The CFO's Guide to AI RCM Vendor Accountability

Why every health system CFO needs an independent diagnostic — and how to use one to protect your investment, strengthen vendor contracts, and close the accountability gap your AI RCM platform cannot close itself.

Published by Duet  ·  April 2026  ·  duet.health/ai-rcm-vendor-accountability

Every health system CFO reading this article has the same problem, whether they know it yet or not.

You have invested — or are about to invest — in AI-powered revenue cycle management. Waystar, R1 RCM, Optum, Ensemble Health Partners, AKASA, Adonis: the category is large, the marketing is confident, and the ROI claims are significant. Prevent denials. Recover revenue. Optimize the cycle. Every vendor in the space makes these promises. Every vendor sends you a dashboard to prove it.

But here is the question none of them will answer: how do you know if it's true?

Not from their dashboard. Not from their quarterly business review. Not from the performance report their account manager emails you before your renewal. From an independent source — one with no financial stake in the answer — using your data, your benchmarks, and a methodology you can audit.

That independent source does not exist in the AI RCM vendor ecosystem. It cannot exist there, because of a structural conflict of interest that is baked into every business model in the category. This guide explains what that conflict is, why it matters for your health system's financial performance, and what CFO-level AI RCM vendor accountability actually looks like in practice.

What AI RCM Vendor Accountability Actually Means

The phrase gets used loosely, so let's define it precisely. AI RCM vendor accountability is the ability of a health system's CFO to independently verify whether an AI revenue cycle management vendor is delivering the specific financial outcomes it promised — measured against an auditable methodology, not the vendor's own reporting.

That definition has three components, and each one matters:

Independently verify

The CFO — or a party acting on the CFO's behalf with no financial relationship to the vendor — produces the assessment. Not the vendor. Not the vendor's analytics partner. Not the vendor's quarterly business review slide deck. Independent means structurally independent: no contract, no referral fee, no ongoing commercial relationship that creates an incentive to shade the findings.

Specific financial outcomes it promised

Every AI RCM vendor makes specific promises during the sales process: denial rate reduction targets, AR days improvement, leakage recovery estimates, automation rates. AI RCM vendor accountability means measuring actual performance against those specific promises — not against a general benchmark the vendor selects, and not against the vendor's own prior-year comparison.

An auditable methodology

The CFO can read the calculation. Every input is disclosed. Every benchmark source is cited. Every weighting factor is explained. A methodology is auditable when a reasonably sophisticated CFO can follow the math from raw input data to final leakage estimate without taking anything on faith. Most AI RCM vendor dashboards do not meet this standard. The models are proprietary, the benchmarks are selected by the vendor, and the attribution logic is a black box.

The Structural Conflict at the Heart of the AI RCM Market

To understand why AI RCM vendor accountability requires an independent party, you need to understand the structural economics of the AI RCM market. Every major vendor in this space has a business model that creates a direct financial incentive to find, and report, the maximum possible leakage — because more leakage justifies more contract value.

The percentage-of-collections model

R1 RCM, Ensemble Health Partners, and most end-to-end RCM outsourcers charge a percentage of net revenue collected — typically 3–8%. At a $500M net revenue health system, that is $15–40M per year in vendor fees. The vendor earns more when collections go up. They earn more when they can attribute that improvement to their own interventions. They have a direct financial incentive to make the attribution as favorable to themselves as possible — and they measure that attribution with their own analytics tool.

The SaaS add-on model

Waystar, Adonis, and AKASA charge SaaS subscription fees, with more sophisticated AI capabilities sold as premium modules. These vendors earn more revenue as health systems purchase additional AI capabilities. Their renewal justification depends on CFOs believing those capabilities are delivering results. The vendor's own performance dashboard is the primary evidence for that belief.

The consulting-to-platform model

Optum and similar vendors with parent-company payer relationships create a layered conflict: the entity analyzing your denial patterns and recommending revenue recovery strategies has a financial relationship — through its parent company — with the payers who are denying your claims. This does not mean the analysis is dishonest. It means the structural independence required for genuine accountability does not exist.

The result of these structural incentives is a market in which CFOs are routinely making multi-million-dollar vendor renewal decisions based entirely on information the vendor produced about the vendor. The vendor selects the benchmark. The vendor runs the attribution model. The vendor writes the QBR. The vendor's account manager presents the results.

This is not fraud. It is a governance gap — and it is the same governance gap that created audit requirements for financial statements, independence requirements for external counsel, and conflict-of-interest policies for board members. The AI RCM market has simply not yet developed the independent accountability function that every other high-stakes financial relationship has.

Why 85% of Health Systems Have Not Achieved AI RCM ROI

The Black Book 2025 survey of over 11,550 healthcare finance professionals found that only 15% of health systems had achieved positive ROI from AI-driven RCM implementations. This number is striking — and it demands an explanation.

The most common explanations offered by vendors are implementation problems, insufficient change management, data quality issues, and insufficient adoption by clinical staff. These explanations are not wrong. They are also conveniently focused on the health system's own execution failures rather than on vendor performance.

A different explanation deserves equal consideration: CFOs cannot distinguish between AI RCM platforms that are delivering value and AI RCM platforms that are producing favorable-looking dashboards. Without an independent benchmark, the CFO cannot determine whether a 12% reduction in denial rate is the result of the AI platform, a change in payer behavior, a shift in the health system's case mix, or seasonal variation in claim volume.

The attribution problem

Revenue cycle performance is affected by dozens of variables simultaneously: payer mix changes, volume fluctuations, coding practice improvements, clinical documentation changes, staffing quality, and macroeconomic factors like Medicare Advantage enrollment growth. AI RCM vendors routinely attribute revenue improvements to their platform without controlling for these confounding variables.

A health system that implemented Waystar in a year when Medicare Advantage denials nationally fell by 8% will see its denial rate improve — and Waystar will show that improvement on their dashboard attributed to AltitudeAI. The CFO has no independent way to decompose how much of that improvement was Waystar, and how much was the broader payer environment.

The wrong buyer problem

The day-to-day vendor relationship for every major AI RCM platform is managed by the Revenue Cycle Director and VP of Revenue Cycle Operations — not the CFO. These are the same people who selected the vendor, championed the implementation, and are professionally invested in the vendor's success. They are structurally positioned to minimize bad news in vendor performance conversations and to accept vendor explanations for underperformance without escalating to the CFO.

The accountability conversation never reaches the person with the authority and the financial motive to demand honest answers. And the vendor's sales and account management teams know this — which is why every major AI RCM platform's QBR deck is designed for a Revenue Cycle Director audience, not a CFO audience.

What AI RCM Vendor Accountability Looks Like in Practice

Independent AI RCM vendor accountability is not a theoretical concept. It has a specific methodology, a specific output, and a specific use case for each type of health system CFO situation.

The accountability diagnostic  ·  what it measures

A rigorous AI RCM vendor accountability diagnostic measures four things against independent benchmarks:

1. Revenue leakage estimate — the gap between what the health system is collecting and what it should be collecting, calculated using the CFO's own financial data and benchmarked against weighted multi-source composites from HFMA, MGMA, Experian, and Kaufman Hall. Every input is disclosed. Every benchmark source is cited. The CFO can audit the calculation.
2. Vendor performance score — a structured assessment of whether the AI RCM vendor's performance on its specific contracted KPIs (denial rate, AR days, clean claim rate, automation rate) is meeting, exceeding, or missing the benchmarks that were used to justify the contract.
3. Execution fog radar — an assessment of the operational and governance risks that prevent AI RCM platforms from delivering their promised value: data quality, integration completeness, staff adoption, model training adequacy, and SLA enforcement. Most underperforming AI RCM deployments have identifiable execution gaps that neither the vendor nor the internal team has escalated.
4. Contract remediation map — the specific, CFO-ready evidence needed to renegotiate vendor contracts, invoke SLA remedies, demand performance credits, or make a vendor change with full financial documentation.

The three CFO situations where accountability diagnostics matter most

Not every CFO situation requires the same accountability response. The three highest-value use cases are:

Pre-renewal assessment. The most valuable window for an independent accountability diagnostic is 12–18 months before a major AI RCM contract renewal. This timing creates the maximum window to act on findings: renegotiate terms, demand SLA remedies, explore alternative vendors, or build the evidence base for a board-level vendor review. A CFO who commissions a diagnostic 30 days before renewal has findings but no leverage. A CFO who commissions one 18 months before renewal has both findings and options.

Post-implementation stall. Many health systems experience an initial improvement in denial rates after implementing an AI RCM platform — and then a plateau or reversal 18–24 months into the deployment. The vendor's explanation is typically implementation maturity, data quality, or market headwinds. An independent diagnostic can determine whether the stall is a genuine market factor, a vendor execution gap, or a model training problem — and produce the CFO-level documentation needed to act on the answer.

New vendor evaluation. A health system considering a new AI RCM vendor investment — or evaluating a move from one vendor to another — benefits from an independent baseline assessment of current leakage levels and vendor performance before any new contract is signed. Without a baseline, the incoming vendor's performance claims cannot be verified, and the outgoing vendor's performance history cannot be accurately assessed.

The AI RCM Vendor Accountability Audit  ·  12 Questions Every CFO Should Be Asking

The following questions are designed to be asked directly to your AI RCM vendor — in writing, before any renewal discussion, with the expectation of a written response. A vendor that cannot answer these questions in writing, with supporting data, has an accountability gap regardless of what their dashboard shows.

Five Principles of Effective AI RCM Vendor Governance

AI RCM vendor accountability is not a one-time event. It is an ongoing governance practice. The following five principles define what effective CFO-level AI RCM vendor governance looks like:

1.  Independent baseline before every major contract

Before signing or renewing any AI RCM contract valued at $500,000 or more annually, commission an independent baseline assessment of current revenue leakage levels and vendor performance. This creates the benchmark against which the new or renewed contract can be measured — and establishes the CFO's independent evidence base before the vendor relationship begins.

2.  CFO-to-CFO accountability structures

Require that every major AI RCM vendor relationship include a formal quarterly CFO-level review — not a Revenue Cycle Director QBR, but a CFO-to-executive conversation that includes independent performance data, not vendor-produced data. Many vendors will resist this. The resistance is itself informative.

3.  Auditable SLAs with real remedies

Negotiate AI RCM contracts with specific, measurable SLAs that include financial remedies — not just “remediation plans” — when performance targets are missed. An SLA without a financial remedy is a target without accountability. Require that performance data submitted against SLAs be drawn from the health system's own source systems, not the vendor's dashboard.

4.  Pre-renewal diagnostic, not post-renewal regret

The optimal window for an independent AI RCM vendor accountability diagnostic is 12–18 months before contract renewal. This is long enough to act on findings — renegotiate, invoke remedies, or plan a transition — but close enough to the renewal to have current, relevant performance data. A diagnostic commissioned after renewal has findings but no leverage.

5.  Separate the accountability function from the vendor relationship

The people managing the day-to-day AI RCM vendor relationship — the Revenue Cycle Director, VP of Revenue Cycle Operations, and their teams — have a professional interest in the vendor's success and a structural disincentive to escalate problems. The CFO should have an independent information channel: a diagnostic tool, an external advisor, or a formal audit process that is not filtered through the people managing the vendor relationship.

What Duet Is — And What It Is Not

Duet is an AI RCM Vendor Accountability Diagnostic. It is the only product in the market built specifically to give health system CFOs an independent, auditable assessment of whether their AI RCM vendor investment is delivering its promised financial outcomes.

Duet is not an AI RCM platform. Duet does not automate denial management, optimize prior auth, or process claims. Duet does not replace Waystar, R1, Optum, Ensemble, AKASA, or any other AI RCM vendor. The value proposition is categorically different.

What Duet produces

The Duet Comprehensive Leakage Report (CLR) delivers four outputs in a 17-day engagement:

• A revenue leakage estimate using the health system's own financial data, benchmarked against weighted multi-source composites from HFMA, MGMA, Optum, Experian, and Kaufman Hall — with every calculation disclosed and auditable by the CFO.
• A vendor performance score measuring the health system's AI RCM vendor against independent benchmarks — not the vendor's own reporting.
• An execution fog radar identifying the specific operational and governance risks that are preventing the AI RCM platform from delivering its full promised value.
• A contract remediation map providing the specific, CFO-ready evidence needed to renegotiate contracts, invoke SLA remedies, or prepare for a vendor transition.

Why Duet can do what AI RCM vendors cannot

Duet charges a fixed $25,000 fee for the CLR. There is no percentage of collections, no recurring license, no referral relationship with any AI RCM vendor, and no financial incentive to report a higher or lower leakage number. Duet earns the same fee regardless of what the Diagnostic finds.

This structural independence is not a feature — it is the entire value proposition. The accountability gap in the AI RCM market exists precisely because no other party in the ecosystem has both the structural independence and the CFO-level framing to produce an honest assessment. Duet is built to fill that gap.

The CFO's Action Plan

If you have read this far, you are likely in one of three situations:

1. You have an AI RCM vendor contract coming up for renewal in the next 12–24 months. This is the highest-value window for an independent accountability diagnostic. Commission the CLR now, before renewal discussions begin, while you still have full negotiating leverage.
2. You have an AI RCM platform that is not performing as expected, and your Revenue Cycle Director is managing the vendor relationship without escalating clear performance evidence to the CFO level. An independent diagnostic will produce the CFO-level documentation needed to have a different conversation — with your vendor and with your board.

3. You are evaluating a new AI RCM investment and want an independent baseline of current leakage levels before you sign. A Duet CLR establishes that baseline — so the incoming vendor's claims can be evaluated against an honest starting point, not the vendor's own pre-sale assessment.

In all three cases, the diagnostic costs $25,000 and takes 17 days. The leakage it is designed to identify — the gap between what your health system is collecting and what it should be collecting, with vendor performance as a variable — is typically measured in millions of dollars annually.

AI RCM vendor accountability is not a luxury governance practice. It is the basic financial oversight that a multi-million-dollar technology investment requires — and that the AI RCM market, by its structural design, cannot provide for itself.