Cleveland Clinic

Cleveland Clinic's revenue cycle staff spent nearly an hour per patient encounter reviewing more than 100 clinical documents — including progress notes, discharge summaries, and pathology reports — to select accurate billing codes. The process was time-consuming, inefficient, and costly, and the health system could not hire enough staff to keep pace with the volume of encounters.

Cleveland Clinic partnered with AKASA to deploy an AI documentation assistant and an AI coding assistant trained on years of internal multi-specialty data. The tools parse complex medical records, stitch together cohesive patient stories, and surface coding suggestions for human coders to review and approve or deny — augmenting rather than replacing the existing workforce.

The tools were rolled out across all domestic Cleveland Clinic locations over four months and have since processed tens of thousands of patient encounters. The AI completes coding work automatically approximately 80–90% of the time with greater consistency than manual review. Cleveland Clinic subsequently expanded the partnership in October to target prior authorization, denials management, and incomplete documentation.

• Training AI on years of system-specific, multi-specialty internal data is key to achieving high accuracy in a complex coding environment.
• A human-in-the-loop model (AI suggests, humans approve) builds trust and maintains quality control while still capturing significant efficiency gains.
• Success requires deliberate application alongside the technology — identifying the right problems is as important as the tool itself.