Hansel
← The village

No. 03 · The observatory

AKASA

Overlay on hospital billing. Sells claim outcomes and prebill review, not software seats.

Status
Open
Case
One of eight recorded cases
The observatory in the gingerbread village
Snapshot
Scored 2026-10-05, reported 2026-10-09
Milestones
6 recorded, 0 with an exact day
Limits
Partial and unavailable dates stay as recorded text. The timeline plots exact days only.

AKASA

AKASA sells AI that runs on top of Epic and other hospital billing systems. Founded as Alpha Health in 2018, it rebuilt around per-hospital language models in 2024. Its customers hold about $180 billion in net patient revenue as of October 2026. That is roughly one in ten US inpatient discharges at customer hospitals. Its October 2026 autonomous coding launch has no named live production account.

Sources:

The market it attacks

The stack has three layers: the EHR core, coding software and human labor. Epic released Resolute billing in 1987 and holds 56.9% of US acute beds in 2025. NYC Health + Hospitals budgeted $289M for Epic revenue-cycle rollout alone. Solventum says over 75% of US hospitals use at least one of its unit’s tools. That unit reports $1.4B annual sales and a leading computer-assisted-coding position. A spinoff or peer combination is planned for the unit. The US employs 200,700 medical records specialists at a $51,140 median wage. A Cleveland Clinic coder can review 100-plus documents and 140,000 codes per case. Outside vendors handle coding at 76% and CDI at 55% of surveyed systems. Complex rules, per-hospital data and an aging workforce keep the layers in place.

Sources:

Origin and the first wedge

Four founders started Alpha Health in 2018 from investing, consulting and technical roles. The insight: the back end of care, not the clinic, blocked progress. Hospitals spend $0.25 to collect each $1 of revenue, per a16z’s investment note. It copied the driverless-car model: software works, remote experts handle exceptions. A 2020 survey found most finance leaders mistook RPA for AI. COVID staffing cuts favored remotely deployed automation. Its 2021 transformer hit a 512-token limit against 1,500-word charts. By 2025 its models read 50,000-word inpatient records.

Sources:

The first customers

The first wedge was claim-status checks on payer portals, routed back to work queues. It changed no claim or payer data, making it the cheapest first automation. One academic center had 43 vacancies on a revenue-cycle team over 400. AKASA says one workflow produced 20 FTEs of work. First-ten customers were never published; three early accounts are on record. Sutter Health was the largest at the June 2020 Series A. Nebraska Methodist’s CFO said automation took over claim-status checks in March 2021. Montage, a 256-bed Epic hospital, chose AKASA by RFP; status checks took four months. It worked about 3,000 claims across five payers in three months there. AKASA reports 13% fewer A/R days and 300-plus staff hours saved monthly there. Sales ran on no seats, no consultants, remote deployment and shared-savings fees. AKASA promised to be running within 90 days of receiving data. How the early accounts were found is undocumented; founder networks and a16z are the inference.

Sources:

How big it is now

Rounds: $20M June 2020, $60M March 2021, $85M total released. TechCrunch reports an unnamed $120M round in 2024 for $205M total. No company release exists for that round; Tracxn still lists $85M. Customer net patient revenue ran $100B-plus through 2023, then $145B, $160B and $180B. By 2022 customers ran 450-plus hospitals across all 50 states. Inpatient volume processed rose almost six-fold in the year to October 2026. A separate 10x claim names no metric. One in ten discharges is a customer share, not volume AKASA actually codes. That implies about 3.5 million stays a year at customer hospitals. The Cleveland Clinic coding tool reached all US locations in four months in 2025. The Clinic’s CDI tool followed a pilot. AKASA says it processes 100% of the Clinic’s enterprise inpatient volume. Named customers include Duke, Johns Hopkins and Stanford; revenue is undisclosed. The reported 2024 round came an inferred 39 months after Series B. US digital-health funding fell from $29B in 2021 to a $2.5B quarter in 2023. US headcount fell 244 to 97 from January 2023 to February 2026; Tracxn estimates 247.

Sources:

What made it hard

Coding sets payment, so captured codes appear to payers as cost. Its CEO estimates inpatient coding drives 50% to 60%-plus of system revenue. The BCBS Association ties $942M in added 2023-25 costs to hospital AI coding. Its analysis did not prove AI caused the increase. A July 2026 GAO spotlight found few independent accuracy studies. AKASA migrates no core system but trains a model per customer on its records. Generic and clinical models fell short; its pre-LLM model never commercialized at scale. It invested in its own GPU capacity to train per-hospital models. Buyers name integration, cost and security as barriers; cost leads at 52.5% for large systems. Epic says over 85% of its customers use some AI. Epic’s Penny autonomously codes radiology and ED; surgery and pathology stay planned. Oracle Health added coding suggestions for ambulatory professional fees only. An analyst reports Toolbox partner apps sit under Epic non-compete terms. AKASA’s 2024 no-replacement position became 2026 designed-for autonomy.

Sources:

The strategy

Stage one, 2018 to 2023: bots plus remote experts inside the customer’s EHR. Stage two, 2024: assistants whose suggestions coders accept or revise. Authorization Advisor arrived March 2024; a CDI product followed in 2025. It added a post-human reviewer of every inpatient encounter before billing. Sales start with a retrospective review; fees begin only after measured gain. Stage three, October 2026: autonomous coding in under 90 seconds, against 30-to-60-minute human review. Blinded third-party tests covered encounters at 65% to 85% of inpatient volume. AKASA says results matched or beat human coders on MS-DRG and principal diagnosis. Quality capture and present-on-admission accuracy were also scored. No named live buyer runs it. Outpatient facility encounters are the stated next step. The billing core stays untouched; overlays take the auditable judgment step. SmarterDx grew from about $50M to $150M after its buyout. It is merging with Datavant, backed by $2.7B of preferred equity. Commure raised $70M at a reported $7B valuation. Capstone says autonomous coding spend rose from $200M in 2024 to $450M in 2025. The counter-bet is Candid Health, which rebuilt the billing core. It raised $120M in July 2026. Olive raised over $900M to a $4B valuation and 900-plus hospitals. Axios found a fraction of promised savings; Olive shut down in October 2023. Epic made Olive stop using its name in marketing. An HFMA column says its AI often proved to be screen scraping. Thoughtful’s stand-alone service ended inside its parent, a portfolio call not proof of failure. Its ending notice ran about 90 days; a competitor reported unsanctioned bots.

Sources:

Milestones

  • June 2020: Sutter Health named largest customer at Series A.
  • March 2021: Claim-status checks automated at Nebraska Methodist.
  • August 2021: Customers pass $100B net patient revenue.
  • June 2024: Generative coding assistant launched.
  • during 2025: Cleveland Clinic coding rollout across US locations.
  • October 2026: Autonomous inpatient coding launched.

Sources:

What others can copy

Enter after the work is done: test whether the first product changes any frontline screen. At Cleveland, coders accepted about half of AKASA’s flags in 60 days. Prove value on last year’s data: test a retrospective dollar figure before full rollout. Pick the unshipped judgment step: read the platform’s published AI feature list. Earn autonomy in stages: publish the volume share that runs with no human. Build for the auditor: log whether software or a person made each call. Stay narrow: test the share of revenue coming from one workflow. These are founder tests to run, not results already achieved.

Sources:

Open questions

Autonomy is validated only by vendor-run blinded tests so far. Fees rise with captured revenue, the same capture a payer group calls cost inflation. Epic’s Penny promises more autonomy; Walaliyadde argues breadth makes deep single-task products unlikely. Revenue, margins, churn and the reported 2024 round stay unconfirmed. Per-hospital fine-tuning may keep margins closer to services than software. How many 2021-23 automation customers remain is not public.

Sources:

Limits

All scale figures are company-reported; no revenue is disclosed. The customer-hospital denominator is not coded volume. Headcount sources disagree by more than two-to-one. Olive’s savings shortfall rests on a secondary Axios capture.

Sources:

Exhibit 8Explore · open a case

Milestones in the AKASA case

Exact-day events on a UTC timeline, one lane per case.

  • milestone

How to readEach dot is a dated milestone. Only exact dates draw dots.

NoteMonth- and year-only milestones stay in the tables, never plotted.

SourceLegacy Software Atlas. Frozen scoring snapshot of 5 October 2026. Report version 4 dated 9 October 2026.

Recorded milestones

Every recorded event for this case, in frozen order. Statuses stay as recorded.

MilestoneDateStatusSources
Sutter Health named largest customer at Series AJune 2020reportedAKASA (as Alpha Health), Series A press release, 10 Jun 2020.
Claim-status checks automated at Nebraska MethodistMarch 2021liveAKASA, Series B press release, 23 Mar 2021.
Customers pass $100B net patient revenueAugust 2021reportedAKASA, press release, 24 Aug 2021.
Generative coding assistant launchedJune 2024reportedAKASA, press release, 18 Jun 2024.
Cleveland Clinic coding rollout across US locationsduring 2025liveAKASA, press release, 13 Oct 2025.
Autonomous inpatient coding launchedOctober 2026reportedAKASA, press release, 2 Oct 2026.

Back to the cases chapter and the full milestone table.