Why ACOs Are Losing Millions to Claim Denials—And How Agentic AI Fixes the Revenue Cycle

By Gautamdev Chowdary, co-founder and CTO, Zynix AI · Updated October 1, 2026

Key takeaways

ACOs lose shared savings to claim denials because unworked denials cost revenue, distort the expenditure data behind future benchmarks and leave risk scores understated. Agentic AI helps by checking claims against payer rules before submission, spotting denial patterns across practices and drafting appeals, with staff handling exceptions, all on the same data that runs quality and care coordination.

KFF found that HealthCare.gov insurers denied about one in five in-network claims in 2024, so denial prevention belongs in ACO strategy, not only in billing.
Shared Savings Program benchmarks are built from an ACO’s own historical spending, so revenue cycle errors can echo into future benchmarks and risk adjustment.
Multi-practice, multi-EHR networks hide denial root causes unless denials are analyzed across the whole network by payer, practice and reason code.
Agentic systems validate claims before submission, adapt to payer rule changes and draft appeals, while staff review exceptions.
A practical start is a 90-day denial audit, a map of your top payers’ rules, a pre-submission validation step and closed-loop analytics.

The Revenue Cycle Is the Silent Killer of ACO Shared Savings

In Performance Year 2024, 75% of the 476 MSSP ACOs earned performance payments, totaling $4.1 billion. That number sounds impressive until you realize how much they left behind. U.S. healthcare organizations lose large sums every year to revenue cycle inefficiency, and ACOs—with their multi-practice, multi-EHR complexity—are disproportionately exposed.

Here is the uncomfortable truth: most ACOs treat revenue cycle management (RCM) as a back-office function. They invest in care coordination, quality reporting, and population health dashboards. Meanwhile, insurers on HealthCare.gov denied 19% of in-network claims in 2024, according to KFF, and every denied claim that is never corrected and resubmitted is revenue lost for good. That is not a billing problem. That is a structural leak in your shared savings model.

Why ACOs Face a Unique Denial Management Challenge

Fee-for-service practices have denial problems. ACOs have denial compounding problems. Here is why:

1. Multi-Practice Heterogeneity

A typical MSSP ACO includes many participating providers across primary care, specialty, and post-acute settings. Each practice may run a different EHR, use different billing workflows, and have different coding accuracy baselines. When you aggregate claims across that network, denial root causes become nearly impossible to diagnose manually.

2. Benchmark Sensitivity

In the Shared Savings Program, CMS builds each ACO’s historical benchmark from the past expenditures of beneficiaries assigned to it and then adjusts it, including for regional spending. That means your own past expenditure data carries real weight. If denied claims lead to delayed or lost reimbursement—and your expenditure data looks artificially low as a result—your future benchmarks tighten. You are punished twice: once on the revenue you never collected, and again on the benchmark that assumes you never needed it.

3. Risk Score Constraints

CMS also caps how much an ACO’s risk scores can rise over an agreement period. For ACOs that are under-coding because claims keep getting denied and reworked, this creates a perverse incentive trap. Your risk adjustment does not reflect the true acuity of your population, your benchmark is lower, and your shared savings shrink.

The Denial Landscape in 2026: What the Data Shows

The numbers are stark for organizations that have not invested in automation:

19% average denial rate across in-network claims in HealthCare.gov marketplace plans (KFF, 2024 data)—similar to 2023, and still catastrophically high for value-based organizations.
A real cost for every reworked claim—and that assumes you rework it at all. Most organizations do not.
Revenue cycle leaders warn that their RCM operations will be less effective in 2026 unless they make structural changes.
Top denial drivers: missing/invalid prior authorization, incorrect patient demographics, coding errors, timely filing failures, and medical necessity documentation gaps.

For ACOs specifically, the problem concentrates in three areas: prior authorization failures (especially as WISeR adds new PA requirements in six states), specialist referral claim mismatches, and post-acute care billing errors where the ACO has the least operational control.

What Agentic AI Actually Changes

The term "AI in revenue cycle" has been thrown around since 2019. Most of what shipped was rules-based automation with a machine learning label. The shift happening in 2026 is fundamentally different: agentic AI systems that can work through multi-step revenue processes and hand exceptions to staff.

Here is what that means in practice:

Pre-Submission Denial Prevention

An agentic AI system does not just flag a claim that is missing a modifier. It identifies the missing documentation, retrieves the correct clinical note from the EHR, validates it against payer-specific rules, cross-references the patient's coverage and authorization status, and corrects the claim—all before it hits the clearinghouse. McKinsey's recent analysis describes this as the "race to a touchless revenue cycle," and the ACOs that get there first will have a structural financial advantage.

Continuous Payer Rule Adaptation

Every major payer updates their claims editing logic quarterly. UnitedHealthcare has different rules than Aetna, which has different rules than regional Blues plans. An agentic system continuously ingests payer rule changes and adapts submission logic accordingly—no manual update cycles, no lag time where claims slip through with outdated formatting.

Cross-Practice Pattern Recognition

This is where agentic AI becomes uniquely powerful for ACOs. When you have 100+ practices submitting claims, an AI agent can identify that Practice A's denial rate for E/M codes spiked 8% last month because of a coding template change, while Practice B's timely filing rate dropped because their clearinghouse integration broke. It surfaces these patterns, assigns root causes, and can even trigger corrective workflows—reassigning claims to a rework queue, alerting practice managers, or adjusting future claim logic automatically.

Appeals Automation

When denials do occur, agentic AI drafts the appeal, attaches the relevant clinical documentation, matches the appeal to the specific denial reason code, and submits through the correct payer channel. Early adopters are reporting lower denial rates and additional recovered revenue.

Why Denial Management Belongs in Your Care Data Infrastructure

At Zynix, we have seen this problem from the inside. Our co-founder and CEO, Jay Chowdappa, MD, is a physician who led ACOs before we built the AI infrastructure Zynix AI provides today. The pattern was always the same: brilliant clinicians doing transformative care work, undermined by revenue cycle failures that nobody had the bandwidth to fix.

What we have learned working with ACOs is that denial management cannot be bolted on. It has to be woven into the same data infrastructure that powers your quality reporting, your care coordination, and your risk stratification. Here is why:

The same clinical data that drives your HEDIS measures drives your claim accuracy. If your AI platform already ingests ADT feeds, lab results, and encounter data for care gap closure, extending that pipeline to pre-submission claim validation is an incremental lift—not a new system.
Prior authorization intelligence feeds both care delivery and billing. With the CMS prior auth transparency rule now in effect and WISeR adding PA requirements in Arizona, New Jersey, Ohio, Oklahoma, Texas, and Washington, you need a unified system that tells the care team what needs authorization and ensures the claim reflects that authorization was obtained.
Denial analytics should inform care model design. If 30% of your denials come from post-acute claims, that is a signal about your discharge planning workflow—not just your billing department.

Where to Start: A Practical Eight-Week Roadmap

If you are an ACO operator reading this, here is where to start:

Week 1-2: Denial Root Cause Audit

Pull 90 days of denial data across your full provider network. Categorize by denial reason code, payer, practice, and service type. Most ACOs have never done this at the network level—they have practice-level visibility at best.

Week 3-4: Payer Rule Mapping

For your top five payers by claim volume, document their current prior authorization requirements, claims editing rules, and appeals submission processes. This becomes the knowledge base your AI system will use.

Week 5-8: Pre-Submission Validation Pipeline

Deploy an agentic AI layer between your EHR/practice management systems and your clearinghouse. This layer validates every claim against payer-specific rules, checks authorization status, flags documentation gaps, and either auto-corrects or routes to human review before submission.

Ongoing: Closed-Loop Analytics

Connect your denial outcomes back to your upstream workflows. Every denial should generate a signal that improves future claim accuracy. Over 90 days, a well-implemented system should measurably narrow your clean claim rate gap.

The Economics Are Unambiguous

Let us do the math for a mid-sized ACO with 30,000 attributed beneficiaries and $400M in total cost of care:

Assume 15% of claims face initial denial
Average rework cost: $50 per claim
Claims volume: ~500,000 annually
Denials: 75,000 claims
Rework cost: $3.75M annually
Unrecovered revenue from denied claims that are never resubmitted: a further loss on top of the rework cost

Now factor in the benchmark impact. If even a small share of those denials create expenditure reporting anomalies, your shared savings calculation shifts. For an ACO operating on thin shared savings margins, that can be the difference between a distribution check and a loss.

The ROI on agentic AI for denial management is not theoretical. It is the highest-leverage investment most ACOs are not making.

The Bottom Line

The ACOs that will thrive under CMS's 2026 benchmark rules, WISeR prior auth requirements, and new quality measure expansions are the ones that treat revenue cycle as a core competency—not a back-office afterthought. Agentic AI is not a nice-to-have. It is the infrastructure layer that connects your clinical operations to your financial performance.

If your shared savings are leaking through denied claims, you do not have a billing problem. You have an architecture problem. And architecture problems require architecture solutions.

Frequently asked questions

Why do claim denials matter for ACO shared savings?

Denials cost money twice. Reworking a denied claim takes staff time, and claims that are never corrected become lost revenue for participating practices. For the ACO, denial patterns can also distort the expenditure and diagnosis data that feed benchmarks and risk adjustment, and they often point to upstream workflow gaps, such as missing prior authorizations or post-acute billing errors, that affect quality and cost as well.

What does agentic AI do in the revenue cycle?

It works through multi-step tasks instead of flagging single errors: checking a claim against payer rules, finding missing documentation, confirming coverage and authorization status, correcting or routing the claim before submission, and drafting appeals when denials occur. Well-run deployments keep billing staff in charge of exceptions and coding questions, and every automated change should be reviewable by the billing team.

How should an ACO start reducing claim denials?

Start with data. Pull at least 90 days of denials across the whole network and group them by reason code, payer, practice and service type. Then document the authorization, claim editing and appeal rules of your largest payers, add a validation step between practice systems and the clearinghouse, and feed every denial back into upstream workflows. Network-level analytics make the patterns visible.

How do the 2026 prior authorization changes affect ACO denials?

Two CMS changes matter. Under the Interoperability and Prior Authorization final rule, impacted payers must publicly report prior authorization metrics, with the first set due by March 31, 2026. The WISeR Model, running in Arizona, New Jersey, Ohio, Oklahoma, Texas and Washington, uses technology with clinician review to assess prior authorization for selected services in Original Medicare. Tracking every authorization removes a frequent cause of denials.

Source notes

CMS: Medicare Shared Savings Program Performance Year 2024 financial and quality results (fact sheet). https://www.cms.gov/files/document/fact-sheet-ssp-py24-financial-quality-results.pdf

KFF: Claims Denials and Appeals in ACA Marketplace Plans in 2024. https://www.kff.org/patient-consumer-protections/claims-denials-and-appeals-in-aca-marketplace-plans-in-2024/

CMS: Medicare Shared Savings Program Shared Savings and Losses, Assignment and Quality Performance Standard Methodology Specifications (April 2026, Version 14). https://www.cms.gov/files/document/medicare-shared-savings-program-shared-savings-losses-assignment-methodology-specifications-version.pdf-0

CMS: WISeR (Wasteful and Inappropriate Service Reduction) Model. https://www.cms.gov/priorities/innovation/innovation-models/wiser

CMS: CMS Interoperability and Prior Authorization Final Rule CMS-0057-F (fact sheet). https://www.cms.gov/newsroom/fact-sheets/cms-interoperability-prior-authorization-final-rule-cms-0057-f

About the author

Gautamdev Chowdary is co-founder and CTO of Zynix AI, where he leads engineering for the Zynix platform, its AI agents and ZynixLLM.