Ambience Healthcare’s AI Surpasses Physicians in Medical Coding Accuracy by 27%
Ambience Healthcare has unveiled a groundbreaking advancement in ambient AI that significantly outperforms board-certified physicians in ICD-10 medical coding. Trained with OpenAI’s Reinforcement Fine-Tuning (RFT) technology, the new AI model achieves a 27% relative improvement in coding accuracy, presenting a significant opportunity to reduce the estimated $266 billion in annual U.S. administrative waste.
Tackling Administrative Waste with “Coding-Aware” AI
At the core of Ambience’s platform is a coding-aware AI scribe that processes patient encounters to identify precise ICD-10 codes. These codes guide the generation of a fully informed medical note, which clinicians can review and approve in real time.
“This is like a scribe who speaks fluent clinical and billing language,” said Dr. William H. Morris, Chief Medical Officer at Ambience Healthcare. “It reduces errors, accelerates documentation, and delivers cleaner, audit-ready charts to revenue cycle teams.”
The system is already live across inpatient, outpatient, and emergency care settings in more than 40 U.S. health systems, including UCSF Health, Houston Methodist, and Memorial Hermann. It supports over 100 medical specialties and integrates directly with leading EHRs such as Epic, Oracle Cerner, and athenahealth.
Also Read: Foxconn's Smart Healthcare Vision at GTC Taipei 2025
Real-Time Compliance, Audit Readiness Built In
Ambience’s platform allows clinicians to simply tap 'Record', conduct a visit, and receive a suggested note with ICD-10 and CPT codes. The reviewed note is then auto uploaded into the EHR, resolving common compliance challenges and reducing time-consuming post-visit coding queries.
“We built compliance into the platform from day one,” said Michael Ng, CEO of Ambience. “Hospitals and clinics can’t afford for AI tools to compromise documentation integrity or reimbursement.”
Validated by Experts, Driven by OpenAI Technology
To validate the model’s capabilities, Ambience benchmarked it against a panel of 18 seasoned physicians using a gold-standard dataset. The AI consistently outperformed human coders, showing that RFT-aligned AI can meet the rigorous demands of high-stakes healthcare environments.
“Ambience shows how reinforcement tuning can enable AI to function in domains where precision is everything safely,” said Shyamal Anadkat of OpenAI.
As the first AI to surpass clinical benchmarks in coding accuracy, Ambience’s platform sets a new standard in healthcare documentation—offering relief from burnout, improved compliance, and smarter workflows.
