418 Background: Medical documentation is necessary to describe encounters for clinical and billing purposes. The increasing time required for medical documentation is contributing to provider stress and burnout. Physicians divide their time and attention between patients and the electronic health record (EHR), negatively impacting the patient physician relationship. Methods: 49 physicians from Texas Oncology participated in a pilot from November 2023 to April 2024 using Deep Scribe (DS) ambient artificial intelligence (AI) technology to document patient clinic visits. Subspecialties of medical oncology, radiation oncology, urology, and breast surgery completed opinion surveys pre and post pilot. Provider adoption, weekly usage, satisfaction scores were measured. Evaluation and management (E&M) diagnosis codes for office visits conducted pre-pilot were reviewed for over 27,000 patient encounters and compared to over 4,500 DS patient encounters. Hierarchical condition category (HCC) codes captured via DS AI were compared to manual capture of billing codes in the pre-pilot period. Results: 77% of providers adopted DS AI technology during the pilot. Median turnaround time for notes was less than 3 minutes. Average satisfaction score was 81%. 91% of providers stated DS was easy to use. 60% of providers believed DS improved the quality of documentation. Providers reported on average 1.5 hours per provider time saved per week. The average number of billed diagnosis codes for non- DS patients was 3 per patient. For the DS patients, the average billed diagnosis per patient was 4.1. Diagnosis codes were categorized as cancer, and other non-cancer HCCs using the CMS-HCC model category version 22. Of the top ten HCC categories, cancer diagnosis codes were better captured by manual coding, whereas DS captured more non-cancer HCC codes across multiple systems, most significantly in endocrine and cardiovascular systems. Conclusions: Data from this pilot study shows physicians reported DS AI improved quality of their documentation and saved physicians’ time. Billing data shows increase capture of non-cancer HCC codes. Use of AI in medical documentation can decrease physician burden and improve quality of patient care. E&M Diagnosis Codes Non-Deep Scribe Deep Scribe Patient visit count 27,236 4,584 Diagnosis code count 82,231 18,702 Average 3.0 4.1 Median 2 3 Mode 2.0 1.0