AKASA has moved deeper into the hospital revenue cycle, launching an autonomous platform for inpatient medical coding and clinical documentation integrity on October 2. The South San Francisco vendor is expanding beyond its AI-assisted prebill review roots.
Coding a patient record drives reimbursement and quality reporting, and AKASA says the work is still mostly manual and error-prone. Its case rests partly on a peer-reviewed 2025 study in npj Health Systems, which recorded coding error rates reaching 20 percent. A Government Accountability Office review from July 2026 drew a related conclusion, naming verifiable accuracy as the hard part for hospitals putting AI to work on notes and coding.
Rather than one general model, AKASA tailors its AI per health system – adjusting for patient mix, clinical criteria and documentation conventions. Health systems keep the controls: they set the audit sampling rate, pick which service lines and payer mixes are in scope, and cap what the system may code. The platform carries ICD-10-CM and ICD-10-PCS natively.
The system ingests whole charts – discharge summaries, operative notes, progress notes, consults, labs, imaging and medications. It then runs a case from start to finish and assigns a complete code set for every specialty. Each code is cited to the precise chart language that supports it, plus the reasoning and a confidence score. AKASA says that trail lets a reviewer validate output in seconds.
A coder needs 30 to 60 minutes per inpatient encounter, and staffing shortages can leave charts untouched for days after discharge. AKASA claims its system finishes in under 90 seconds. In blinded third-party evaluations, its AI matched or beat human experts on MS-DRG assignment, principal diagnosis, clinical quality capture and present-on-admission accuracy, AKASA says. Cleveland Clinic, already a prebill customer, has said it will explore the autonomous tools.
