Stop Augmenting Doctors: Why We Need to Replace the Entire Clinical Unit with AI

I inhabit two worlds—practicing physician devoted to patient care and founder of a health tech startup. Living in this contradiction has led me to a conclusion that's heretical in medical circles: we need to stop the charade of doctor "augmentation" and start building full-stack AI clinics that replace the entire clinical unit.

This isn't some academic thought experiment. I've invested my own time and efforts building technology to help doctors work better, only to watch those efforts crash against the reality of healthcare integration. The truth is controversial but clear: the future of healthcare isn't augmented doctors—it's autonomous systems with human guides.

A Reality Check at Superhuman Speed

Last week, four Purdue undergrads shattered the Guinness World Record with a robot that solves a Rubik's Cube in 0.103 seconds. Not seasoned engineers or AI specialists—undergrads. Their machine processes information, makes decisions, and executes physical movements faster than a human eye can blink.

Meanwhile, healthcare celebrates saving doctors 42 seconds per patient note.

The contrast is absurd. We're celebrating incremental efficiencies while the technological capability for transformative change already exists. It's like watching the Wright brothers take flight while insisting we focus on breeding faster horses.

Augmentation Is a Failed Experiment

My own company set out to build the ultimate clinician copilot. We designed it, tested it, refined it based on clinician feedback. I am biased, but I believe it would have been revolutionary. That is, except for one stubborn reality: the clinicians themselves.

The evidence keeps piling up. A multi-site JAMA study showed documentation time dropping from 10.3 to 8.2 minutes per note. Another saved physicians a mere 0.9 minutes. Kaiser's massive 2.5 million-encounter rollout with ambient AI scribes? A paltry 42-second reduction per note.

Each integration touchpoint—EHR connections, workflow retraining, authentication protocols, sign-off policies—breeds fresh friction. Convincing decision makers at health systems to adopt a new software takes charisma and skills of persuasion practically at the level of scam artists. Not to mention, many brilliant doctors still regard AI as some form of technological witchcraft. They excel at medicine, not parsing the nuances of generative models, retrieval-augmented generation techniques, or diverse forms of agentic architecture.

The Two Souls of Healthcare

Working as both clinician and technologist has revealed a fundamental truth: healthcare performs two distinct functions that we've mistakenly bundled together:

Transactional care runs on algorithms and protocols. It's the diagnostic decision trees, prescription rules, test ordering, and documentation that consumes 80% of clinical time. It's fragmented across doctors, nurses, and techs, forcing everyone to work below their capabilities (or below the “top of their license).

Relational care operates on human connection. It's interpreting what health means in the context of someone's life, guiding them through uncertainty, motivating behavior change, and navigating value-laden decisions. It's what most of us became doctors to provide—yet it's increasingly suffocated by transactional demands.

Our current approach to healthcare AI—layering technology onto overtaxed clinicians—only reinforces this broken model. We're using technology with incredible potential only to shave seconds off tasks that shouldn't exist at all.

The Shift from Augmentation to Replacement Is Already Happening

Look at where regulations are clearer and risks are lower:

• Digital triage & prescribing: K Health's AI symptom checker guides millions of patients without human intervention. Patients already trust algorithms for initial advice.
• Pharmacy operations: Automated dispensing systems have cut error rates and staff workload in hospital pharmacies. Logistics-heavy work is being fully automated, not merely assisted.
• Surgical procedures: Virtual Incision's MIRA robot completed its first autonomous hysterectomy this year. High-stakes interventions are moving toward hands-off operation.
• Regulatory frameworks: 46 surgical robots cleared by the FDA since 2015 are advancing toward Level 5 autonomy. The regulatory path for replacement—not just assistance—is being mapped out in real time.

This isn't speculative futurism—it's happening now, following a predictable trajectory we've seen across industries: automation first tackles bounded, repetitive tasks, then steadily climbs the complexity ladder.

The Full-Stack AI Clinic: Blueprint for Medicine's Future

This transformation is unfolding in distinct phases:

Phase 0: Digital & Low-Risk (happening now)Symptom triage, prescription management, and chronic-disease monitoring—requiring minimal human oversight.

Phase 1: Automated Logistics (beginning)Robotics for pharmacy operations, self-service sample collection, and autonomous imaging systems.

Phase 2: Procedure Autonomy with Human Oversight (2-5 years)AI-led diagnosis and planning, endoscopy, laparoscopic procedures, and interventional radiology with humans monitoring but rarely intervening.

Phase 3: Closed-Loop Autonomous Care (5-10 years)AI systems that develop treatment plans, monitor patients proactively, and orchestrate interventions without human decision points.

The safety paradigm shifts from human oversight to system design—redundant verification pathways, formal logic validation, and real-time auditing provide more consistent protection than overtaxed clinicians who might be distracted, fatigued, or biased.

Reclaiming the Human Element in Medicine

Consider Dunbar's number—the cognitive limit on meaningful relationships humans can maintain, roughly 150. For physicians, this translates to about 30-40 patient relationships if we want any semblance of work-life balance. Yet we manage panels of hundreds or thousands, forcing shallow interactions that frustrate everyone involved.

I've experienced this tension firsthand. I've deliberately chosen depth over volume in my practice, focusing intensely on fewer individuals rather than maximizing throughput. This allows me to notice when they've changed jobs, lost parents, or made significant life transitions—all factors that profoundly shape health outcomes but vanish in high-volume care.

The autonomous AI clinic doesn't threaten this approach—it enables it at scale. By automating the routine, protocol-driven aspects of medicine, we create space for human clinicians to practice relationship-centered care that machines will never replicate. In other words, AI is liberating us from the chains of today’s broken healthcare system.

Breaking the Economics of Transactional Medicine

The current economic model rewards transactions over relationships—more patients, more procedures, more billing codes. But that framework is crumbling under its own weight.

What if, instead of clinicians spending 80% of their time on transactions and 20% on relationships, we inverted that ratio? What if physicians primarily served as guides and meaning-makers, while autonomous systems handled routine workflows?

Amazon now employs 750,000 robots and projects $10 billion yearly savings by 2030 while flattening its human workforce growth. The same economic forces will drive healthcare organizations toward autonomous systems that eliminate—not marginally improve—human cost centers.

A Call to Action

To clinicians: Redirect your expertise toward patient guidance, system design, and complex decision support. Your judgment remains invaluable—but it won't be tethered to order entry and documentation much longer.

To regulators: Shift from evaluating individual algorithms to certifying autonomous systems with continuous performance monitoring.

To payers and employers: Pilot autonomous care models in low-acuity domains now, before consumer-tech companies beat you to market with superior experiences at lower costs.

That Rubik's Cube robot reminds us that technological execution can leap past generations of incremental thinking. Healthcare shouldn't celebrate saving seconds when we can eliminate entire bottlenecks.

The future isn't augmented doctors struggling against integration friction. It's autonomous systems handling the transactional core of medicine, paired with human clinicians practicing the deeply relational care they trained to provide. That's the healthcare I'm working to build.

‍