Medical coding has long been a provider problem. Physicians submit claims; payers process or deny them. The two sides of that transaction rarely bet on the same technology. So when CVS Health Ventures — the venture capital arm of one of the largest payers in the United States — announced a strategic investment in Fathom, a leader in autonomous medical coding, on May 4, 2026, it marked something genuinely new: a payer putting money into the accuracy of the claims its counterparts submit.
The investment is worth examining not just as a financial headline but as a signal about where autonomous coding is headed and what it means for every revenue cycle team still running a traditional coding workflow.
What Fathom Does
Fathom’s platform applies deep learning and natural language processing to the full patient encounter, automatically assigning ICD-10, CPT, and HCC codes across inpatient and outpatient settings. Unlike earlier computer-assisted coding tools that prompted human coders to approve or reject suggestions, Fathom is fully autonomous: it completes the coding workflow and routes only edge cases to human review.
The company reports 90% or higher automation rates across multi-specialty health systems in a single deployment, an approach that contrasts with the department-by-department rollout model most legacy vendors require. That system-wide implementation speed attracted attention in the KLAS “Autonomous Coding 2025” report (August 2025), where Fathom received a 95.5/100 overall performance score and was subsequently named the #1 solution for Reducing Cost of Care in KLAS’s 2025 Emerging Solutions Top 20 Report.
Why a Payer Is Backing a Provider Tool
The CVS Health Ventures investment raises an obvious question: why would a payer invest in technology that helps providers code more completely?
Carter Prince, Partner at CVS Health Ventures, addressed this directly in the May 4 announcement (BusinessWire). He described Fathom’s accuracy as having “potential to become a trusted, consistent, and neutral representation of the care delivered,” creating new opportunities to strengthen payer-provider relationships.
The phrase “neutral representation” is the key. From a payer’s perspective, accurate coding is not a zero-sum game. When autonomous coding produces documentation that reflects actual clinical work — no undercoding, no overcoding — claims data becomes more reliable for risk stratification, prior authorization, and care management. Inaccurate coding costs payers too: upcoded claims inflate medical spend, while undercoded populations skew actuarial models and risk adjustment calculations.
CVS Health, through Aetna, covers tens of millions of beneficiaries. When the care those members receive is coded accurately, the data downstream becomes more useful to both sides of every transaction.
The KLAS Numbers: What 95.5/100 Means in Practice
Establishing the Autonomous Coding Benchmark
The KLAS score deserves context. The “Autonomous Coding 2025” report (August 2025) was the first time KLAS formally evaluated autonomous coding as a distinct market category. Fathom’s 95.5 overall performance score came from customer interviews at real health systems — actual users, not marketing claims — with above-average ratings on 12 of 14 numeric indicators. Customers unanimously validated Fathom’s 90%+ automation rates at high accuracy.
What Deployment Looks Like
The benchmark translated to published results in March 2026, when Your Health, a multi-specialty physician group, released outcomes from its Fathom deployment: a 95.5% automation rate at 98.3% accuracy across all service lines (BusinessWire, March 19, 2026). That combination — automation above 90% and accuracy approaching 99% — is the threshold revenue cycle leaders cite as the point where autonomous coding becomes economically viable without a dedicated human review layer on every chart.
What This Investment Signals for the Industry
The Fathom deal is not an isolated event. It connects to a broader convergence that is accelerating autonomous coding adoption across the market:
- Documentation complexity is rising. FY2027 ICD-10-CM adds 238 new codes effective October 1, 2026, and HCC Model V28 expanded risk-adjustment categories from 86 to 115, raising the documentation burden on coding teams working manually.
- Staffing constraints persist. Experienced coders are retiring faster than new credentialed coders are entering the workforce; automation reduces direct dependency on headcount to manage volume.
- Regulatory pressure is intensifying. The DOJ’s 2026 national healthcare fraud takedown charged 455 defendants over $6.5 billion in false claims, CMS forced Elevance Health to repay $342 million in Medicare Advantage overpayments, and RADV audit activity is ongoing — all of which make auditable, defensible coding a compliance priority, not just an efficiency one.
The Bain & Company 2025 Provider and Payer Healthcare IT Survey found that revenue cycle management had risen to the top priority for health IT investment across both providers and payers. Autonomous coding sits at the center of that priority. When a payer’s own venture capital arm backs a provider-facing coding platform, it signals that AI coding accuracy has crossed a credibility threshold that matters to the entire ecosystem.
What Coders and RCM Leaders Should Take From This
How to Evaluate Your Current Stack
The Fathom KLAS results establish a useful market floor for comparison. Any autonomous or computer-assisted coding tool currently under evaluation should be benchmarked against automation rate (the industry baseline is moving toward 90%+ for mature platforms), first-pass accuracy (high-performing systems are approaching 98-99%), specialty coverage breadth (single-specialty tools create coverage gaps across a health system’s full service mix), and auditability (how the system documents and justifies each code assignment for compliance review).
Teams still running primarily manual coding workflows should revisit their business case. The staffing math has shifted: at 90%+ automation rates, autonomous coding is no longer a pilot-stage experiment. At that performance level, it changes how RCM departments are staffed, how coders spend their time, and how quickly claims move through the revenue cycle.
A payer investing in autonomous coding is not a curiosity. It is a concrete signal that AI-generated coding is becoming the expected standard of care for revenue cycle operations — and that the accuracy bar is high enough that both sides of the claims transaction are prepared to rely on it.
To see how autonomous AI coding applies to your specific organization, explore Medikode’s automated medical coding platform and evaluate what auditable, multi-specialty AI coding can deliver for your revenue cycle.