AI startup Ambience’s new medical coding model outperforms doctors by 27%

In March, Microsoft co-founder Bill Gates warned that AI will replace most human jobs, including doctors and teachers, within a decade. Now, we’re starting to see signs of that prediction inch closer to reality.

On Tuesday, AI startup Ambience Healthcare announced a new medical coding model that’s beating physicians at a task critical to the business of medicine: ICD-10 classification. According to the company, the AI outperformed doctors by 27% in coding accuracy.

Ambience builds AI tools that listen to doctor-patient conversations in real time and generate clinical notes. Its latest model can also identify ICD-10 codes—used to classify everything from flu diagnoses to complex chronic conditions—as doctors record patient visits. The model was trained using OpenAI’s reinforcement learning tools, customized for the medical field.

Ambience Healthcare’s AI Platform Surpasses Clinician Performance by 27% in Medical Coding

There are over 70,000 ICD-10 codes. Getting them right is tedious, but important—it affects how hospitals bill insurers, how diseases are tracked, and how patient outcomes are measured. Traditionally, doctors or trained coders enter these manually, often under pressure and at the end of long shifts. Mistakes can lead to incorrect billing, compliance issues, and hours of follow-up work.

Ambience says its model can reduce that friction. In a benchmark test, the system was evaluated against a “gold panel” of expert-agreed codes. It then went head-to-head with 18 board-certified doctors. The result: the AI performed 27% better than the physician average.

 “Trained using OpenAI’s Reinforcement Fine-Tuning (RFT) technology, Ambience’s AI platform achieved a 27% relative improvement over physician benchmarks—reducing coding errors by approximately one-quarter and offering a powerful tool to tackle the $266 billion in annual administrative waste, including $19 billion in diagnosis coding & documentation substantiation errors,” Ambience said in a blog post.

“We’re not replacing doctors or coders,” said Brendan Fortuner, head of engineering at Ambience. “What we’re doing is we’re liberating them from administration, and we’re fixing mistakes that help make health care better, safer, more cost-effective.”

Dr. Will Morris, the company’s chief medical officer, emphasised that ICD-10 data isn’t just red tape—it’s the standard used to compare outcomes across hospitals and clinicians. “If you think about it from a data perspective, it’s how you can compare and contrast clinician A to B, or health system A to B,” he said. “It’s the cornerstone for quality,” he said in an interview.

Founded in 2020, Ambience is already used in more than 40 health systems, including the Cleveland Clinic and UCSF Health. The company has raised over $100 million from Kleiner Perkins, Andreessen Horowitz, and the OpenAI Startup Fund. According to The Information, it’s now seeking fresh capital at a valuation north of $1 billion. Ambience declined to comment.

The ICD-10 breakthrough adds to a growing suite of features already offered by Ambience. The startup previously rolled out tools for CPT (Current Procedural Terminology) codes, and it’s now eyeing other bottlenecks in healthcare admin, like prior authorisations, utilisation reviews, and clinical trial matching.

The new model will roll out to Ambience’s customers this summer.

“Getting it right at the point of care is a fundamental change,” Morris said.

Founded in 2020 by Mike Ng and Nikhil Buduma, San Francisco–based Ambience is backed by Andreessen Horowitz, the OpenAI Startup Fund, Kleiner Perkins, and other top-tier investors.

Ambience Founders