The Peterson Health Technology Institute (PHTI) released a new report on October 14, 2026, warning that the administrative AI tools US healthcare organizations are buying to cut costs are, under today's payment incentives, having the opposite effect. The report, Driving Adoption of High-Value AI in Healthcare, concludes that AI used for documentation, coding, billing and prior authorization is entrenching inefficient workflows and driving higher billing instead of eliminating waste.
The findings are based on a series of workshops PHTI held across 2025 and 2026 with leaders from health systems, health plans, technology developers, investment firms, academia and federal agencies, supplemented by expert interviews. PHTI is an independent evaluator of digital health technologies, and the report was prepared under the direction of its executive director, Caroline Pearson.
What the Report Found
The report's central argument is about incentives. Under fee-for-service payment, providers are paid for activity and volume. Administrative AI makes that activity faster and more complete, so organizations document and bill more of it. The technology works as designed. The payment model decides what it produces.
AI scribes are the report's clearest example. As Becker's Hospital Review reported, after one multihospital health system deployed an AI scribe across its ambulatory and emergency department settings, Level 5 encounters for established patients rose 5% and Level 4 encounters rose 7%, adding an average of $1,004 in monthly revenue per provider. The report cited a KLAS Research validation of the scribe vendor's results. Health plans, meanwhile, told PHTI they see coding intensity jump as soon as providers switch these tools on, and several plans are responding with their own algorithms to flag outliers.
Prior authorization shows a similar pattern. According to the report, insurers and vendors have not reported evidence that AI reduces the average cost per claim once the price of the tool is included. Faster submissions on one side and faster denials and appeals on the other are instead increasing the volume of back-and-forth between providers and payers.
At the same time, the report finds that the clinical AI applications with the most potential to expand access to care, such as hypertension management tools and mental health chatbots, remain stalled by payment models, liability rules and licensure frameworks that were built around human clinicians.
The Roadmap PHTI Proposes
The report organizes its recommendations into four areas that it says must advance together:
- Understand what works: generate real-world evidence on AI tools, monitor their ongoing performance, and assess clinical and economic outcomes so purchasers can trust what they are buying.
- Set rules for coverage and use: update licensure and scope-of-practice rules for AI-supported care, clarify medical liability and insurance, and define coverage consistently across payers.
- Purchase effectively: design payment models that reward outcomes rather than activity, and evaluate vendors against established outcome measures.
- Redesign care delivery: rebuild care models around clinical AI, evolve workforce roles, and remove administrative processes the report describes as adverse.
As a concrete example of the payment direction it favors, the report pointed to the Centers for Medicare and Medicaid Services' ACCESS model, a 10-year outcome-based program that launched in July 2026 with more than 160 participating organizations, as Becker's reported.
What This Means for AI Buying Decisions
Most clinics and hospitals now signing AI contracts are buying administrative tools first: scribes, coding assistants, billing automation and prior authorization software. This report is the strongest signal yet from an independent evaluator that those purchases should be judged on total cost and outcomes, not on documentation speed or captured revenue alone. If payers respond to AI-driven coding intensity with tougher audits and their own counter-algorithms, the short-term revenue gains the report describes could turn into denial and compliance costs.
For smaller practices, the practical exposure is simpler: a tool that speeds up a broken workflow makes the workflow more expensive to run, not less. The report gives practice leaders a framework, and independent backing, for asking vendors harder questions before renewing.
Analysis: How Clinics Should Read This Report
Analysis (Med Clinic X): This report is not an argument against AI in healthcare, and it should not be read as one. PHTI's position is that AI is genuinely different from earlier health IT because it can perform work that previously required human expertise, and that its clinical potential is real. The warning is narrower and more useful: value depends on what the tool is paid to produce. Front-office and administrative AI should be evaluated on measurable outcomes a clinic can verify itself, such as fewer missed calls, lower no-show rates, faster intake completion and staff hours returned to patient care, with the full cost of the tool counted. Contracts that tie vendor payment to those outcomes align with the direction PHTI recommends. Clinics reviewing how front-office automation is priced and measured can see Med Clinic X's approach here: https://medclinicx.com/solutions/ai-medical-receptionist
Sources and References
- Peterson Health Technology Institute, Driving Adoption of High-Value AI in Healthcare, October 14, 2026: https://www.phti.org/research-insight/driving-adoption-of-high-value-ai-in-healthcare/
- Becker's Hospital Review, AI tools drive up healthcare costs: Report, October 14, 2026: https://www.beckershospitalreview.com/healthcare-information-technology/digital-health/ai-tools-drive-up-healthcare-costs-report/
- HIT Consultant, PHTI Releases Landmark Framework to Scale High-Value Clinical AI Over Administrative Waste, October 14, 2026: https://hitconsultant.net/2026/10/14/phti-report-driving-adoption-high-value-ai-healthcare-caroline-pearson-roadmap-payment-reform/
Reporting based on the sources above, accessed October 15, 2026. Med Clinic X did not conduct original reporting or interviews for this article. Corrections: contact us via https://medclinicx.com/contact