Payer AI vs. Provider AI: Why the Bot Wars Are Costing Healthcare More

Payer AI vs. Provider AI: Why the Bot Wars Are Costing Healthcare More

The promise of AI in healthcare administration was straightforward: automate the paperwork, shrink the overhead, and let clinicians focus on patients. The reality documented in a major April 2026 report from the Peterson Health Technology Institute (PHTI) is considerably messier. When both payers and providers deploy AI simultaneously, the two systems can end up fighting each other -- adding cost and complexity rather than removing it. In the revenue cycle world, medical coders are caught in the middle.

What PHTI Found

The Peterson Health Technology Institute published its findings in April 2026 from a workshop convened with senior leaders across health systems, payer organizations, and federal agencies. The central finding was blunt: there is currently no evidence that AI in healthcare administration "translates to lower average cost per claim factoring in the cost of the AI solution."

The report identified a pattern it called the "bot wars" dynamic. When providers deploy AI to automate prior authorization submissions, and payers deploy AI to triage and scrutinize those same requests, the result is often an escalating back-and-forth. More rounds of automated appeals. More automated denial responses. More volume without resolution. The per-transaction cost falls on individual steps, but the total number of transactions multiplies.

The same dynamic is playing out with medical coding. AI ambient scribes on the provider side generate more thorough documentation -- which generates higher-acuity code suggestions. Payers, in turn, have deployed algorithmic downcoding tools to challenge those submissions. Neither system "wins" on its own terms; they simply create more work for everyone, including human coders who still have to review the escalated cases.

How the Escalation Loop Works in Medical Coding

The mechanics are worth understanding in detail. Consider a hospital system that deploys an AI ambient scribe. The scribe captures more complete clinical details -- comorbidities, severity indicators, procedure complexity -- that would previously have been missed or underdocumented. The coding AI then surfaces higher-specificity ICD-10-CM codes and more accurate CPT codes, many of which carry higher reimbursement.

On the payer side, claims with higher complexity codes now flag more frequently in the automated review queue. The payer's AI requests additional documentation or issues an automated denial pending review. The provider's AI generates an automated appeal. The loop continues through multiple rounds -- each step handled by software, but each step also adding latency, staff review time, and overhead that wasn't there before the AI was deployed.

PHTI workshop participants specifically cited three patterns that make the bot wars costly:

• Multiplied communication volume -- automated back-and-forth exchanges generate more correspondence per claim without resolving the underlying coding question faster.

• Limited impact on complex cases -- AI performs well on routine authorizations and straightforward coding scenarios, but complex cases still require human intervention at multiple points.

• Unintended inflation -- ambient scribes are increasing documentation thoroughness in ways that payers characterize as upcoding and providers characterize as accurate capture, with no agreed standard for resolution.

The Missing Standard: Who Decides What Accurate Coding Looks Like?

Underneath the bot wars is an older, unresolved dispute. ICD-10-CM and CPT coding guidelines require coders to capture diagnoses and procedures "to the highest degree of specificity supported by the documentation." AI scribes are getting better at creating documentation that supports higher specificity. But payer contracts and fee schedules were calibrated to a world where that documentation was often incomplete.

There is no current CMS policy or AMA guideline that adjudicates what happens when AI-assisted documentation routinely surfaces codes that were technically always correct but previously undercoded. The PHTI report noted that payer and provider leaders who attended the April 2026 workshop agreed on almost nothing -- except that AI scribes are raising costs. What they disagreed on was whether that rise represents legitimate revenue recovery or inflation of claims.

For medical coders and coding managers, this ambiguity is consequential. A coder validating AI-suggested codes is now implicitly making a judgment call about what a payer's AI will accept, not just what the guidelines require. That is a new type of pressure that did not exist two years ago.

What Revenue Cycle Teams Should Do Right Now

The PHTI findings do not argue that AI is wrong or that providers should pull back. They argue that the current deployment wave is happening without a shared framework for measuring whether it actually reduces system-wide administrative burden. For revenue cycle and coding teams, that means several practical steps matter more than waiting for the industry to sort itself out.

First, track denial rates by AI-suggested code versus human-suggested code. If your AI coding tool is surfacing codes that are denied at significantly higher rates than your baseline, that is a signal that the bot war is costing you more than the automation saves. Second, build appeal workflows that can distinguish between algorithmic denials (which often resolve with standard documentation) and human-reviewed denials (which require more nuanced responses). Third, audit whether your payer contracts have kept pace with what AI documentation actually supports -- in many cases, the fee schedule and the medical necessity criteria were written before AI scribes existed.

What This Means for AI Coding Tools

The bot wars dynamic is a reason to be more exacting about which AI coding tools you choose -- not less likely to adopt AI. A tool that generates high-specificity codes without also generating the supporting documentation rationale is a liability in the current environment. A tool that helps coders understand why a payer's AI will likely flag a given code, and what addendum or clinical detail would resolve the flag, is considerably more valuable.

This is also where agentic AI has a structural advantage over static auto-coding. An agentic system can reason through the payer's likely review criteria, identify which clinical facts in the note support the code assignment, and flag where documentation gaps exist before the claim is submitted. That kind of pre-submission reasoning directly addresses the bot wars problem: it reduces the volume of automated denials rather than automating the appeals that follow them.

The PHTI April 2026 report is worth reading in full, available at phti.org. The findings will likely inform CMS and payer policy discussions through 2026 and into 2027, making them directly relevant to anyone building or buying AI-assisted coding workflows.

The Bottom Line

AI in healthcare administration is not failing -- but it is not yet delivering the system-wide savings that justified the investment. The bot wars are a symptom of deploying AI in an environment where payers and providers have incompatible incentives and no shared standard for what accurate coding looks like. Coders who understand this dynamic are better positioned to evaluate AI tools, manage denial workflows, and make the case internally for the right kind of automation.

If your organization is evaluating or deploying AI for medical coding, Medikode's automated medical coding platform is built specifically to improve accuracy and reduce denials -- with the kind of reasoning transparency that helps coders stay ahead of payer scrutiny, not just respond to it after the fact.