A peer-reviewed analysis published in the January 2026 issue of Health Affairs offers a sobering assessment of where healthcare AI is headed.
Stanford University authors reviewed 21 AI tools currently deployed across the utilization management landscape, spanning prior authorization, concurrent review, claims adjudication, and appeals. Their conclusion: rather than resolving longstanding tensions between payers and providers, AI risks “supercharging” the flaws already embedded in these processes. The result is what the authors call an AI arms race — payers accelerating denials with automated tools, providers countering with AI-drafted appeals, and patients caught in the administrative crossfire.
For hospital and health system leaders, this trajectory should prompt a hard question: Is the AI you’re deploying helping your organization — or escalating a conflict no one wins?
The Risks Are Closer Than You Think
The Stanford authors identified several failure modes that are already surfacing in real-world deployments. Automation bias — the tendency to over-rely on AI recommendations even when clinical judgment suggests otherwise — ranked among the most pressing concerns. So did anchoring effects, in which AI-generated case summaries unconsciously shape reviewer conclusions before a human ever engages with the clinical record.
Perhaps most telling, in assessments conducted at Stanford Health Care, some staff using generative AI tools were unable to explain how the tools worked or identify potential sources of bias or limitations — a finding that likely reflects broader industry patterns.
Opacity compounds the problem. The analysis notes that no comparative studies exist examining denial rates or wrongful denials in AI-assisted versus traditional review processes. Providers are operating in an environment where payer AI decision-making is largely a black box — and the data to challenge it doesn’t exist yet.
Concurrent Review: The Overlooked Opportunity
While prior authorization has dominated healthcare AI headlines — fueled in part by the CMS mandate requiring impacted payers to implement open APIs by January 2027 — the Stanford analysis draws attention to a stage of utilization management that receives far less scrutiny: concurrent review.
Only three of the 21 tools evaluated address concurrent review. That’s a significant gap, given what’s at stake. Medical necessity denials during concurrent review represent roughly $2.5 billion in annual losses for healthcare organizations — approximately $5 million per provider per year. Unlike prior authorization, which patients encounter when scheduling care, concurrent review happens while patients are actively hospitalized. Misalignment at this stage doesn’t just create administrative burden; it delays care and drives costly downstream appeals.
Collaboration as a Design Principle
The Stanford authors weren’t simply sounding the alarm. Embedded in their analysis is a clearer path forward — one built on transparency, meaningful human oversight, and what they describe as collaborative AI design.
Among all 21 tools evaluated, only two were identified as “collaborative platforms” — tools designed to align payers and providers around shared clinical data, rather than optimizing for one party’s advantage. The authors advocate for AI that “helps insurers approve requests more efficiently, improves communications with providers and patients, and conserves reviewers’ time for hard decisions.”
That last point matters for health system leaders. The goal isn’t automation for its own sake — it’s AI that elevates clinical judgment rather than replacing it. Platforms that provide shared visibility into decision-making, automate straightforward cases, and surface complex ones for appropriate human review offer a fundamentally different value proposition than tools built to win an administrative battle.
A Framework for Evaluation
As health systems assess their own AI strategies, the Stanford analysis offers a useful lens. The researchers recommend demanding transparency in how tools make recommendations, ensuring human review is substantive rather than perfunctory, investing in staff training around AI limitations, and establishing governance structures that monitor for underperformance and disparate impacts.
For organizations evaluating vendors, the questions to ask are pointed:
Does this tool work with your payer partners or against them?
Can you see how it reaches its conclusions?
Is there peer-reviewed evidence that it performs accurately across diverse patient populations?
The AI arms race is real. But it isn’t inevitable. The healthcare organizations that come out ahead will be those that choose collaboration over escalation — before the costs of the alternative become impossible to ignore.
Learn how one AI platform is closing this gap, built from the ground up to align providers and payers during concurrent review.
At the Becker's 11th Annual IT + Revenue Cycle Conference: The Future of AI & Digital Health, taking place September 14–17 in Chicago, healthcare executives and digital leaders from across the country will come together to explore how AI, interoperability, cybersecurity, and revenue cycle innovation are transforming care delivery, strengthening financial performance, and driving the next era of digital health. Apply for complimentary registration now.
