and OpenAI (and others coming soon) – technology companies with fundamentally different approaches to problem-solving and vastly greater resources for AI development.

If you’re not watching the tech world, you may not see what’s coming to the healthcare world. While healthcare organizations debate regulatory compliance and incremental improvements, tech giants are building foundational AI capabilities that will reshape entire industries.

This matters because these companies don’t operate under healthcare’s traditional constraints. They’re not burdened by existing revenue models, regulatory capture, or professional guild thinking. When Microsoft’s researchers achieved superhuman diagnostic performance, they weren’t thinking about how to preserve physician billing models – they were solving for optimal patient outcomes at minimal cost.

The important aspect? These companies are treating healthcare as just another domain where AI can demonstrate superior performance, much like they’ve done with chess, Go, protein folding, and software development. To them, medical diagnosis is an information processing challenge, not a sacred professional domain.

The Incumbency Trap: Why Current Players Will Struggle

Here’s the uncomfortable truth: I do not believe the largest transformation will emerge from existing healthcare companies. Today’s hospital systems, physician groups, and insurers are structurally bound to scarcity-based economics. When your revenue model depends on billing for physician time, procedures, and facility usage, AI that dramatically reduces those needs represents an existential threat.

This creates what I call the “productivity paradox” in healthcare. For fee-for-service providers, reducing costs literally means reducing revenue. The Microsoft study revealed this dynamic clearly – off-the-shelf AI models showed a strong correlation between diagnostic accuracy and cost, with the most capable systems ordering more tests and procedures. But when properly orchestrated, AI achieved better outcomes at lower cost by being more strategic about information gathering.

I believe healthcare leaders are overweighting the regulatory and professional nuances that they think will act as barriers to tech companies entering their domain. Yes, there are compliance requirements, licensing considerations, and liability concerns. But these represent engineering problems, not insurmountable obstacles, to companies that have solved far more complex technical challenges.

I believe the revolution will come from new entrants who build business models around abundance rather than scarcity – companies that capture value from better outcomes rather than more procedures.

The Technology Infrastructure Advantage

Consider the fundamental advantages tech companies bring to healthcare AI:

Scale and Resources: Google, Microsoft, and OpenAI have AI research budgets that dwarf the R&D spending of most healthcare companies combined. They’re already building the foundational models that healthcare organizations will eventually license.

Data and Compute: These companies have the infrastructure to process vast amounts of medical literature, clinical data, and research findings at scales impossible for traditional healthcare organizations.

Talent and Culture: They attract top AI researchers and engineers, creating teams that can rapidly iterate and deploy at global scale. Their culture rewards breakthrough innovation over incremental improvement.

Platform Thinking: Rather than focusing on specific medical specialties or procedures, they’re building general-purpose AI systems that can be applied across all domains of medicine – exactly what we saw in both the Microsoft and Google studies.

The Capitated Care Advantage

All of this said – there is one segment of our healthcare system uniquely positioned to benefit from all of this immediately: capitated care models, particularly Centers for Medicare & Medicaid Services Advantage plans. When providers receive fixed payments to manage complete patient care, AI can become an amazing ally rather than a competitor – every accurate diagnosis, prevented readmission, and avoided unnecessary test directly improves margins while enhancing patient outcomes.

Medicare Advantage has grown from 19% of Medicare beneficiaries in 2009 to over 50% today, representing a $400+ billion market naturally aligned with AI-driven efficiency. These organizations have both the financial incentives and regulatory framework to be early adopters of transformative AI applications. (HINT, HINT)

The Path Forward: From AI Interns to AI Physicians

The immediate future – I am thinking the next 3-5 years – we will see what a good friend of mine (Aditya Paul Berlia) calls the “AI intern” phase. Every practicing physician will effectively gain multiple AI assistants, handling routine tasks, flagging potential diagnoses, and continuously monitoring patient data. As one study participant noted, the AI was “required to pull some additional diagnoses that may not have been the final diagnosis but would be important to think about.”

But the AI interns won’t stay interns forever. Today’s frontier AI systems already demonstrate “polymathic ability to reason across specialties,” combining “the generalist’s range with specialists’ depth.” As the Microsoft researchers noted, this challenges traditional healthcare structures where “medicine is too vast for any one person to master in full.”

When these systems accumulate experience across millions of patients and integrate real-time research advances, they’ll eventually surpass human expertise in most clinical areas. The question becomes: why are we maintaining the human bottleneck for routine medical decisions?

Beyond Regulatory Resistance

Of course, professional licensing boards, the American Medical Association, and existing healthcare power structures will resist this transition. Current regulatory frameworks and professional licensing requirements will create friction, but as researchers note, “regulatory capture can only slow innovation, not stop it.”

Healthcare regulation might similarly protect traditional practice models in some contexts while entirely new care delivery systems emerge elsewhere.

Tech companies have proven adept at navigating complex regulatory environments – from financial services to autonomous vehicles to privacy law. They typically approach regulation as an engineering constraint to be systematically addressed, not an existential barrier. Expect them to pursue multiple pathways: direct-to-consumer applications, partnerships with forward-thinking health systems, and international markets with different regulatory frameworks.

The Strategic Imperative

The potential for AI-assisted diagnosis to address “urgent challenges in healthcare delivery” and “improve quality of care globally, helping to mitigate the costly impact of clinical workforce shortages” represents both opportunity and competitive threat.

For healthcare leaders, the question isn’t whether this transformation will happen – it’s whether your organization will be disrupted by it or help drive it. In resource-limited settings especially, “cost-effective strategies may enable health systems to impact more lives per dollar spent, allowing scarce medical resources to be reserved for those with the most urgent clinical needs.”

The winners will be organizations that answer this core question: In a world where medical expertise is abundant and nearly free, how do you create and capture value by focusing on outcomes, access, and patient experience rather than provider productivity? Mind you – I am not saying this is easy but I am saying it is necessary to be able to move to through this sooner rather than later.

The Way I See It

We’re witnessing the rapid and inevitable transition from healthcare scarcity to healthcare abundance, driven not by healthcare companies but by technology giants with different constraints, incentives, and capabilities. The evidence shows AI systems can already “advance both diagnostic precision and cost-effectiveness in clinical care” while achieving “superhuman performance” on complex diagnostic challenges.

Please do not misunderstand or misinterpret what I am saying – This isn’t about replacing doctors – it’s about fundamentally reimagining how medical expertise is delivered and consumed. The organizations that recognize this shift and rebuild their models accordingly will capture enormous value. Those that optimize existing fee-for-service approaches while ignoring the technology transformation will find themselves in a very difficult position IMHO. So, I will say it again just to be clear – this is a business model shift that is required.

If you’re not paying attention to what Microsoft, Google, OpenAI, and their peers are building, you’re missing the most important development in healthcare since the discovery of antibiotics. The transformation is accelerating faster than most realize, and it’s being driven by players who don’t think like traditional healthcare companies.

The only question is whether you’ll be ready for it.

One suggestion: Play out the simulation with your teams: What Happens When All Medical Expertise Becomes Essentially Free? Don’t argue if it can or cannot happen – play it out and see what you come up with….

References:

McDuff, D., Schaekermann, M., Tu, T., et al. (2025). Towards Accurate Differential Diagnosis with Large Language Models. Nature, pages 1-7.

Nori, H., Daswani, M., Kelly, C., et al. (2025). Sequential Diagnosis with Language Models. arXiv preprint arXiv:2506.22405v2.