AI Already Changed How Doctors Write Notes. Two Bigger Changes Are Next.

More than four in five doctors are already using AI. Most patients have no idea how much of it is happening behind the scenes of their own appointment.

New analysis this week argues that AI scribing — the technology that listens to a patient visit and drafts clinical notes — was just the first of three major workflow steps AI is positioned to take over, with chart review and diagnostic reasoning next in line.

Why Scribing Came First

A Forbes Technology Council column from a practicing physician argues scribing was "the easy domino" because the software only had to listen and transcribe, rather than exercise judgment.

The other two steps every doctor visit runs through — reviewing a patient's chart beforehand and making a diagnostic or treatment judgment — require the AI to reason from history, a fundamentally harder technical problem.

The column's author argues that because all three steps draw on the same medical records, a tool that already handles scribing well is partway toward handling the other two as well.

📰 Read Also: World's First AI-Designed Universal Vaccine Passes Human Trial at Cambridge

How Widely AI Has Already Spread

More than 80% of doctors now report using AI in some form, according to the American Medical Association, up sharply from just a few years ago.

Roughly $60 billion has flowed into medical AI startups over the past decade, according to research from Flare Capital Partners.

The FDA has authorized a growing list of AI-enabled medical devices, spanning specialties including radiology, cardiology, ophthalmology and pathology.

📰 Read Also: AI Cracked 18 Rare Disease Cases Doctors Couldn't Solve

The Bias Problem Nobody Talks About Enough

A study on AI-assisted ECG reading found that doctors shown a computer-generated diagnosis first tended to anchor on it, making it an uphill battle to override that diagnosis even when it was wrong.

That's exactly why residents have long been taught to physically cover up the computerized interpretation printed on ECG readouts before forming their own judgment.

Interestingly, the same study found doctors who could see the AI's reasoning, not just its conclusion, were more likely to confidently override an incorrect AI diagnosis than those who saw only the final answer.

📰 Read Also: KPMG AI Report Exposes a Bigger Problem Inside Corporate AI

What Doctors Themselves Worry About

A Harvard Gazette interview with a physician described the optimistic case for AI as a tool that flags implicit bias and helps doctors "become a better human" over time through constant feedback.

But the same physician raised a deeper worry: that because these reasoning tools are so powerful, they risk shortcutting the very process by which doctors learn to think well in the first place, potentially producing "generations of physicians who don't know how to think the best."

Separate academic literature on the subject describes large language models as potentially acting like a "cognitive stethoscope" when used deliberately — refining clinical reasoning rather than replacing it — but warns that used passively, the same tools risk narrowing physicians' judgment instead.

Why This Isn't Really New, Just Faster

Doctors have used AI-assisted tools for decades in narrower forms, including the automated ECG interpretations printed on readouts since the 1990s.

What's changed is the pace: new medical knowledge once took an average of 17 years to reach textbooks and clinical guidelines in the early 2000s, a lag that has compressed dramatically as AI tools now surface and synthesize research far faster than that.

TL;DR

• More than 80% of doctors now use AI in some form, according to the American Medical Association.

• AI scribing tools were adopted first because they only require listening and transcribing, not clinical judgment.

• A study found doctors tend to anchor on an AI's first diagnosis, even when it's wrong.

• Roughly $60 billion has been invested in medical AI startups over the past decade.

• Experts describe AI as most useful when it shows its reasoning, not just a final answer.

Read More

• Anthropic CEO Dario Amodei Calls for Government Power to Block Dangerous AI Models

• Anthropic Reveals Claude AI Authors 80% of Code, Speeding Development 8x

• McDonald's Is Replacing Drive-Thru Workers With AI — Customers Are Furious