Ambience Healthcare, an AI startup, has introduced a new medical coding model powered by OpenAI that reportedly outperforms doctors by 27% in identifying ICD-10 codes—standard classifications used globally for various diseases and medical conditions. The model listens to patient-doctor interactions and automatically assigns the correct codes, aiming to reduce billing errors and lighten the administrative load on healthcare workers.
Built using OpenAI’s reinforcement fine-tuning technology, the model was tested against a gold-standard set of cases curated by expert clinicians. When benchmarked against the performance of 18 board-certified physicians, the AI demonstrated a 27% improvement in coding accuracy.
Ambience emphasizes that its tool is not designed to replace medical professionals but to support them by streamlining administrative tasks. The technology is already in use at over 40 healthcare institutions, including UCSF Health and Cleveland Clinic, and the startup has raised over $100 million from notable investors such as Kleiner Perkins, Andreessen Horowitz, and the OpenAI Startup Fund. The company is reportedly pursuing additional funding at a valuation exceeding $1 billion.
In addition to ICD-10, Ambience already supports CPT codes and plans to expand into areas like prior authorizations, utilization management, and clinical trial matching. The new model is expected to roll out to customers in the coming summer, marking what Ambience sees as a transformative step in healthcare documentation and efficiency.
