Denial by Algorithm: Medicare's AI Prior Authorization Pilot Puts Care at Risk

Medicare's WISeR AI Pilot: What Insurance Teams Need To Know Now

The Trump administration will launch a Medicare pilot that uses an AI algorithm to help decide prior authorization for select services. The program, WISeR (Wasteful and Inappropriate Service Reduction), starts Jan. 1 and runs through 2031 in Arizona, Ohio, Oklahoma, New Jersey, Texas, and Washington.

Medicare has largely avoided prior authorization in traditional fee-for-service. This move borrows from private insurance, even as the administration publicly presses plans to reduce delays and denials elsewhere. Expect scrutiny from both parties in Congress, providers, and the press.

Scope: What's In, What's Out

• Services in scope (initial): skin and tissue substitutes, electrical nerve stimulator implants, knee arthroscopy. More may be added.
• Services out of scope: inpatient-only, emergency care, and anything that would pose substantial risk if delayed.
• Process: AI assists determinations; a "qualified human clinician" must review before any denial.
• Vendor incentives: vendors can share in savings, but are barred from compensation tied directly to denial rates, according to CMS.

Why This Matters For Insurance Teams

Public patience for delay-or-deny playbooks is thin. Lawmakers and hospital groups see shared-savings models as a potential incentive to reduce utilization, including cases that are medically necessary. That means higher regulatory, legal, and reputational risk if patient harm is tied to AI-driven decisions.

At the same time, leaders want measurable fraud and waste reduction without undermining access. The bar is simple: faster, cleaner decisions with defensible outcomes, transparent rules, and proof that care isn't inappropriately delayed.

Operational Mechanics To Prepare

• Human-in-the-loop clarity: Define what "meaningful human review" means in your workflow (time-on-case, documentation standards, clinical credentials).
• Tiered routing: Auto-approve low-risk, guideline-concordant cases; route medium risk to clinical review; escalate high-risk or high-cost to senior clinicians.
• Time guarantees: Set strict SLAs for initial and appeal decisions. Fast lanes for urgent requests.
• Exception handling: Codify rules for complex comorbidities and unique clinical contexts where algorithms are weakest.
• Transparent rationale: Provide clear, clinical criteria with every decision and an easy path for peer-to-peer review.

Governance And Compliance Essentials

• Policies: Document clinical criteria, model use, human review requirements, and escalation triggers.
• Model governance: Version control, validation plans, drift monitoring, and independent audits of fairness and performance.
• Vendor management: Contracts that ban denial-linked compensation, require audit access, log "who decided what and when," and enforce data privacy.
• Appeals oversight: Track overturn rates and time-to-overturn by condition, site of care, and population.
• Safety guardrails: Automatic bypass for cases flagged as risk of harm from delay; standing clinician authority to override.

Metrics That Will Make Or Break This Pilot

• Speed: Average time-to-decision, time-to-appeal decision, and peer-to-peer response time.
• Quality: Overturn rate (initial and appeal), rate of retroactive approvals, and clinician agreement rate.
• Access: Denial rates by service and population; monitor for patterns that suggest bias.
• Outcomes: Admission/ED visit rates within 30-60 days post-denial or delay; complications by service line.
• Provider experience: Peer-to-peer cycle time, documentation burden, and complaint volume.
• Member impact: Grievances, surprise bills tied to denials, and communication clarity scores.

Regulatory And Political Risk

Lawmakers across parties have flagged concerns about overreach if AI denies doctor-recommended care. A House measure seeks to block funding for the pilot in the FY 2026 HHS budget. Expect hearings, data requests, and pressure to define "meaningful human review."

Public sentiment is negative. Surveys show clinicians believe AI is increasing denials and patient harm in prior authorization flows. Algorithms perceived as "deny by default" will face immediate backlash.

Provider Relations: Keep Friction Low

• Clarity first: Publish criteria, medical policies, and decision rationales in plain language.
• Friction reduction: Auto-approve repeat approvals with stable clinical status; minimize re-auth cycles for chronic care.
• Fast escalation: Same-day peer-to-peer for urgent cases; named clinical contacts for high-volume providers.
• Closed-loop feedback: Let physicians flag model errors and trigger rapid policy updates.

Data You Need On Day 1

• High-quality claims + clinical context: Accurate diagnosis, prior treatments, contraindications, and recent outcomes.
• Service-level nuance: Distinguish low-value patterns from clinically justified outliers.
• Appeals corpus: Mine overturned denials to refine criteria and retrain models.

Communications Playbook

• Member letters: Clear next steps, deadlines, and how to reach a clinician reviewer.
• Provider notices: What changed, criteria sources, and escalation channels.
• Public stance: AI supports clinicians; it does not replace them. Share governance and audit commitments.

Key Quotes And Positions To Track

• CMS leadership said prior authorization delays "erode public trust," yet the pilot expands PA within Medicare to cut waste.
• Critics argue shared savings can still pressure utilization. Hospital and policy experts warn about incentives to deny high-cost care.
• CMS states vendors are prohibited from denial-linked pay, and every denial requires review by a qualified human clinician.

Action Checklist For Insurance Leaders

• Stand up an AI governance committee with clinical authority and legal oversight.
• Define meaningful human review standards and audit them weekly.
• Implement tiered risk routing and safety bypass rules for cases where delay could harm patients.
• Publish medical policies and update them on a predictable cadence.
• Instrument metrics: speed, overturns, outcomes after denial, provider friction, and member impact.
• Pressure-test communications and appeal instructions with real physicians and members.
• Run pre-launch simulations using historical cases; compare model vs. clinician decisions and outcomes.

What To Watch Next

• Any additions to the service list and expansion to more states.
• Congressional moves to restrict or reshape the pilot's funding and scope.
• Formal definitions of "meaningful human review" and required reporting to CMS.
• Litigation trends around AI-driven denials and the standard of care.

Helpful References

• CMS: Prior Authorization and Interoperability Final Rule
• AMA: Prior Authorization Research and Resources

Build Internal AI Literacy (Optional)

If your UM, claims, and compliance teams need fast, practical AI upskilling for policy, auditing, and vendor oversight, see these resources:

• AI courses by job

Bottom line: This pilot will test whether AI can trim waste without harming access. Insurers that prove speed, fairness, and clinical integrity will have the advantage. Those that can't will face audits, appeals, and headlines.