The integration of artificial intelligence into the administrative workflows of the American healthcare system has reached a critical inflection point, as industry leaders, clinicians, and patient advocates grapple with the ethical and operational boundaries of automated decision-making. At the recent HLTH conference, a premier gathering for healthcare innovation, a panel of experts convened to discuss one of the most contentious friction points in modern medicine: prior authorization. The consensus among the participants—representing major insurance associations, patient advocacy startups, and health policy consultancies—was clear: while AI offers transformative potential for efficiency, the final determination of medical necessity must remain the sole province of a licensed human clinician.
The debate arrives at a time when the healthcare industry is under intense scrutiny regarding the use of algorithms to manage care. Prior authorization, the process by which providers must obtain advance approval from health plans before delivering specific treatments or medications, has long been criticized for creating administrative burdens and delaying patient care. The introduction of AI into this process has the potential to either alleviate these delays through rapid processing or exacerbate them through opaque, algorithmic denials.
The Mandate for Human Oversight in Clinical Denials
Sean Robbins, Executive Vice President and Chief Operating Officer of the Blue Cross Blue Shield Association (BCBSA), provided a definitive stance on the limits of automation. During the panel discussion, Robbins emphasized that the ethical framework of insurance operations must prioritize human judgment in sensitive clinical scenarios. He argued that while AI is an invaluable tool for organizing data and streamlining the initial stages of a review, it lacks the nuanced understanding required to issue a medical necessity denial.
Robbins articulated a specific distinction between AI-assisted review and AI-driven decision-making. He stated that a licensed clinician should not merely "review" a recommendation generated by an algorithm but should be the primary architect of the final decision. This distinction is vital in the context of liability and medical ethics. If an AI system flags a request for further review, the subsequent evaluation by a human doctor must be a de novo assessment of the patient’s clinical needs, rather than a rubber-stamp of an automated suggestion. Robbins’ comments reflect a growing sentiment within the payer community that maintaining public trust requires a "human-in-the-loop" approach that is both transparent and accountable.
The Conflict Between Payer Clinicians and Treating Physicians
While the panel agreed on the necessity of human intervention, a significant point of contention emerged regarding which human should have the final word. Dr. Warris Bokhari, CEO of Claimable—a company dedicated to assisting patients in appealing denied insurance claims—offered a sharp critique of the current payer-led model. Bokhari argued that even when a human clinician at an insurance company makes a decision, they are often doing so in a vacuum, devoid of the personal and clinical context that only a treating physician possesses.
Bokhari’s perspective is particularly relevant in high-stakes fields such as oncology and transplant surgery. He contended that an "unnamed person" at an insurance company, regardless of their medical license, is effectively practicing medicine when they override the recommendations of a patient’s direct provider. This "arm’s-length" clinical review often fails to account for the physical presence of the patient, their diagnostic history, or the emotional and social nuances of their care. According to Bokhari, a coverage decision is, in the eyes of the patient, a clinical decision. When an insurer denies a life-saving treatment, the distinction between "coverage" and "necessity" becomes a semantic one that offers little comfort to those in need of care.
Accountability and the Use of Outdated Clinical Evidence
A central theme of the discussion was the need for accountability when insurers make incorrect or outdated clinical calls. Dr. Bokhari highlighted a specific allegation involving Blue Cross and Blue Shield of Illinois, claiming the insurer had denied liver transplants for cancer patients based on medical evidence dating back to 1994. He used this example to illustrate the dangers of static clinical guidelines in an era of rapid medical advancement.
The reference to 1994 served as a stark reminder of how far clinical medicine has progressed. In the mid-90s, many of the targeted therapies, immunotherapies, and advanced surgical techniques used today were either in their infancy or non-existent. Bokhari argued that using thirty-year-old data to justify current denials is not only a failure of the prior authorization system but a direct threat to patient safety. He urged insurers to stop "standing in the way" of physicians when time-sensitive, life-critical treatments are at stake.
In response, Robbins acknowledged that inappropriate denials are unacceptable in any context. He maintained that the industry’s goal should be to provide the right answers to the right people as quickly as possible, emphasizing that transparency and accountability are the tools needed to refine the system.
A Structural Revamp: Moving Beyond Speed
Dr. Mandy Cohen, a national advisor for Manatt Health and former Director of the Centers for Disease Control and Prevention (CDC), moved the conversation toward a broader systemic analysis. She argued that the industry should not focus solely on speeding up a flawed process but should instead rethink the entire prior authorization model.
Dr. Cohen pointed out that the friction between payers and providers is rooted in the underlying payment incentives of the American healthcare system. In a fee-for-service environment, providers are incentivized to increase the volume of services, which can lead to the ordering of "low-value care"—procedures or tests that do not significantly improve patient outcomes but do increase costs. Insurance companies implemented prior authorization as a "guardrail" to control these costs and ensure that care aligns with established evidence-based guidelines.
However, Cohen suggested that the current "cat-and-mouse" game between insurers and doctors is unsustainable. The "rethink" she proposed involves moving toward value-based care models where incentives are aligned toward outcomes rather than volume. In such a system, the need for intrusive prior authorization would naturally diminish as providers take on more financial risk and responsibility for the total cost and quality of care.
Data Trends and the Administrative Burden
The urgency of the HLTH panel discussion is supported by a wealth of industry data. According to a 2023 survey by the American Medical Association (AMA), 94% of physicians reported that prior authorization requirements lead to delays in care for patients with significant medical needs. Furthermore, 80% of physicians reported that these administrative hurdles can lead to patients abandoning their recommended course of treatment altogether.
The administrative cost of these processes is equally staggering. Research published in Health Affairs indicates that the U.S. healthcare system spends approximately $31 billion annually on the administrative tasks associated with prior authorization. For a typical primary care physician, these requirements can consume up to 13 hours per week of staff and physician time—time that could otherwise be spent on direct patient care.
The rise of AI in this sector was initially seen as a "silver bullet" to reduce these costs. However, recent legal challenges have highlighted the risks of over-reliance on technology. In 2023 and 2024, several major insurers faced class-action lawsuits alleging that they used AI algorithms to mass-deny claims without adequate human review. These cases have spurred regulatory bodies to take a closer look at how automation is deployed in utilization management.
Regulatory Responses and the Path Forward
The federal government has already begun to intervene. In early 2024, the Centers for Medicare & Medicaid Services (CMS) finalized a rule (CMS-0057-F) aimed at improving the prior authorization process. The rule requires Medicare Advantage organizations, Medicaid managed care plans, and other payers to implement electronic prior authorization processes and provide specific reasons for any denials. Crucially, the rule sets shorter timeframes for decisions, requiring payers to respond to urgent requests within 72 hours and standard requests within seven days.
While the CMS rule does not explicitly ban the use of AI, it reinforces the necessity of transparency. If an AI system is used to generate a denial, the insurer must be able to explain the clinical basis for that denial in a way that can be understood and contested by the provider and the patient.
The insights from the HLTH panel suggest that the future of prior authorization lies in a hybrid model. In this vision, AI serves as a powerful "co-pilot" for clinicians, handling the ingestion of thousands of pages of medical records, identifying relevant clinical data points, and highlighting where a request meets or fails to meet specific criteria. However, the "pilot"—the licensed human clinician—must remain at the controls for any decision that results in a "no."
Conclusion: Balancing Innovation and Ethics
The intersection of AI and prior authorization represents one of the most complex challenges in healthcare technology today. As Sean Robbins, Dr. Warris Bokhari, and Dr. Mandy Cohen highlighted, the solution is not as simple as faster algorithms or more clinicians. It requires a fundamental shift in how the industry views the relationship between cost control and clinical autonomy.
For AI to be successfully integrated into this space, it must be viewed as a tool for empowerment rather than a tool for restriction. By automating the mundane aspects of data collection and initial review, AI can actually give human clinicians more time to focus on complex, high-stakes cases. However, as the panel concluded, the moral and professional responsibility of making a life-altering medical necessity decision is a burden that technology cannot, and should not, carry alone. The human element remains the most vital component of the clinical equation, ensuring that medicine remains a practice of care, context, and compassion.
