Analysis
Hospitals' use of artificial intelligence tools to document patient conditions during insurance claims added $942 million in extra costs across Blue Cross and Blue Shield companies in 2024 and 2025, compared with 2023 baseline levels, according to an analysis by the Blue Cross Blue Shield Association reported by TechCrunch and Technology.org.
The analysis found a sharp increase in patients being documented as having complex conditions, but BCBSA says there is "no evidence of a corresponding change in care delivered" -- meaning hospitals are coding sicker patients without treating them any differently, in the association's read of its own claims data. About $653 million of the $942 million increase came specifically from secondary diagnoses that pushed hospital stays into higher-paying billing categories, and BCBSA calculates hospitals were paid an average of nearly $12,000 more per case as a result.
Two Sides Of The Same AI Wave
This is a dispute between hospitals and insurers that predates AI by decades, but BCBSA's framing makes clear AI tools are accelerating it on both sides. Hospitals have spent the past two years deploying AI clinical-documentation and coding assistants -- the same category of product Microsoft-owned Nuance, Solventum (3M's spun-off health unit), and venture-backed startups like Fathom, Nym Health and CodaMetrix all compete in -- pitched explicitly as a fix for physician burnout and years of undercoded, underpaid claims. BCBSA's analysis argues those same tools are now systematically upcoding, whether or not that's the vendors' or hospitals' intent.
Insurers are running their own AI in parallel, on the other side of the same claims: AI-assisted prior-authorization and claims-review tools that both UnitedHealth and Cigna have faced lawsuits over for allegedly denying claims at scale. BCBSA's report doesn't address that side of the ledger, but the timing matters -- an association representing 33 independent Blue Cross and Blue Shield plans is making its cost case publicly right as scrutiny of insurer-side AI denials is also intensifying.
The Numbers In Context
$942 million is a meaningful number for BCBSA member plans specifically, but it is a fraction of the roughly $5 trillion the US spends on healthcare annually, and BCBSA's own framing -- a disconnect between coding and treatment -- is an assertion the association has an obvious financial interest in making; hospitals would likely argue the more complex coding reflects real acuity that earlier, less rigorous documentation missed. Neither side's numbers are independently audited in what's public so far.
What the fight doesn't resolve: whether AI coding tools are creating new costs or simply surfacing costs that under-documentation previously hid from insurers. Hospital groups frame the same tools as finally capturing conditions clinicians always saw but never had time to fully document -- a genuinely different read of the identical $942 million, and a real risk that either side's framing overstates its case before independent auditors weigh in.
For investors in the clinical-documentation AI category, the diligence implication is direct: any startup selling coding or documentation AI to hospitals should now expect insurer pushback on reimbursement, not just hospital enthusiasm about efficiency, and that dynamic will likely show up in slower enterprise sales cycles before it shows up in any lawsuit.
