An AI-Powered Startup Can Now Perform This Important Task Better Than Doctors

A healthtech startup has used OpenAI’s technology to develop a new, customized AI model that they claim can outperform real physicians at generating accurate medical codes and notes. Ambience Healthcare, a San Francisco-based company that provides healthcare workers with an AI-powered platform, says that its new model is specifically designed to identify ICD-10 codes, which are used to classify diseases and medical conditions for billing and record-keeping purposes. By recording a doctor-patient conversation, the AI model can intuit the conditions being discussed and generate notes with accurate codes. From corporate giants like Amazon, which recently launched AI tools for medical practitioners, to startups like Suki AI, businesses large and small are creating AI-powered products aimed at helping doctors and clinicians spend less time on administrative work, like writing and filing medical notes. According to a 2024 study published by Google Cloud and The Harris Poll, clinicians spend an average of nearly 28 hours per week on administrative tasks, and 94 percent of them say their administrative workloads prevent them from spending more time helping patients. In a press release, Ambience chief medical officer William H. Morris said that its platform can be understood “as a scribe that fluently speaks both clinical language and the intricate healthcare billing rulebook from day one.” In addition to helping clinicians finalize their notes faster, he claims, it will also ensure that “revenue cycle teams receive cleaner, more accurate, and audit-ready charts.” Ambience’s platform also integrates natively with several popular electronic health record providers, such as Epic and Oracle. This means that after clinicians review and sign off on notes made in the platform, they’re immediately added to the patient’s record. In February 2024, Ambience raised $70 million in a series B round co-led by OpenAI’s startup fund, and the two firms started collaborating. In partnership with OpenAI’s startup solutions team, Ambience created a dataset of complex clinical cases, all labeled with accurate ICD-10 codes. By fine-tuning an OpenAI model with this high-quality proprietary data, Ambience was able to create a powerful medical coding model. In a test that pitted Ambience’s new model against 18 board-certified physicians, the company said that “Ambience’s AI demonstrated a 27 percent relative improvement in coding performance compared to the expert physician baseline.” These coding improvements could help medical practitioners ensure that they always get paid for the work they perform, and mitigate the risk of billing errors. ‍ BY BEN SHERRY @BENLUCASSHERRY