There is a quiet war inside American healthcare. It is fought in milliseconds. At the center sits one force reshaping everything: AI in the US healthcare revenue cycle. This technology now decides who gets paid, who gets denied, and who gets stuck waiting.
On one side stand the payers. Their AI scans your claim in milliseconds. It cross-references hundreds of denial triggers. Then it rejects the claim before your biller finishes a first coffee. The AI does not sleep. It does not negotiate. It does not give your documentation the benefit of the doubt.
On the other side stand the providers. They race to deploy their own AI. These tools draft appeals, predict denials, and recover lost revenue.
A regulator sits in the middle. CMS now tries to rewrite the rules while the game is already underway. This is the defining strategic tension in US healthcare today. Let me break down what is happening. I will also explain why most organizations fight the last war.
How AI in the US healthcare revenue cycle broke the old economics
Start with an irony almost nobody saw coming. AI did not just make providers more efficient. It changed the math of who gets paid.
The primary impact of generative AI is not what providers bill for. It is the accuracy and speed of how they do it. AI tools can auto-generate evaluation and management codes, identify missed charges, and augment diagnosis specificity. This can shift risk scores. Even modest changes in risk assessment move payer reimbursement and overall costs.
Now multiply that across millions of encounters. You get a structural shift. Coding tools have fundamentally shifted the economics of reimbursement, resulting in complex claims paid out with greater speed than payers have ever experienced. This creates new dynamics for cost oversight.
In plain terms, providers found revenue they were leaving on the table. And it works. One 200-physician orthopedics group used AI-powered contract management. It recovered $10.3 million in hidden underpayments. That came from just seven commercial payers.
This efficiency lands directly on the payer balance sheet. AI-enabled coding now adjudicates claims on the first pass, often at higher levels. Traditional cost-containment cannot keep pace. So payers did the predictable thing. They armed up.
The counterattack โ AI denials at scale
Here is where it turns into an arms race. The surge in AI-enabled coding pressures payers. Many now lean harder on denials, payment delays, and prior authorizations.
This is not theoretical. Health insurers and providers increasingly use AI tools in prior authorization and claims. The denial numbers climb fast. Between 2022 and 2023, care denials rose 20.2% for commercial claims. They jumped 55.7% for Medicare Advantage claims.
But the system denies claims it should not. Insurers eventually overturn 75% of Medicare Advantage care denials. That reveals deep inefficiency. KFF data confirms the pattern. Medicare Advantage insurers made 52.8 million prior authorization determinations in 2024. They denied 4.1 million, or 7.7%, of requests. Patients appealed only 11.5% of those denials. Yet appeals succeeded 80.7% of the time.
Read that again. Four out of five appealed denials get reversed. Yet barely one in ten denials ever gets appealed. That gap is the business model. The denials that stick are not always correct. They are the ones nobody fought.
This is exactly why providers now weaponize AI of their own.
Provider AI: fighting fire with fire
The provider toolkit matured fast. It points squarely at the denial machine.
Predicting denials before submission. Predictive AI assigns each claim a probability-of-denial score. Teams then intervene before they submit. The results are real. Kaiser Permanente ran a six-month pilot of an AI denial-management platform. It cut initial denial rates by 17%. The tool flagged eligibility mismatches and missing documentation.
Drafting appeals at machine speed. Provider-facing tools rank denials by likelihood of reversal. They weigh the payer and the service. Generative AI then drafts the appeal letter. It locates clinical information in the EHR and summarizes it to answer the insurer’s objection. In practice, generative AI drafts appeal packages roughly three times faster than manual work.
Catching systemic underpayment. This is the quiet killer. Imagine a commercial payer underpaying one CPT code by 3% across hundreds of locations. Human auditors would miss it. A well-tuned AI catches the discrepancy almost every time.
The strategic logic is brutal and clear. Insurance companies did not invest in AI to improve your cash flow. They invested to protect their own. Providers simply respond in kind.
The referee enters: CMS-0057-F
Now the regulator steps in. The CMS Interoperability and Prior Authorization Final Rule (CMS-0057-F) took effect January 1, 2026.
The rule matters. CMS now requires impacted payers to send urgent prior authorization decisions within 72 hours. Standard decisions must arrive within seven calendar days. And here is the part that fuels the appeals war. Starting in 2026, payers must give a specific reason for every denied prior authorization.
The rule also forces transparency. Payers must publicly report prior authorization metrics. They must explain denials clearly. This builds accountability. CMS projects roughly $15 billion in savings over ten years.
Here is the strategic catch most people miss. The rule covers only a slice of the market. It applies to Medicare Advantage, Medicaid and CHIP managed care, state Medicaid and CHIP fee-for-service, and QHP issuers on federal exchanges. State laws govern commercial payers outside these categories. Those laws vary widely.
The broader regulatory patchwork keeps forming. By early April 2026, at least 25 states issued guidance based on a 2023 NAIC model bulletin. That bulletin covers AI across the full insurance life cycle, including claims payment.
The twist: the rule may be making denials worseโfor now
Here is the counterintuitive part. The reform meant to reduce friction has, for now, coincided with more denials.
Prior authorization denials jumped 31% year-over-year in 2026. That number is not an anomaly. It is a signal. CMS-0057-F went live on January 1, 2026. It aimed to speed up approvals. Yet many practices now face more denials, not fewer.
Why? Several factors collide at once. Payers expanded the list of services that require prior authorization in 2026. More services mean more chances for a submission to fail. On top of that, 288 new CPT codes went live on January 1, 2026. Another 614 new ICD-10-CM codes took effect October 1, 2025. Practices that skipped updates now submit outdated codes. Payers reject them.
In other words, faster and clearer denials are still denials. Speed and structure do not equal approval.
The deeper problem – an automation feedback loop
Zoom out and you see something worrying. In an AI-versus-AI dynamic, errors do not just happen. They propagate.
A single modifier error on a multi-procedure case costs you more than one claim. It establishes a denial pattern. Payer AI then flags that pattern for future scrutiny. One side’s mistake becomes the other side’s training data.
The governance literature sounds the alarm. AI in utilization review raises hard questions. These include the role of the human in the loop, the opacity of algorithmic decisions, and automation bias. Researchers give a blunt assessment. So far, institutional governance has not met the challenge of responsible use.
The legal system already engages too. Patients have filed class action lawsuits over specific denial algorithms. They argue the denials skipped individual assessment. They also cite a lack of transparency about the data behind the AI. These cases still move through the courts.
What healthcare leaders should do now
So where does this leave you? The AI in US healthcare revenue cycle race will not slow down. But you can compete smarter. Here is where I would focus.
Invest in denial prevention, not just appeals. Stop the denial before it happens. Clean coding and eligibility checks beat reactive letters.
Update your code sets immediately. Refresh fee schedules, superbills, and EHR templates now. Outdated codes hand payers an easy denial.
Build governance before you scale. Put a human in the loop. Audit your models. Document your logic. Regulators and courts will ask.
Treat CMS-0057-F as leverage. Use the new denial-reason requirement to sharpen appeals. A specific reason is a specific target.
The next 18 months will define the economics of US healthcare. The winners will not own the most AI. They will govern it best.
For more such articles : RoamReadReflect
Related posts
https://www.linkedin.com/pulse/from-gut-feel-gen-ai-how-rewriting-ma-playbook-girish-kumar-kdhkc
https://www.linkedin.com/pulse/why-change-management-makes-breaks-ai-healthcare-girish-kumar-7lpec
“๐๐๐ฒ ๐ฒ๐จ๐ฎ๐ซ ๐๐ก๐จ๐ข๐๐๐ฌ ๐ซ๐๐๐ฅ๐๐๐ญ ๐ฒ๐จ๐ฎ๐ซ ๐ก๐จ๐ฉ๐๐ฌ, ๐ง๐จ๐ญ ๐ฒ๐จ๐ฎ๐ซ ๐๐๐๐ซ๐ฌ
Leave a Reply