Research context: Forvis Mazars, Leveraging Automation, Analytics, & AI in the Revenue Cycle · October 2026

Automation Is Not an RCM Strategy. The Real Opportunity Is a Closed-Loop Revenue Cycle.

Healthcare organizations have spent years buying technology to work claims faster. The more important opportunity is building a revenue cycle that learns why claims fail, changes the workflow that caused the failure, and verifies whether the fix actually improved cash. Automation executes. Analytics finds patterns. AI can help interpret complexity and prioritize action. Without a closed loop between those capabilities, providers risk buying faster versions of the same broken process.

Forvis Mazars recently published “Leveraging Automation, Analytics, & AI in the Revenue Cycle,” bringing together three capabilities that are frequently marketed independently. The title highlights a real operational challenge: success depends less on whether an organization has access to advanced tools than on whether those tools change financial outcomes.

That is the RCAI lens on the topic. A sophisticated revenue cycle needs more than a collection of task-specific bots or dashboards. It needs a system that connects the clinical event, payer requirements, claim adjudication, staff action and final payment into one measurable process.

Start With the Denial Nobody Should Have Worked Twice

Consider a health system's authorization-denial queue. Staff spend hours pulling the payer response, checking eligibility and coverage criteria, locating clinical evidence, submitting an appeal and then waiting for payment. A workflow robot may eliminate clicks. A language model may draft the appeal. A denial dashboard may identify the payer causing the largest backlog.

All three can be useful. But if the originating authorization failure remains uncorrected, the provider will continue producing the same denials at scale.

That is where the operating model needs to change. Every resolved denial should tell the organization whether the cause was registration, payer rule interpretation, scheduling, clinical documentation, coding, claim formatting or a payer-processing issue. That answer should be tied to a specific upstream owner and tracked for recurrence.

Forvis Mazars' Denials Management Monitoring materials emphasize root-cause analytics and prevention. The firm reports that, in its experience, organizations often experience initial claim-denial rates of at least 10% and may lose 3% or more of annual net revenue to denial write-offs. These are Forvis Mazars' experience-based statements about its client environment, not universal industry benchmarks or proof of the benefit of any particular AI tool.

Three Technologies Are Solving Different Problems

Too many RCM transformation plans use “AI” as a catch-all label. It obscures which technology is actually needed and makes outcomes hard to measure.

• Automation: executes rules-based tasks such as pulling payer status, routing work, reconciling remittances, validating required fields and initiating approved follow-ups. Its value usually comes from fewer manual touches and consistent throughput.
• Analytics: surfaces variation across payer, service, provider, location and denial category. Its value comes from identifying root causes and prioritizing the problems that matter most financially.
• AI: can synthesize narrative evidence, summarize clinical or payer correspondence, classify ambiguous denials and suggest next steps. Its value depends on accuracy, evidence, governance and the ability to complete appropriate work in existing systems.

These capabilities are complements, not substitutes. A model can recommend a next action but cannot guarantee that a payer will accept it. An automated task can execute perfectly while relying on an outdated policy. A dashboard can surface trends without any process owner acting on them.

The Strategic Moat Is Payer-Specific Execution Data

The most useful intelligence for an RCM operation is often not a generalized AI model. It is the combination of what the payer contract requires, what the payer actually did, which steps were taken and what ultimately resulted in payment.

For example, knowing that a payer rejects a particular service for a recurring documentation reason can improve a work queue. Knowing which evidence successfully overturns those denials—and when it is better to prevent the claim from going out—can improve the underlying business.

That dataset becomes more valuable when it is standardized across multiple billing teams, hospitals or specialties. It also becomes harder to build than a single AI feature because it requires integrations, permissions, operational feedback and reliable definitions of the outcome.

Contract allowances, actual negotiated rates, adjudication responses and payer policy changes all matter. A denial recovered at less than the correct contracted amount is not necessarily a complete financial success. A claim posted as paid but later recouped is not equivalent to durable cash.

Use Revenue Integrity, Not Bot Utilization, as the Scorecard

RCM leaders need a measurement system that prevents inflated automation narratives. Counting tasks, agent executions or “identified opportunities” does not establish that the provider is better off.

Instead, RCAI would evaluate each use case against a defined baseline using metrics that are economically connected:

• Correct net reimbursement: dollars collected against expected contracted payment, adjusted for reversals and recoveries.
• Avoidable denial rate: true preventable denials by payer and root cause, not only the percentage appealed successfully.
• Cost to collect: all-in labor, vendor fees, licensing, integration and exception-handling expense.
• Time to cash: elapsed time from service to final paid adjudication, by comparable patient and claim cohorts.
• Human exception burden: the volume, severity and cost of tasks that automated systems fail or escalate.
• Sustained improvement: whether the underlying defect rate stays lower after policy, staffing and patient-mix changes.

Health systems should measure these outcomes by use case and compare them against appropriate baseline cohorts. A pilot can appear successful simply because it touches easier claims, operates at lower volume, or receives extra staff support that will not exist in production.

Where to Deploy First: Three High-Value Workflows

1. Prior authorization and eligibility. Start where missing payer requirements cause rework before the patient reaches the billing queue. Automation can verify and route; analytics can reveal recurring payer and service-line errors; AI can help staff interpret complex evidence or payer correspondence. Actions still need safeguards and payer-specific policy verification.

2. Denial intelligence and recovery. Connect the denial reason, supporting evidence, actions taken and final resolution, then push root-cause fixes to registration, clinical operations, coding or scheduling. The goal should be fewer denials entering A/R, not just more appeals sent.

3. Contract performance and underpayments. Compare remittances with the expected allowance for the actual payer, procedure, location and terms. Analytics can identify leakage; AI may help investigate discrepancies; controlled workflows can route disputes. Realizing the dollars requires accurate contract data and a way to distinguish payer error from billing error.

In a related initiative, the DFW Hospital Council and Forvis Mazars announced a November 5, 2026 session on prior authorization, focusing on workflow design, documentation and the use of automation, analytics and AI to reduce denials. That context reinforces why the technology question cannot be separated from the underlying process.

Why Most RCM AI Plans Break at Integration

The technical challenge is not only extracting data from an EHR. It is connecting provider operations with payer rules, clearinghouses, claims, remittance records, clinical documentation and contract terms—while protecting sensitive patient information and complying with applicable rules.

The organizational challenge may be larger. An A/R team can identify a scheduling problem but cannot independently change how appointments are authorized. A coding team may flag poor clinical documentation but cannot make the physician modify a note. An underpayment team can detect contract variances but still needs the payer contracting team to confirm terms and escalation paths.

A closed-loop RCM system therefore needs accountable process owners, a shared taxonomy of root causes, transparent audit trails, human approval thresholds and regular reviews of realized financial performance.

Without those elements, more software simply increases the number of systems showing the same operational problems.

The Vendor Implications

As AI functions become easier to integrate, vendors selling only a narrow productivity feature may face pricing pressure. Competitive advantage may move toward companies that own reliable payer and workflow intelligence, have access to verified payment outcomes and can deliver approved actions directly inside existing provider systems.

This does not mean every health system needs one giant RCM platform. Specialized tools may remain the best option when their financial return is clear and data can flow cleanly across the system. The danger is buying isolated point solutions that cannot learn from one another.

For revenue cycle services companies, the same shift is underway: basic claim handling is only part of the value proposition. The more defensible offer is specialized knowledge combined with technology that systematically improves outcomes across clients.

RCAI Take

Automation is the execution layer. Analytics is the intelligence layer. AI can be the interpretation and prioritization layer. The value comes when all three change the operating system that produces revenue.

Forvis Mazars' October discussion arrives at an important moment for the industry: hospital finance teams are being asked to modernize while payers, staffing constraints and reimbursement complexity keep adding pressure. The right response is not to automate every possible action. It is to find the highest-cost recurring failures and build governed feedback loops that eliminate them.

Revenue cycle leaders should be able to answer a simple question about every technology investment: What process changed, what financial outcome improved, and how do we know it will keep improving?

If the answer is only that the AI completed more tasks, the strategy is not finished.

Sources and editorial note: Forvis Mazars, Leveraging Automation, Analytics, & AI in the Revenue Cycle (October 2026); Forvis Mazars, Denials Management Monitoring; DFW Hospital Council/Forvis Mazars webinar announcement. The linked FORsights article was visible, but its full body text was not retrievable for review; accordingly, this is an independent RCAI analysis inspired by its published topic. No unverified claims are attributed to that article. Product statements and experience-based denial estimates are attributed to the separately accessible Forvis Mazars materials.