Find my path

Three possible futures

Plan against all three. Bet on none.

These are not predictions. They're three coherent stories about how things could go, each with its assumptions in plain view, so you can see which one the news is pointing toward and change course.

Five horizons

Choose how far ahead to look

Training and your early career, where income is set

Plausible scenario

Constrained adoption

Early-career jobs look much like today's, with the usual pressures on payment.

Accelerated augmentation

Early-career jobs are redesigned around higher capacity. Income may split by specialty, setting and ownership.

Broad autonomy

Some routine roles shrink noticeably, while new oversight and hybrid roles appear.

In all three: Debt taken on earlier is still due, whichever future arrives. Illness and fear still bring people to care.

Each one, with the case against it

Constrained adoption

Plausible scenario

AI keeps improving, but clinical evidence, safety incidents, regulation, trust, payment and institutional habit slow it down. Human roles shift gradually.

What must be true

  • Benchmark gains keep failing to show up as better outcomes for real patients.
  • A visible safety failure or legal ruling makes institutions more cautious.
  • Payers and regulators keep requiring a licensed person on most decisions.

The case against it

  • Physician use is already climbing fast, and it is mostly voluntary.
  • Patients are using AI directly, without waiting for institutions to approve it.
  • Cost pressure on health systems is a strong reason to adopt quickly.

Signals to watch

  • Prospective trials repeatedly show no benefit over usual care.
  • Major malpractice rulings against AI-assisted decisions.
  • Regulators tighten the pathways most devices use today.

If it unfolds

  • Traditional training keeps most of its value, so the path you know is still a sound plan.
  • Being the person who can evaluate these tools carefully becomes a real advantage.
  • Keep building AI literacy anyway. Slow adoption is still adoption.

Income: Today's income structure mostly holds. The pressures that already exist, such as payment rates and practice costs, matter more than AI does.

Next review by February 6, 2027.

Accelerated augmentation

Plausible scenario

Advanced AI becomes normal across workflows. Clinicians supervise, step in, handle uncertainty and coordinate far more care. Training and economics change significantly.

What must be true

  • Tools prove reliable enough for routine use, with clinicians still accountable.
  • Payers and employers reorganize work around higher capacity per clinician.
  • Training programs adapt, though unevenly.

The case against it

  • New tools can add work instead of removing it, and saved time may simply fill up.
  • Skill loss could undermine the supervision this future depends on. Most physicians already worry about it for trainees.
  • Supervising something usually right is a hard job people do poorly over time.

Signals to watch

  • Panel sizes or daily patient counts rise across large health systems.
  • Billing codes and payment rules appear for AI-supervised care.
  • Residency programs change what they teach and how they assess it.

If it unfolds

  • Judgment, communication, procedures and handling the unusual case become the core of the job.
  • Learn to catch a confident wrong answer. Practice it now on real cases.
  • Choose training that keeps your hands-on skills strong while tools do more of the routine work.

Income: This is where income is hardest to call. Higher capacity could raise earnings for clinicians paid by volume or lower them where employers and payers capture the gains, and it may split widely by specialty and ownership.

Next review by February 6, 2027.

Broad autonomy

Plausible scenario

Validated AI and robotics independently carry out substantial parts of care in authorized settings. Workforce structures, professional boundaries and prices could change dramatically. This is a possibility, not an endorsed arrival date.

What must be true

  • Systems are validated in real patients, across populations, with reliable failure handling.
  • Law and insurance settle who is accountable when an autonomous system causes harm.
  • The public accepts it, and physical robotics advances far beyond screen-based work.

The case against it

  • We found no evidence of a major specialty working this way in real care today.
  • Physical care, messy data and rare events are all far harder than published test cases.
  • Licensing, liability and politics can hold this back long after the technology is ready.

Signals to watch

  • Regulators authorize AI to make clinical decisions with no clinician reviewing them, in a mainstream use.
  • Payers reimburse fully autonomous care at scale.
  • Large trials show autonomous care matching or beating clinicians on hard outcomes.

If it unfolds

  • Skills that transfer between jobs, and low fixed costs, protect you more than any one title.
  • Roles shift toward governance, exceptions, building and overseeing systems, and human care where people still want it.
  • Keep the size of your debt in proportion to how much you would still want this path in this future.

Income: Substantial displacement in some roles is possible, along with lower prices and wider access for patients. Heavy debt tied to one income assumption is the most exposed position in this future.

Next review by February 6, 2027.

Method

How we build these, and why you won't see a percentage

No probabilities. A number like "40% chance by 2040" needs a method tested against real outcomes. No one has one for this, so any such number would be invented precision.

No single curve. Fast progress in some measures is real. But capability can plateau or jump, and adoption moves by its own rules. Exponential change is a scenario to investigate, not a law of medical progress.

Tasks, not titles. A task can be automated long before a job disappears, and a job can change completely while its title stays the same.

Revised in public. Horizons roll forward from today, and we say what changed. If a trigger fires, the scenarios change with it. The further out you read, the less weight it should carry.

Sources last checked November 6, 2026

Next review by February 6, 2027.

What changed, and how we decide

Career guidance, not medical, legal or financial advice. For a major loan or licensing decision, talk with a qualified professional, and bring your questions from here.

Do you disagree?

If you have counterevidence, or we got something wrong, we want to see it. Don't take our word for any of this.

Send us the evidence