Bots vs. Bots: What the Blue Cross AI Coding Fight Means for Your Revenue

On September 24, the Blue Cross Blue Shield Association released an analysis estimating that rising hospital coding intensity, which it links to AI coding tools, added $942 million in spending for Blue plans between 2023 and 2025. Reuters, the New York Times and much of the trade press covered it within days. BCBSA's Luke Chalker put it this way, per TechCrunch: "It's not a war. It's a completely one-sided blood bath." In his telling, insurers are the ones losing.

Hospitals make the mirror-image complaint. Becker's, drawing on revenue cycle leaders at 17 health systems, described hospitals as losing an "increasingly unsustainable" payer automation race.

Both sides can be right, because they are describing different parts of the same machine. Payers lead on automated denial and claim review. Providers lead on documentation and coding. Neither side is winning. The loser is the system that pays for both, and the clearest measure of that cost is the $43 billion hospitals spent in 2025 trying to collect for care they had already delivered.

What BCBSA found, and what it concedes

According to Fierce's account of the white paper, the share of Blue inpatient claims coded as medically complex rose from 37% in early 2023 to 40% by late 2025. Roughly 70% of the added cost, more than $650 million, came from secondary diagnoses that moved claims into higher-paying DRGs. BCBSA says this happened as more than 60% of hospital systems adopted AI tools that scan lab results and records.

The core argument is that diagnoses rose while treatment did not. In major bowel surgery, hospitals with the fastest complexity growth coded posthemorrhagic anemia in 13.7% of cases, compared with 9.9% elsewhere, yet transfused fewer of those patients (16.9% versus 19.3%). "If patients are truly sicker, we'd expect to see more treatment," Chalker said.

BCBSA's own caveat matters. The analysis relies on claims, not clinical documentation, which the association acknowledged would be a more direct test. Even its vice president of clinical affairs, Razia Hashmi, hedged: "There may be an element of correct coding there, but the likelihood that this is technology-enabled upcoding is higher, in my view."

The hospital rebuttal

The American Hospital Association had already called these claims "unsubstantiated" in a fact sheet dated August 2026. Its case is that patients are sicker. Hospital case-mix index rose about 5% from 2019 to 2024, and simpler care keeps moving to outpatient settings. The AHA also turned the charge back on insurers, citing a MedPAC finding that upcoding contributed to $40 billion in overpayments to Medicare Advantage plans. Its sharpest line: insurers "want it both ways," claiming sicker enrollees for their own risk scores and healthier ones when claims arrive.

Coding vendors make a structural point. "We live in a fee-for-service system, we pay for volume, not for value," Solventum's Dr. Travis Bias told Healthcare Dive. Even CMS Administrator Dr. Mehmet Oz conceded that "short term, AI is going to be inflationary."

The other front: denials

On review and denials, payers hold the advantage, and the economics explain why. The AHA estimates hospitals spent $43 billion in 2025 collecting payments insurers owed, including nearly $18 billion on overturning denials alone.

A Health Affairs study of 2019 data found Medicare Advantage plans denied 17% of initial claims, and 57% of those denials were ultimately overturned. Most denials ended up as delayed payments, not avoided ones, but providers still lost about 7% of net MA revenue. For a payer, a denial is an automated edit that costs close to nothing. For a provider, every appeal takes staff time, clinical documentation and time waiting to be paid.

The AHA describes insurer "downcoding programs" that "use automated edits to unilaterally reduce reimbursement... without reviewing medical documentation." HCA CFO Mike Marks said on the company's April 2026 earnings call that "the denials and underpayments are still really high," with Medicare Advantage the main driver. That is the largest U.S. for-profit hospital operator, after years of revenue cycle investment.

Outpatient and ASC groups are on the denial side of this fight

BCBSA measured inpatient DRG coding. Multi-site outpatient, ASC and procedural groups mostly feel the pressure from the denial side: prior authorization, medical necessity review and automated downcoding of claims that were coded correctly. For those groups the arms race shows up less in code capture and more in the rising share of staff hours spent defending payment after the case is done. Chalker also said BCBSA plans future analyses of outpatient care, so coding scrutiny is likely coming next.

Systems are rebuilding the machinery

Large systems are treating revenue cycle as core infrastructure. CommonSpirit agreed to pay Tenet about $1.9 billion over three years to exit a Conifer outsourcing contract that ran through 2032, while Conifer paid CommonSpirit about $540 million to redeem its 23.8% stake. Conifer then filed a WARN notice for 1,037 layoffs, citing "restructuring and technology initiatives related to, among other things, no longer providing services to CommonSpirit Health after October 30, 2026."

On September 9, UF Health and R1 announced an AI-native revenue cycle model on R1's Phare OS platform, with engineers from R1's R37 lab embedded alongside UF Health teams. The release describes agent-driven workflows paired with continuous human oversight.

The backdrop is getting harder

Premier projects a hospital revenue impact of about $68.6 billion over 2026 and 2027 from rising uninsurance tied to the 2025 federal budget law. Fewer insured patients means every paid claim counts for more.

States are pushing back on the payer side of the fight. As of May 5, ASCO had tracked over 130 prior authorization bills in 42 states in 2026, with five signed into law. Utah's SB 319 sets 72-hour decisions for urgent requests and seven calendar days for nonurgent ones, keeps authorizations for chronic conditions valid for 12 months, and bars retroactive denials or clawbacks for approved services, with limited exceptions. It also requires insurers to disclose AI use in prior authorization. Washington now prevents insurers from using AI as the sole means to deny, delay or modify services.

What this means

The BCBSA report does not settle whether AI coding is accurate or inflationary. What it does show is that the argument has moved from anecdotes to claims-level evidence, and that Blue plans say they are using data to set expectations for hospitals that use AI tools. Providers are now defending on two fronts: how they code, and how they collect.

The questions worth asking are financial ones. How much of your revenue cycle spend goes to recovering payment for care you already delivered? When a payer's automated edit reduces a claim, how quickly do you know, and what share do you let go? If your coded acuity rose over the past two years, would your documentation show the treatment that goes with it? And do you know whether you are losing more to denials than you gained from better capture?

Both sides are spending heavily on software to fight over the same dollars. In that kind of arms race, organizations that still treat revenue cycle as a back-office function are funding the other side's operating leverage with their own revenue.