October 2, 2026 — South San Francisco
AKASA has launched an autonomous AI platform for inpatient medical coding and clinical documentation integrity (CDI), expanding from its AI-powered prebill review into what CEO Malinka Walaliyadde calls the industry’s “holy grail”: an autonomous mid-cycle for the healthcare revenue cycle. The system reads a full patient chart at discharge and assigns the complete code set itself, linking every code back to the chart language that supports it.
Bottom line
Hospital billing is one of those industries where AI actually fits the job: repetitive, rule-bound, and drowning in backlogs. AKASA claims coding that takes human coders 30–60 minutes per inpatient encounter now finishes in under 90 seconds after discharge. If the blinded evaluations hold up in the real world, this is one of the most concrete labor-automation stories AI has produced yet.
What happened
AKASA, a South San Francisco company founded in 2018 (formerly Alpha Health), announced the launch of an autonomous AI platform for the mid-cycle of the healthcare revenue cycle — the stage where a patient’s clinical record is translated into the codes that drive reimbursement, quality reporting, and risk adjustment. The platform covers autonomous inpatient coding plus clinical documentation integrity, a unified AI layer the company says keeps documentation complete before codes are assigned.
The company says its customers represent more than $180 billion in aggregate net patient revenue and roughly 10% of U.S. inpatient discharges, and that inpatient volume processed through its AI products grew nearly 6x in the past year. Health systems set the operating boundaries: they define thresholds, service-line rules, payer mix, and audit sampling, and the system only codes what it has been approved to code. It works alongside existing EHR and billing systems, so hospitals don’t need to replace anything.
The accuracy claims
AKASA says the system was tested in third-party blinded evaluations where its AI and expert human coders coded the same inpatient encounters, covering 65% to 85% of inpatient volume at most health systems. Independent reviewers — not told whether each code set came from the AI or a human — found the AI matched or exceeded expert coders on key accuracy measures, according to the company: MS-DRG assignment, principal diagnosis, clinical quality capture, and present-on-admission accuracy.
Worth noting: these are company-reported results from a paid evaluation, not a peer-reviewed study. The announcement itself cites a 2025 peer-reviewed study in npj Health Systems reporting coding error rates up to 20%, and a July 2026 GAO review flagging verifiable accuracy as a central challenge for healthcare AI. The honesty about accuracy being hard counts in AKASA’s favor — but independent replication is still the real test.
Early customer response
Cleveland Clinic, which uses AKASA’s prebill review products, intends to explore the autonomous mid-cycle offerings. Chief digital officer Rohit Chandra said the clinic seeks to improve speed and precision in its revenue-cycle work under “a compliance-first approach.” Nebraska Methodist Health System CFO Jeff Francis called autonomy “the natural next step,” pointing to revenue integrity, denials, and write-offs. Andreessen Horowitz general partner Julie Yoo framed the bet simply: healthcare can’t meet coming demand through human labor alone.
Why it matters
This is the “everything AI, tested” angle in its purest form. Medical coding is not glamorous — but it is exactly the kind of high-friction, high-cost bottleneck AI is good at. AKASA’s framing is smart: coding happens at discharge, when the chart closes, and the company claims it cuts discharged-not-final-coded volume, shortens accounts-receivable days, and speeds cash collection without adding headcount. For hospital CFOs staring at staff shortages and three-to-four-day coding backlogs, that’s a pitch that sells itself. Watch for whether payers and auditors accept AI-generated code sets at scale — that’s where this either becomes infrastructure or hits a compliance wall.
Frequently asked questions
What is AKASA’s new platform?
An autonomous AI system that reads complete inpatient charts at discharge and assigns medical codes (ICD-10-CM/PCS and MS-DRG) itself, with an audit trail linking each code to supporting chart language.
How fast is it compared to human coders?
AKASA claims under 90 seconds per encounter after discharge, versus the 30–60 minutes a human coder typically needs — and no waiting days for a coder to start.
Is the AI coding without oversight?
Health systems define what the system is allowed to code, with thresholds and audit sampling. Customers control the operating boundaries.
Has it been independently validated?
AKASA reports third-party blinded evaluations found its AI matched or exceeded expert human coders on key accuracy measures. These results have not been peer-reviewed.
Sources: Unite.AI; AKASA press announcement (akasa.com)