.png)
By Kacie Geretz, Director of Growth Enablement
February 20, 2026
Payer audits are exposing black box AI gaps. Health systems that can't trace code assignments back to clinical documentation face costly compliance risks and delayed reimbursements in 2026's tightened regulatory environment.
Transparent AI creates instant, defensible audit trails. Explainable coding systems show the exact clinical evidence behind every code, eliminating time-consuming chart reconstruction when auditors question billing decisions.
Autonomous medical coding with built-in transparency delivers measurable results. Organizations using explainable AI see 25–50% reductions in code-related denials while strengthening physician trust and accelerating payment cycles.
Payer audits keep getting more sophisticated, and regulators keep asking harder questions about AI-driven coding decisions. In 2025 alone, total at-risk payer audit amounts increased 30% year-over-year, with outpatient coding-related denials rising 26% (1). Health systems are adopting AI to support medical coding and revenue cycle team members, but the tools they pick today will shape their compliance outcomes for years. The problem? Many don't realize they're using systems that can't explain how they arrive at code assignments. Unfortunately, that gap could cost them when auditors come calling.
Black box AI models have spread quickly across medical coding because these systems crunch through massive datasets to produce code assignments (2). The drawback is that they don't show their work. You submit a chart, you get codes back. However, you don't get the clinical evidence behind those choices, which means you can't defend them when payers ask questions.
It's the same as asking a coworker to code a tricky case. They give you the final codes but won't say where they found the supporting documentation. You probably trust their skills, but six months later, when an auditor questions one of those codes, you're stuck hunting through the chart with no idea what you're looking for.
That's the same risk here. Payers are questioning AI-generated codes more aggressively, and health systems that can't trace the path from documentation to code are struggling to defend their billing. In fact, one large nonprofit health system reported that underpayments or initial denials increased by more than 50% over a two-year period as payers adopted AI tools to review claims (3). The AI did its job and generated codes, but nobody can point to the specific clinical language that justified them.
Transparent AI (sometimes called explainable AI) takes a different approach. These systems don't just give you codes. They show you a clear path from the clinical documentation to each assigned code.
Here's how it works: When a transparent AI system assigns a code, you can see exactly which phrases, lab values, or clinical findings drove that decision. The technology identifies the precise evidence an auditor would need to see. Even better, it demonstrates actual clinical language understanding instead of just matching patterns.
This explainable AI reasoning is based on documentation, and that reasoning has to hold up when payers, regulators, and compliance teams dig into it. A transparent system allows health systems to focus more on patient care rather than defending coding decisions, since the evidence trail is already there. This AI model is all about accountability.
The audit world has changed. Payers use sophisticated analytics to flag questionable code assignments, conducting targeted data-driven reviews to validate that billed evaluation and management (E/M) levels align with service complexity (4). They want organizations to back up every charge with solid documentation. If your coding AI is a black box, you're facing three real problems.
First, audit responses eat up time. If an auditor questions a code from three months ago, your team must dig through the entire chart for what triggered the AI's decision. That takes hours of work from your skilled coders and slows everything down.
Second, your team can't learn. Coders who see AI suggestions without understanding why the system made those choices can't improve their skills. They're approving outputs without understanding the reasoning, which limits professional growth. That undercuts your ability to support medical coding and revenue cycle team members who need to grow professionally.
Third, you miss warning signs. If your AI keeps assigning codes based on shaky documentation, you won't know until the denials start rolling in. A transparent system flags documentation gaps right away, so you can fix issues before claims go out the door.
The American Health Information Management Association points out that coders' expertise sets them apart from AI models. Their clinical and regulatory understanding and professional judgement put them in a position to take on more responsibility—entering managerial roles (5).
In other words, AI should support, not sideline, human expertise in coding. Black box systems make that almost impossible because coders can't see what the AI is reading in the documentation.
While many autonomous medical coding solutions vary widely in their transparency and auditability, Nym's engine works differently. For every code it assigns, the system points to the specific clinical evidence that supports it. Every chart gets a complete audit trail.
Powered by Nym's proprietary Clinical Language Understanding (CLU) technology, the engine doesn't just find statistical correlations with certain codes. It identifies the actual phrases, values, and clinical findings that meet coding criteria. For example, the engine might highlight "patient presented with acute palpitations" and explain why that particular wording supports a specific cardiac code.
This brings greater accuracy and efficiency to healthcare revenue cycle management (RCM) without sacrificing the transparency auditors expect. When a payer questions a code, your team can pull up exactly what Nym found in the documentation during the original coding. Geisinger, for example, processes encounters in less than 2.5 seconds with over 96% accuracy using Nym's engine.
Nym also reduces administrative burdens by eliminating those time-consuming chart reviews during audits. Your compliance team doesn't have to reconstruct what happened; they can just show the evidence Nym tagged when the coding was done.
Transparency helps in ways that go beyond surviving audits. Physicians trust automation more when they can see the reasoning behind it. The American Medical Association reinforced this in 2025, adopting new policy calling for explainable clinical AI tools that provide clear reasoning physicians can access, interpret, and act on (6). A surgeon reviewing a post-op note can assess whether the AI caught the procedure's complexity by reviewing the specific documentation the system flagged.
That trust strengthens collaboration between clinical teams and revenue cycle staff. When physicians see the AI reading their documentation accurately instead of guessing, they start paying more attention to documentation quality. They can spot which phrases make code assignments stronger and which ones leave room for confusion.
Transparency also creates more predictable revenue cycle outcomes. Black box systems can shift behavior based on training data updates, suddenly coding cases differently without explanation. Transparent systems maintain consistent logic, which helps you forecast performance and track trends with confidence.
Organizations implementing explainable AI see substantial improvements. In the context of autonomous medical coding, the technology's accuracy contributes to a 25% to 50% reduction in code-related denials (7). This level of precision helps health systems demonstrate clear evidence for code assignments, reducing disputes with payers and accelerating payment cycles.
The question for healthcare leaders isn't whether to use AI for medical coding. Most have already decided that. The real question is whether to build revenue cycle operations on transparent systems or black box models that leave you exposed during audits.
In 2026's regulatory climate, transparency has become essential for sustainable operations and ethical AI adoption. Black box models might boost efficiency in the short term, but they create long-term instability. The compliance risks, eroded physician trust, and inability to understand or fix performance issues make them a growing liability.
Health systems choosing transparent AI can defend their billing, train their teams properly, and keep physicians and payers confident in their work. They're also showing commitment to ethical AI by keeping automated decisions understandable and accountable.
The technology that enables health systems to spend more time on patient care while meeting strict compliance standards is available now. How organizations handle the choice between transparency and opacity will determine who succeeds under tighter scrutiny and who struggles when auditors start asking questions.
Moving past black box models isn't just about audit preparation. It's about building revenue cycle operations that can adapt and hold up over time. Pick systems that show their work. Everyone benefits: auditors, staff, and patients.
Landi, H. (20 November 2025). Payer audits, denial amounts rise again in 2025, vendor data show. Fierce Healthcare. Retrieved February 19, 2026, from https://www.fiercehealthcare.com/finance/payer-audits-denial-amounts-rise-again-2025-vendor-data-show
Kosinski, M. (29 October 2024). What is black box artificial intelligence (AI)? IBM. Retrieved on January 28, 2026, from https://www.ibm.com/think/topics/black-box-ai
Kacik, A. (24 October 2024). 'AI arms race' underway as payers, providers jockey for upper hand in claims review. Healthcare Dive. Retrieved February 19, 2026, from https://www.healthcaredive.com/news/artificial-intelligence-claims-review-payers-providers/730872/
Minemyer, P. (16 December 2025). How payers are 'embracing AI's benefits' — and what's next: AHIP. Becker's Payer Issues. Retrieved February 19, 2026, from https://www.beckerspayer.com/payer/how-payers-are-embracing-ais-benefits-and-whats-next-ahip/
Weber, S. (28 March 2024). Success of Revenue Cycle AI Hinges on Health Information–Physician Partnerships. Journal of AHIMA. Retrieved February 5, 2025, from https://journal.ahima.org/page/success-of-revenue-cycle-ai-hinges-on-health-information-physician-partnerships
American Medical Association. (11 June 2025). AMA adopts new policy aimed at ensuring transparency in AI tools. Retrieved February 19, 2026, from https://www.ama-assn.org/press-center/ama-press-releases/ama-adopts-new-policy-aimed-ensuring-transparency-ai-tools
Nair, S. (16 January 2024). The Reality of Autonomous Coding. Health IT Answers. Retrieved February 5, 2026, from https://www.healthitanswers.net/the-reality-of-autonomous-coding/
Kacie Geretz, RHIA, CPMA, CPC, CCA is the Director of Growth Enablement at Nym, where she aligns Nym’s product roadmap with the evolving needs of health system partners and serves as the externally-facing expert on Nym’s autonomous medical coding engine. A graduate of The Ohio State University’s Health Information Management program, Kacie brings deep expertise across the revenue cycle—having led revenue integrity programs, built managed care contracting and credentialing infrastructure, and driven denials and A/R process improvement initiatives. She is passionate about advancing healthcare automation and regularly shares insights on coding innovation and RCM transformation.
