How AI Will Reshuffle Healthcare: A Systems View With Sangeet Paul Choudary
Most AI talk in healthcare gets stuck on a narrow question: can the tech do tasks faster than humans? That misses the real shift. AI is the mechanism that lets fragmented systems finally work together - and that's where the value moves.
Sangeet Paul Choudary, co-author of Platform Revolution and Senior Fellow at UC Berkeley, argues the real question is: what constraints does AI remove, which new ones appear, and who controls them? That's the game.
From Tasks To Systems
AI acts at three levels at once: tasks, organizations, and ecosystems. The biggest changes begin at the ecosystem level, where competition moves and past differentiators turn into commodities. That forces organizations to redesign how they create and capture value, which then reshapes jobs and tasks on the ground.
"AI does not simply change jobs because the tasks within them change, but because the entire system around jobs changes."
What Actually Makes A Task Valuable
Every task carries three kinds of value:
- Intrinsic: the value of doing the task well.
- Economic: what the market pays - driven by scarcity.
- Contextual: value in specific situations.
Knowledge once gave physicians economic value due to scarcity. AI makes knowledge abundant. Like Google Maps did for cab drivers, that flattens performance and compresses premiums on pure knowledge work.
Where Physician Value Moves: Risk And Coordination
As knowledge becomes accessible to more people, the frequency of physician consults may fall. But the premium doesn't disappear - it shifts. Physicians who manage risk and coordinate complexity still hold the keys.
Think anesthesia: many tasks are automatable, but when something goes wrong, the person who carries the risk still holds economic value. Expect a shift from simple cases to complex ones that require judgment, escalation pathways, and cross-disciplinary coordination.
Jobs Exist To Manage Constraints - Not To Perform Tasks
Before word processors, typists weren't just typing - they were managing the cost of errors. Remove the constraint, remove the job. Same in medicine. Focus less on the tasks you do, and more on the constraint you own: risk, coordination, and outcome accountability.
"Jobs are not just bundles of tasks. Jobs primarily exist to manage constraints."
Skills To Build Next: Curiosity, Curation, Judgment
When answers get cheap, questions become expensive. Borrow a page from sommeliers: information is abundant; conviction at the point of decision is scarce. Your edge is the ability to synthesize in context and help patients decide with clarity.
- Curiosity: ask sharper questions to frame the problem correctly.
- Curation: filter noise and elevate the few options that matter.
- Judgment: apply context, risk, and patient values to make the call.
Why "Doctors Using AI Will Replace Doctors Who Don't" Falls Short
It's true - and incomplete. Tools alone can flatten performance and shrink premiums if you don't redefine your role. The constraint you manage matters more than the tool you use.
- Don't just "use AI." Shift what you own - risk, complexity, coordination, outcomes.
- Redesign workflows around new constraints. Tasks will follow.
For Health Systems: Treat AI Like Barcodes At Walmart, Not K-Mart
When barcodes arrived, K-Mart sped up checkout. Walmart rebuilt the business: centralized data, dynamic pricing, supply-chain leverage. Same logic here. If you only chase efficiency, you'll miss the control points where value concentrates.
Identify the control points: data aggregation, triage, routing, care coordination, prior auth, and outcome assurance. Then rebuild operations around them.
AI's Superpower: Coordination
Today, coordination is slow (humans wrangling unstructured info) or brittle (automation that requires perfect schema). AI bridges both. Train an internal knowledge model on your transcripts, notes, emails, PDFs, and structured data so every team works from the same institutional brain - queried directly and integrated into workflows.
Across organizations, coordination improves when you either codify a shared language or use models that consume messy inputs. Auto claims offer a useful analog: platforms like CCC standardize shared inputs across insurers and repair shops, while firms like Tractable use vision models to streamline handoffs with simple photos.
Interfaces Don't Win. Coordination Does.
Alexa had users and partners, but not end-to-end coordination. Ordering a pizza required too many hops, so people defaulted to simple commands. Digital front doors in healthcare face the same trap.
To avoid becoming a pretty brochure, your front door must own the handoffs: intake, triage, benefit checks, scheduling, prep, follow-up, and escalation - with permissions, documentation, and billing happening in the background.
Practical Steps For Healthcare Leaders
- Map constraints: where do risk, coordination, or ambiguity block flow today?
- Build the institutional brain: unify unstructured and structured knowledge; make it callable in workflows.
- Redesign care paths: AI handles routine; humans own exceptions and escalations.
- Instrument outcomes: measure handoff success, time-to-decision, rework, and risk events.
- Governance: set guardrails, audit trails, and clear escalation authority.
- Negotiate control points: own triage, scheduling, and routing - or someone else will.
What This Means For Clinicians
- Move up the acuity ladder: more complex cases, fewer routine consults.
- Get great at question design, option curation, and clear risk communication.
- Coordinate teams and systems, not just visits.
- Develop data and model literacy; know when to override, escalate, or change course.
Learn More And Level Up
Sangeet Paul Choudary's book Reshuffle won the 2025 Thinkers50 Strategy Award, which underscores how system-level design beats tool-level tweaks. You can explore Thinkers50 here: Thinkers50.
Want to upskill your team for AI-era roles by job function? See curated options here: Complete AI Training - Courses by Job.
The Bottom Line
AI shifts scarcity from knowledge to coordination and risk. Clinicians who lean into judgment and synthesis - and organizations that redesign around new control points - will set the pace. Everyone else will be stuck doing faster tasks inside the same old system.
Your membership also unlocks:
