The rapid proliferation of artificial intelligence within clinical environments promises to revolutionize patient outcomes and operational efficiency, yet a new report from the Peterson Health Technology Institute (PHTI) warns that the United States’ existing healthcare payment models are fundamentally ill-equipped to handle this transition. The comprehensive analysis, released following an intensive multi-sector workshop, argues that without a radical shift in how AI-enabled care is reimbursed, the healthcare system risks either stifling innovation through underfunding or inflating national health expenditures through misaligned incentives.
The PHTI report is the culmination of a high-level summit convened in May 2026, which brought together a diverse cohort of industry leaders, including executives from major health systems and insurance plans, technology developers, venture capital investors, academic researchers, and representatives from federal agencies. The primary objective of the gathering was to dissect the financial barriers preventing clinical AI from reaching its full potential. Clinical AI, as defined by the institute, encompasses a wide spectrum of tools—from assistive technologies that provide decision support for clinicians to autonomous systems capable of performing complex diagnostic and management tasks without direct human intervention.
The Structural Misalignment of Legacy Payment Models
The central thesis of the PHTI report is that today’s primary healthcare payment structures—fee-for-service (FFS), pay-for-performance (P4P), and capitation—were designed for a world of human-delivered care and are structurally incompatible with the scalable nature of software.
Under the traditional fee-for-service model, providers are reimbursed based on the volume of services, procedures, and visits performed. This model creates a direct link between human labor and revenue. However, AI breaks this link. Unlike a human physician whose capacity is limited by hours in a day, an AI algorithm can process thousands of data points, images, or patient records in seconds. If the healthcare system continues to apply FFS logic to AI, the report warns of a "volume-based explosion" in costs.
"Under fee-for-service, reimbursement increases with the volume of billable services, rather than the value created," the report states. "AI will enable healthcare organizations to deliver more services and generate more billable work at unprecedented scale. Layering this onto the existing fee-for-service payment chassis would allow reimbursement to grow far faster than the true cost of delivering that care."
While alternative models like pay-for-performance and capitation (fixed per-patient payments) offer some protection against volume-driven inflation, the PHTI analysis found that even these models lack the specific incentives required to encourage the high-upfront-cost adoption and long-term maintenance of AI tools. Capitated models, for instance, might discourage the adoption of expensive new AI technologies if the immediate financial return is not clearly defined, even if the long-term clinical benefit is significant.
Three Pillars for Future AI Reimbursement
To address these systemic flaws, the PHTI workshop participants established three foundational principles that should guide the development of future AI-specific payment models.
First, reimbursement must be "deflationary" and outcomes-based. In most sectors of the economy, technology serves to lower costs over time as it scales. In healthcare, the opposite has often been true, with new technologies frequently adding to the total cost of care. The PHTI argues that AI should only receive premium reimbursement when it demonstrably improves patient outcomes, lowers the overall cost of a care episode, or significantly expands access to care in underserved areas.
Second, payment models must be dynamic and evolve alongside the evidence. Unlike a surgical procedure, which remains relatively static once mastered, AI algorithms are iterative. They improve with more data, but their performance can also "drift" as clinical environments change. The report suggests that payment rates should initially be set high enough to incentivize early-stage innovation and the high costs of integration. However, these rates should be subject to regular, longitudinal reviews. As the marginal cost of running the software declines and real-world evidence of its effectiveness accumulates, the reimbursement rates should adjust accordingly.
Third, there is a recognized need for "value-linked" payments. Instead of paying for the "use" of an AI tool, payers should pay for the "result" the AI achieves. This shifts the focus from the process to the product, ensuring that the financial rewards are aligned with the patient’s health rather than the provider’s utilization of a specific software suite.
The Challenge of Autonomous Clinical AI
One of the most significant findings of the report is the distinction between assistive AI and autonomous AI. While assistive AI acts as a "co-pilot" for doctors, autonomous AI represents a paradigm shift where the software itself performs clinical tasks, such as diagnosing diabetic retinopathy from retinal scans or managing insulin levels in real-time.
The PHTI workshop participants concluded that autonomous clinical AI cannot be shoe-holed into existing billing codes with minor tweaks. Instead, it requires entirely new payment frameworks. These frameworks must answer difficult questions regarding professional liability, the definition of a "service," and the eligibility of non-human entities to "receive" payment.
"No single payment model will support AI adoption across all clinical settings and use cases," the report noted. The participants emphasized that the industry must develop context-specific models. For example, the payment structure for an autonomous diagnostic tool in a primary care setting should look vastly different from an AI-driven remote monitoring system used in chronic disease management.
Industry Context and the Economic Landscape
The release of the PHTI report comes at a critical juncture for the U.S. healthcare economy. National health spending is projected to reach nearly $7.2 trillion by 2031, accounting for roughly 20% of the Gross Domestic Product. At the same time, the healthcare industry is facing a massive labor shortage, with some estimates suggesting a deficit of over 100,000 physicians by the early 2030s.
AI is often cited as the primary solution to this labor gap, but the financial incentives are currently pointing in the wrong direction. For a hospital system to invest millions in an AI infrastructure, they need a clear path to Return on Investment (ROI). If the current billing codes do not recognize AI-assisted work, or if they reimburse it at a rate that doesn’t cover the technology’s licensing and maintenance costs, adoption will remain confined to wealthy academic medical centers.
Conversely, if the Centers for Medicare & Medicaid Services (CMS) and private insurers create "add-on" payments for AI—similar to the New Technology Add-on Payments (NTAP) used for medical devices—without strict outcome requirements, it could lead to a surge in unnecessary utilization. The PHTI report serves as a warning to regulators that the "business as usual" approach to medical coding will not suffice for the digital age.
Stakeholder Reactions and Potential Implications
While the report reflects a consensus from the May workshop, it also highlights the tensions between different stakeholders in the healthcare ecosystem.
Health plans (payers) are increasingly wary of "coding creep," where AI tools are used to find more billable diagnoses (upcoding) without necessarily changing the patient’s treatment plan. For payers, the PHTI’s call for deflationary and outcomes-based models is a welcome safeguard.
On the other side, technology developers and venture capitalists argue that overly stringent reimbursement requirements could kill innovation in its infancy. They point out that the research and development costs for clinical-grade AI are astronomical, often requiring years of clinical trials to gain FDA clearance. If the "dynamic" payment model cuts reimbursement rates too quickly, it may discourage investors from funding the next generation of life-saving algorithms.
For health systems and providers, the primary concern remains integration. Clinical AI is not a "plug-and-play" solution; it requires significant changes to clinical workflows and Electronic Health Record (EHR) systems. Hospital administrators are looking for payment models that recognize the "total cost of ownership," including the human labor required to oversee and maintain these digital tools.
Conclusion: A Decisive Moment for Healthcare Policy
The Peterson Health Technology Institute’s findings suggest that the United States is at a crossroads. The decisions made by policymakers and insurance executives over the next 24 to 36 months will dictate whether AI becomes a tool for sustainable healthcare transformation or a driver of further financial instability.
"The decisions we make today on how to pay for AI-enabled clinical care will shape not only the pace of clinical AI adoption, but the impact on our healthcare system for years to come," the report concludes.
As the industry moves forward, the focus will likely shift toward pilot programs that test these new, outcomes-based reimbursement models. The success of these pilots will depend on the ability of payers and providers to share data transparently and agree on what constitutes "value" in an increasingly automated clinical landscape. For now, the PHTI report stands as a definitive roadmap for the complex work of aligning the economics of healthcare with the possibilities of artificial intelligence.
