AI in Healthcare

AI Scribes Save 16 minutes a Day, Says JAMA. Diagnostic Reporting Needs More Than That

The largest study of ambient AI scribes ever conducted just delivered a reality check. Here's what it found, and why radiology and pathology were never really the subject of study.

PRAID AI Team · Sep 2, 2026 · 8 min read

AI Scribes Save 16 minutes a Day, Says JAMA. Diagnostic Reporting Needs More Than That

In April 2026, JAMA published the largest real-world study of ambient AI scribes to date. More than two years of data, five major U.S. academic health systems, and 8,581 clinicians. The headline number, a 16-minute daily reduction in documentation time, was immediately treated as either a triumph or a disappointment, depending on who was doing the marketing. It's neither. It's a precise answer to a question that most of healthcare AI has been asking wrong.

16 minutes of daily documentation time saved, 21% of eligible clinicians adopted, 8,581 clinicians tracked across 5 US health systems

What the study actually measured

Researchers from Mass General Brigham and UCSF, working as part of the Ambient Clinical Documentation Collaborative, tracked ambulatory clinicians across Mass General Brigham, Emory Healthcare, UCSF, Yale New Haven Health, and UC Davis between 2023 and 2025. Of the 8,581 eligible clinicians, 1,809 adopted an ambient AI scribe. The remaining 79% chose not to.

Per 8-hour clinical day: 13.4 minutes reduction in total daily EHR time, 16.0 minutes reduction in documentation time, +0.49 additional patient visits per week, and no change in after-hours pajama time charting

The senior author, Mass General Brigham's Chief Health Information Officer Rebecca Mishuris, was candid that these are modest reductions, unlikely on their own to explain the larger burnout improvements ambient scribes have been credited with elsewhere. That candor is the most useful part of the study, it separates two things the industry has been bundling together: time saved and documentation transformed.

Why is the number modest and why that's the right result

Ambient scribes like the ones studied, Ambience, Nuance DAX Copilot, and Abridge, were built to solve a specific problem: turning a spoken conversation between a clinician and a patient into a reasonably good clinical note. That's a real problem, and they're getting better at it. But it's a fundamentally different problem from structured diagnostic reporting. A primary care visit is a conversation. A radiology or pathology case is not. There's no patient in the room to talk to, there's an image, a slide, a set of findings that need to be dictated against a fixed clinical vocabulary, mapped to a structured template, and signed off in a form that a RIS, a PACS, or a referring physician can actually use. Shaving 16 minutes off a conversational note is a workflow tweak. Getting a CT report from spoken findings to a signed, structured, standards-aligned document is a different category of work entirely.

What the JAMA study tested

Ambient listening during patient conversations
General ambulatory note generation
Primary-care and outpatient encounters
Free-text summarization, not structured coding

What diagnostic reporting requires

Terminology-aware dictation, no patient conversation
Section-by-section structured reports
Template alignment to ACR / CAP / RCPath standards
Direct, sign-off-ready output into RIS/PACS

The category is about to be forced to specialize

Two pressures are converging on radiology and pathology reporting right now. The first is evidentiary: JAMA-level data is now publicly establishing that general-purpose ambient scribes deliver real but modest gains, which raises the bar for any tool that wants to claim more than that. The second is structural, the reporting software radiologists have used for over a decade is being retired.

Early 2026 Microsoft begins sending end-of-life notices for PowerScribe 360, the dictation platform that has held roughly 75% of the U.S. radiology reporting market since its 2010 launch.
August 31, 2026 Annual renewals and maintenance for PowerScribe 360 end. Departments are pushed toward a cloud subscription successor, with existing pricing agreements no longer honored.
August 31, 2027 Full support ends. Radiology groups that haven't migrated are left running unsupported, mission-critical reporting infrastructure.

At the same time, more than twenty AI scribe vendors are now competing in a market that's starting to consolidate, most of them built for general ambulatory documentation, not for the regulated, structured reporting that radiology and pathology departments actually run on. The JAMA study is, in effect, the evidence base that will separate "AI scribe" as a broad category from "diagnostic reporting engine" as a specific, higher-bar discipline.

The gap this leaves open

A generation of ambient scribes has proven it can shave minutes off a conversation. Almost none of them are built, certified, or templated for a radiologist or pathologist signing off a structured, standards-aligned report under a medical device framework. That's a narrower problem, and a harder one.

Where PRAID Scribe fits

This is the specific gap PRAID Scribe was built for. Not another ambient listening layer for conversations that don't happen in a reading room, a reporting engine built from the ground up for radiology and pathology, as a CE-marked medical device under MDR EU, with ISO 13485 and EUDAMED registration.

Built for the case, not the conversation

Real-time speech turned into structured, section-by-section reports — findings, impressions, measurements — without a patient dialogue to summarize.

Standards-aligned by default

Templates aligned to ACR radiology reporting guidance and CAP / RCPath pathology dataset structures, recognized by clinicians and auditors alike.

Regulated, not just fast

CE Marked (MDR EU), ISO 13485:2016, EUDAMED-registered — a compliance posture general ambient scribes don't carry.

Multilingual reporting

English native, with Arabic conversion — built for reporting workflows across GCC, MENA, and EU markets.

Zero procurement, zero upfront cost

Individual sign-up, monthly on/off billing — a working reporting seat in a day, not a quarter-long procurement cycle.

A platform, not a point tool

Standalone today, with the full PRAID diagnostic AI operating system available behind it when a department is ready to scale.

The JAMA study's real message isn't that AI scribes don't work — it's that "documentation time saved" and "diagnostic reporting solved" are two different claims, and the evidence so far only supports the first. For radiology and pathology departments facing a forced platform migration this year, that distinction is worth making before signing another multi-year contract with a tool that was never built for the reading room in the first place.

Sources

Rotenstein LS, et al. "Changes in Clinician Time Expenditure and Visit Quantity With Adoption of Artificial Intelligence-Powered Scribes: A Multisite Study." JAMA, April 2026. massgeneralbrigham.org
HIT Consultant — "JAMA Study: AI Scribes Deliver Modest EHR Time Savings Across 5 Major Health Systems." hitconsultant.net
Healthcare IT News — "AI scribes can reduce EHR charting, more so when used frequently." healthcareitnews.com
The Imaging Wire — "Microsoft Sunsets PowerScribe 360 Reporting Software." theimagingwire.com
Radiological.ai — "What Is PowerScribe 360? A Plain Guide for Radiology Groups." radiological.ai

Start using Praid AI today

AI-assisted radiology and pathology tools
Secure, compliant, and easy to integrate
Manage all your work in one place

Bring PRAID AI to your practice

Seamlessly integrate AI-powered precision into your clinical workflow and unlock faster, more accurate diagnoses for every patient.