# How $40M RCM Firms Cut Claim Denials From 18% to 8% Canonical: https://granular.to/blog/rcm-cut-claim-denials-18-to-8 Published: 2026-05-31 Updated: 2026-05-31 Author: Trey Category: Playbook Tags: professional-services, automation, ai-agents, operations, playbook > A field-tested workflow for $40M revenue cycle management firms to cut initial claim denial rates by 50-60% through front-end eligibility, prior-auth automation, and denial-routing triage that does not require ripping out your clearinghouse or replacing your core platform. > **TL;DR.** Mid-market RCM firms running 12-18% initial denial rates are paying for the same workflow problem three times: a claim that should never have been submitted, the rework labor to fix it, and the 65% chance the denial gets written off entirely. The fastest path under 8% is not buying a new clearinghouse. It is reordering three steps you already do: front-end eligibility at the second layer, prior-auth detection before scheduling, and denial-routing triage by recoverable dollar value. Most firms can get there in 90 days without replacing their core platform. If your $40M RCM firm is running an initial denial rate north of 12%, you are losing two ways at once. You are paying $25 to $118 per claim to rework denials that should never have left your shop, and you are writing off the 65% of denials that never get resubmitted at all. The fix is operational, not technical. The firms that get under 8% changed the order in which their staff touches claims, not the platform underneath them. ## Why your denial rate is probably worse than you think The national initial denial rate hit 11.81% in 2024, up 2.4% year over year according to [Kodiak Solutions' analysis](https://www.techtarget.com/revcyclemanagement/news/366625109/Initial-claim-denial-rates-put-revenue-cycle-in-tough-spot) of 1,800 hospitals. Forty-one percent of providers told [Experian Health's 2025 State of Claims survey](https://www.experian.com/blogs/healthcare/state-of-claims-2025/) that more than 10% of their claims now get denied. That number was 30% three years ago. The HFMA MAP Keys target for initial denial rate is under 5%. Best-in-class is under 3%. If you are at 12% or 18%, you are not running an outlier shop. You are running a typical mid-market RCM operation that has been quietly bleeding margin while the industry got worse. Two reason codes cause most of the bleed. CO-16 (missing or incorrect information) is roughly 20% of denials. CO-197 (prior authorization absent or invalid) is another 20-25%. Together they account for about 40% of every denial that lands in your work queues. Both are front-end workflow failures, not coder errors and not payer policy changes. The financial weight is heavier than most operators realize. [Becker's Hospital Review put the all-in rework cost at $118 per claim](https://www.beckershospitalreview.com/finance/denial-rework-costs-providers-roughly-118-per-claim-4-takeaways/) for hospital settings and $25.20 for professional billing. Forty-eight percent of avoidable denials are non-recoverable according to Change Healthcare's Denials Index. Sixty-five percent of denied claims never get resubmitted at all per HFMA. On a $40M RCM book serving 200,000 claims a month at 15% denials, the math is roughly $700,000 a month in rework labor and another $1.2 million in writeoffs that should have been collected. ## Step one: stop relying on the basic 270/271 eligibility check Most RCM firms run eligibility through their clearinghouse's free 270/271 transaction layer. That tells you whether the patient has active coverage on the date of service. It does not tell you whether the specific service is covered, whether there is a deductible left to meet, whether prior auth is required, or whether the patient has a secondary policy you need to coordinate with. You need a second eligibility layer that runs before the appointment is finalized, not after the claim is submitted. The mid-market platforms that handle this credibly are pVerify, Experian eCare NEXT, Waystar's eligibility module, Inovalon Eligibility, and Infinx. Availity's Essentials tier is free and covers 2,600+ payers at the 270/271 layer, but it does not normalize benefits or estimate patient responsibility. The operational change is the sequence, not the tool. Today your registration staff probably runs eligibility at intake, captures the response, and moves on. The firms running under 8% denials run eligibility twice: once at scheduling, again 48 hours before the appointment. The second pass catches coverage terminations, plan switches, and benefit period resets that the first pass missed. It also flags claims that need prior auth before the patient walks in, instead of after the claim hits the payer's adjudication queue. This single change typically pulls CO-22 (coordination of benefits) and CO-29 (timely filing) denials down by 30-50% within 60 days. Those denials cluster heavily in the elective-procedure book, which is where your margin lives. ![Eligibility verification dashboard showing two-layer payer benefit verification workflow at a mid-market RCM firm](/images/blog/rcm-cut-claim-denials-18-to-8-eligibility.jpg) ## Step two: detect prior auth requirements before the scheduler picks up the phone Prior authorization denials, CARC code CO-197, are the most expensive type to absorb. The appeal success rate on PA denials runs 60-70%, which sounds good until you remember each appeal costs roughly $43 in labor and 14-30 days of delay before the payment hits. The right place to catch a PA requirement is at the moment the appointment gets scheduled, not at the moment the claim hits the clearinghouse. That means the PA-detection logic has to live inside the scheduling workflow, attached to the patient's payer and the CPT codes the clinician intends to bill. The post-Olive AI landscape has narrowed to four mid-market-credible options. [Cohere Health](https://www.coherehealth.com/) auto-approves up to 90% of submissions for participating plans and serves 660,000 providers. Rhyme integrates inside the EHR for 300+ payers and runs 4 million authorizations a year. Infinx claims 98%+ accuracy on PA detection and a 99.5% turnaround. Waystar's auth module is bundled with their broader RCM stack. The operational test is whether the PA system can answer "is auth required, and if so what is its status" inside the scheduling call, not in a separate workflow the back office reviews later. If your PA process still relies on a person logging into a payer portal after the appointment is booked, you are catching too many auth gaps too late. [Olive AI's collapse in 2023](https://www.fiercehealthcare.com/health-tech/once-high-flying-unicorn-olive-ai-sells-two-key-businesses-winds-down-operations) shook the prior-auth automation market and pushed many mid-market firms to either Cohere or Rhyme. Either is defensible for a $40M shop. Waystar's bundled offering makes more sense if you are already on their platform end to end. ## Step three: stop working low-yield denials by hand The third change is the one operators resist most. You have to write off denials your team has been chasing for years. The Adonis 2026 RCM Report found that 83% of organizations using AI-driven denial automation see at least a 10% reduction in denials within six months. The mechanism is not magic. It is denial scoring: every denial that hits the work queue gets scored by payer, dollar amount, claim age, and CARC type. Low-yield denials get written off immediately. High-yield denials get routed to the senior analysts who can actually move them. Half your team is currently working denials that have less than a 20% recovery probability. They are working them because that is what they have always done and the work feels productive. It is not. The labor cost of working a $35 denial through three appeal cycles exceeds the recovery value most of the time. The cleanest test is to pull your last 90 days of denial work, categorize by CARC code and payer, and calculate the labor hours per recovered dollar. You will find two clusters: high-yield work (typically CO-50 medical necessity, CO-197 prior auth, and any denial over $500) and low-yield work (typically CO-29 timely filing past appeal window, CO-11 dx mismatches on professional claims under $200, and any payer with a less than 30% appeal-overturn rate). The high-yield cluster needs your best people and AI-augmented appeal packet generation. [Waystar's AltitudeAI](https://www.waystar.com/news/waystar-advances-ai-leadership-with-next-generation-denial-prevention-and-reimbursement-recovery-innovations/) cut one mid-size health system's appeal package creation from 38 hours to 2 hours, and prevention work from 133 hours to under 6. Adonis (which raised a $40M Series B in March 2026) and Janus Health offer similar denial-prediction layers that integrate with your existing workflow without replacing the underlying clearinghouse. The low-yield cluster gets written off and removed from the work queue. ## What the named operators are showing The case studies in this space are mostly enterprise health systems, but the operational pattern translates straight to a mid-market RCM book. [BAYADA Home Health Care](https://www.waystar.com/insights-resources/resources-page/) cut its denial rate by 72% and recovered $2.5 million in six months running Waystar's denial management module. Their clean claim rate moved from the low 90s to 97% and average days to payer receipt dropped 51%. [St. Luke's Health System](https://www.experian.com/blogs/healthcare/case-study-how-st-lukes-health-system-cut-denials-by-76-with-enhanced-claim-status/) cut denials 76% using Experian Health's Enhanced Claim Status, dropped discharged-not-billed by $15 million a month, and saved the equivalent of three FTEs annually. Neither shop ripped out their core platform. Both layered denial-management workflow on top of what they already had. ## The 90-day timeline Most mid-market RCM firms can hit 8% in 90 days if the leadership team commits. The sequence we see in the field: - **Days 1-30**: Add the second eligibility layer at scheduling and 48 hours pre-service. Train registration staff on the dual-pass workflow. Expected denial-rate movement: 12-18% down to 10-13%. - **Days 31-60**: Roll out PA detection inside the scheduling workflow with either Cohere, Rhyme, or your incumbent's PA module. Audit your last 90 days of CO-197 denials to confirm the new workflow catches the patterns you are seeing. Expected denial-rate movement: 10-13% down to 8-10%. - **Days 61-90**: Score your existing denial work queue by recovery probability. Write off the bottom 30%. Redirect freed labor to high-yield denials with AI-augmented appeal generation. Expected denial-rate movement: holds at 8-10% but cash collected per labor hour roughly doubles. The post-[Change Healthcare cyberattack](https://www.hfma.org/technology/cybersecurity/cyberattack-on-change-healthcare-brings-turmoil-to-healthcare-operations-nationwide/) reality is that you probably need a backup clearinghouse anyway. Most $40M RCM firms now run two: Availity for connectivity plus Waystar or Inovalon for the analytics and AI layer. That diversification is independent of the denial-rate work, but it is the right time to do both. ## FAQ **Is 18% really a normal starting point for a $40M RCM firm?** Yes. Forty-one percent of providers in Experian Health's 2025 survey said more than 10% of their claims get denied, and the survey under-counts firms that have not invested in tracking. We see new client engagements walking in at 14-20% initial denial rates routinely. **Will this work without replacing our clearinghouse?** Almost always. The three changes (dual-pass eligibility, PA detection at scheduling, denial-routing triage) are workflow and add-on tooling, not platform replacements. Replacing a clearinghouse mid-contract is rarely worth the disruption unless you are still on a single-vendor Change Healthcare setup that has not been diversified. **How much of this needs AI versus just better process?** Steps one and two are mostly process and existing tooling. Step three is where AI earns its keep. The denial-prediction and appeal-generation layers from Waystar AltitudeAI, Adonis, or Janus Health are doing genuine work that humans cannot do at scale. **What does this cost?** A second eligibility layer runs $0.15-0.40 per transaction at mid-market volume. PA automation runs $2-5 per detected requirement at most plans. AI denial layers price by claim volume, typically $0.05-0.20 per claim depending on the vendor. On a 200,000-claim-per-month book, total added tooling cost is usually $30,000-60,000 per month, against $700,000-1,200,000 in current rework and writeoff cost. Payback is rarely longer than 90 days. **Where do you start if you only get to do one thing?** The dual-pass eligibility change at scheduling. It has the highest denial-prevention yield, costs the least, and does not require new platform integration. --- If your initial denial rate is north of 12% and you have been working from the same denial queue for the last three years, we can help. Granular builds the denial-routing layer, the PA-detection workflow, and the score-and-triage tooling that mid-market RCM firms need to get under 8% without replacing the core platform. Fixed price, four weeks, working tool. [Book 30 minutes with us](/) and we will walk through your current denial mix before the call ends. --- ## Keep Reading - **[Why Healthcare RCM Is the Next Mid-Market AI Roll-Up](/blog/healthcare-rcm-next-mid-market-ai-roll-up)**: Why PE is targeting mid-market RCM firms next and what that means for your strategic options. - **[How to Evaluate AI Vendors When You Don't Have a CTO](/blog/evaluate-ai-vendors-without-cto)**: A practical framework for comparing Waystar, Adonis, Janus, and the rest without a technical team to vet them.