Summary
- As currently deployed, AI in healthcare administrative processes is likely to achieve only some of its goals.
- When applied on top of flawed administrative workflows, data complexity, and incentive structures, AI exacerbates the underlying issues.
- Realizing the potential for AI to reduce administrative waste will require redesigning the processes on which the technology is being deployed.
Key Takeaways
In January 2026, PHTI convened senior leaders from health systems, health plans, technology developers, investment firms, and federal agencies to discuss how technology and policy can enable AI to reduce administrative costs, accelerate payment cycles, and promote appropriate high-value care.
Participants explored the following questions:
- How can AI simplify and reduce administrative processes by removing wasteful steps, while maintaining necessary oversight and accountability?
- To what extent has current adoption reduced administrative friction?
- What market-shaping actions can mitigate inflationary risk to ensure that AI adoption delivers net savings?
The workshop focused on two administrative use cases where AI adoption is accelerating: prior authorization and medical billing.
Key Takeaways
Prior Authorization
Prior authorization is intended to promote evidence-based care and control costs, but inconsistent requirements and operational inefficiencies make the process administratively burdensome for health plans, providers, and patients alike.
AI may reduce the cost for individual organizations to execute prior authorizations, but it has not reduced overall system-level costs.
AI may reduce the cost for individual organizations to execute prior authorizations, but it has not reduced overall system-level costs.
Both plans and providers are making substantial investments to deploy AI tools. On the provider side, for example, AI solutions can draft medical necessity justifications, complete submission forms, and generate appeals. On the health plan side, AI solutions triage requests by complexity and resolution pathways and provide decision support, among other capabilities.
These tools enable providers to submit more complete requests with less effort, and health plans to process higher volumes of submissions at a lower cost per decision. While insurers and vendors have reported that AI can expedite steps in the prior authorization process, there is no evidence yet that this translates to lower average cost per claim factoring in the cost of the AI solution. As such, participants raised concern that optimizing each side of the transaction risks making the overall process more activity-intensive, rather than more efficient.
Real-time prior authorization at the point of care is an emerging model, but current proofs of concept are narrow and not yet scalable.
Real-time prior authorization at the point of care is an emerging model, but current proofs of concept are narrow and not yet scalable.
While current AI tools automate individual steps within the existing prior authorization workflow, emerging solutions aim to complete the full authorization process during the patient visit. Real-time adjudication models aim to integrate clinical documentation, policy matching, and authorization determination at the point of care, compressing the prior authorization process from days into minutes.
Early pilots suggest real-time adjudication models could reduce initial denials and accelerate approvals of prior authorization requests. Self-reported pilot results of Optum Rx’s PreCheck Prior Authorization with Cleveland Clinic reported an 88% reduction in appeals and a 68% reduction in denials caused by missing information.
The gap between these proofs of concept and a production-ready, generalizable solution is substantial. Real-time adjudication requires reconciling fragmented medical policies across health plans; navigating proprietary medical necessity criteria; managing variation in clinical complexity; and building infrastructure for real-time, bidirectional data exchange.
Data standards and digitization of medical policies can reduce information asymmetry, but AI’s impact is limited by variation across medical policies.
Data standards and digitization of medical policies can reduce information asymmetry, but AI’s impact is limited by variation across medical policies.
Current policy efforts focus on standardizing and digitizing prior authorization data and infrastructure. Finalized in 2024, the CMS Interoperability and Prior Authorization Rule (CMS-0057-F) requires Medicare Advantage, Medicaid fee-for-service, Medicaid managed care, Children’s Health Insurance Program (CHIP), and federally facilitated Marketplace plans to implement a prior authorization application programming interface (API) by January 1, 2027.
Workshop participants noted that CMS-0057-F and health plan pledges primarily support real-time adjudication of straightforward prior authorizations or resolution of prior authorizations delayed by administrative misalignment, such as unclear documentation or data requirements.
While digitization of medical policies can help reduce friction driven by information asymmetry, it does not address the variation in medical necessity criteria across health plans that drive clinically contentious and complex cases. Each health plan defines medical necessity according to its own clinical guidelines, benefit design, and coverage rules—and digitizing the exchange of this information does not standardize or ensure transparency of its content. As a result, a request that is straightforward for one plan may be contentious for another.
AI is exposing and exacerbating fundamental issues within the underlying prior authorization process.
AI is exposing and exacerbating fundamental issues within the underlying prior authorization process.
The application of AI to prior authorization reveals deeper structural limitations that technology alone cannot resolve.
The discussion raised fundamental questions about what AI should solve for within the prior authorization workflows. Before investing in technology to make a process run faster, is there a need to reconsider the design of the overall process?
Participants suggested potential structural alternatives to prior authorization that may accomplish utilization management goals, such as prepayment review (i.e., monitoring appropriateness after care delivery) or offering providers a discounted payment rate to bypass upfront review requirements entirely. Both alternatives would represent a fundamental restructuring of health insurance business models—which would require significant policy and industry alignment to adopt and implement.
Key Takeaways
Medical Billing
Medical billing requires providers to translate clinical encounters into standardized codes for reimbursement, but the current process where documentation drives billing incentivizes providers to document as completely as possible to maximize payment.
Provider deployment of AI is increasing billing intensity and inflating medical spending.
Provider deployment of AI is increasing billing intensity and inflating medical spending.
AI scribes and automated coding tools are now standard across health systems. These tools deliver modest time savings for physicians and enable more complete capture of patient complexity.
One multihospital health system found that after deploying an AI scribe, improved documentation and coding led to a 5% increase in Level 5 encounters and a 7% increase in level 4 encounters for established patients, translating to an average revenue increase of $1,004 per provider per month.
Health plans are beginning to respond to AI-driven increases in billing intensity with across-the-board downcoding and other reimbursement reductions, but the impact of these cuts is not yet known.
Health plans are beginning to respond to AI-driven increases in billing intensity with across-the-board downcoding and other reimbursement reductions, but the impact of these cuts is not yet known.
Reported health plan responses include deploying algorithms to identify and automatically adjust outlier high-complexity E/M codes, reducing reimbursement for select modifiers, and comparing submitted E/M levels against peer providers. These approaches vary across payers, are not well documented, and their impact on total cost of care is not yet known.
Health plan responses are also generating pushback from providers and state legislators, and several health plans have paused, delayed, or rescinded policies following provider opposition and state regulatory scrutiny. These responses may also disproportionately harm providers that have not adopted AI documentation and coding tools, who are disproportionately smaller, more rural, critical access, and independent.
Reimbursement policy is the strongest lever to drive administrative efficiencies and system-level cost savings.
Reimbursement policy is the strongest lever to drive administrative efficiencies and system-level cost savings.
Current health plan responses are likely not sufficient to address AI-driven medical inflation; a coordinated policy response is needed. Determining the appropriate response requires careful analysis of how AI is impacting medical charges, patient outcomes and equity, and provider compensation.
Participants suggested potential near-term guardrails to detect and understand drivers of medical inflation, as well as payment reforms to recalibrate the price of medical services in response to deployment of AI. Participants generally supported adjusting the price of medical services to address medical inflation.
AI Applications in Healthcare
PHTI convened a three-part workshop series in Washington, D.C., bringing together senior leaders from health systems, health plans, technology developers, academia, investment firms, and federal agencies to explore pathways for responsible AI adoption across healthcare.