The United States faces a health care crisis that goes far beyond rising costs. Physician burnout is accelerating, access to care is eroding, particularly in primary care and rural or underserved communities, and administrative burden continues to reduce time spent with patients. Clinicians are expected to diagnose and treat diseases in addition to navigating an increasingly complex maze of documentation, billing rules and compliance requirements. This burden has become one of the leading drivers of workforce attrition across the system.
To address this, healthcare organizations are turning to artificial intelligence tools as a practical necessity. AI writers streamline clinical documentation, while AI-powered coding solutions translate notes into accurate billing codes in real time. These technologies allow clinicians to focus more on delivering care and less on paperwork, a result that almost all stakeholders agree is long overdue.
The operational benefits are clear. AI coding systems aren’t just about reducing administrative workloads; they can also significantly improve financial performance. Mercyhealth, for example, reported a Revenue increase of 5.1 percent after implementing an AI coding solution. Health systems using automated coding also see a significant reduction in claim denials, a problem that can be costly for large organizations. up to $5 million per year. At a time when hospitals operate on razor-thin margins, these efficiencies do not constitute marginal gains.
Yet as providers increasingly adopt AI to stabilize their operations, payers react with suspicion. Insurers have begun calling the use of automated coding “upcoding,” and executives at major companies including UnitedHealthcare and Centene have recently announced plans to deploy additional AI tools to counter what they describe as aggressive billing practices. The result is a new AI arms race across the revenue cycle that risks deepening distrust rather than solving the underlying problem.
The problem is structural. The American health insurance model relies on utilization management practices that deny, delay, or reduce claims as a means of controlling costs. Although insurance plays a vital role in society, its economic incentives are fundamentally mismatched with those of providers and patients. In response, clinicians and health systems have been forced to document and code with extraordinary precision simply to receive payment for care already provided. What could have been a simple process has become a system defined by complexity, opacity and constant rule changes.
In this environment, manual billing and coding is no longer realistic. The volume of documentation requirements, regulatory updates, and coding reviews has exceeded what even highly trained human coders can reasonably handle. AI is not a shortcut or a tool for revenue manipulation. It’s the only scalable way to navigate a billing ecosystem that has become too complex for human cognition alone. In the modern healthcare landscape, AI has become fundamental infrastructure.
This tension is exacerbated by the regulatory delay. Much of the U.S. reimbursement framework – largely shaped by the Centers for Medicare & Medicaid Services – was built for a manual, human-coded era. Yet these same rules now govern AI-assisted workflows, without updated guidelines for how automation should be evaluated, audited, or incentivized. Without modernization, policy risks penalizing efficiency rather than rewarding accuracy, leaving providers stuck between outdated compliance standards and operational reality.
As a doctor with direct experience, fears that AI-based coding could inflate bills misunderstand both the technology and the problem it aims to solve. Proper coding is not about embellishment. Automated systems ensure that services provided are entered correctly the first time, reducing the need for rework, calls and extended reimbursement cycles.
Several years ago, I became the chief medical officer of an assisted living facility for dementia patients, to fill gaps in care for residents who could no longer be seen by their primary care physicians. Although I enjoyed the experience of caring for patients in their skilled care environment, I struggled to understand coding for care provided outside of my office. After nine months of endless delays and rejections, I also realized that I was making about 25 cents on the dollar per patient visit. So, I “burned out” before a year had passed and quit, entirely because of the burden of coding.
Expecting providers to accept underpayment for care provided would be like asking a grocery chain to allow consumers to leave their store having paid only part of what is in their cart. Delayed reimbursements and denied claims drive far higher costs into the system than accurate coding ever could – costs that ultimately pass on to patients through reduced services, longer wait times and, in some cases, facility closures. Just this year, 23 hospitals and emergency services have closed. From a policy perspective, the accelerated closure of hospitals, particularly in rural and underserved areas, raises questions that go beyond individual tolls. Reimbursement delays, denials and administrative difficulties increasingly determine which communities retain access to care. This is not a sustainable model for anyone.
Importantly, modern AI coding platforms are highly auditable systems in which every billing decision can be traced back to specific clinical documentation. This transparency provides clear justification for claims, paving the way for greater accountability for both providers and payers.
Notably absent from this debate is scrutiny of how payers themselves deploy AI. Insurers are increasingly using automated systems to report, delay or deny claims at scale, often with far less transparency than provider-side tools. Regulating one set of algorithms while leaving the other uncontrolled only worsens asymmetry and distrust. A constructive way forward would focus on shared norms rather than mutual suspicion. Establishing clear guidelines for AI-assisted coding that define auditability requirements, documentation traceability, and acceptable use in payer and provider systems would replace escalation with a common accountability framework.
Payers and providers ultimately share the same stated goals: providing high-quality care, operating efficiently, and maintaining financial sustainability. Treating AI coding tools as ammunition in an ongoing battle undermines all three. Used correctly, these technologies offer the potential to simplify an over-engineered system, reduce friction, and refocus resources on patient care.
Ending the AI arms race will require a change in mindset. Progress depends on collaboration and recognizing that automation can be a shared tool for clarity, equity and sustainability. Without this reset, the system risks continuing down a path that exhausts clinicians, destabilizes institutions, and leaves patients with fewer recourses.