CMS Is Building an AI Engine for Medicare Prior Auth — Here's What Every Practice Needs to Know

The Centers for Medicare & Medicaid Services is quietly laying the groundwork for algorithmic prior authorization decisions that will touch 45 million+ Medicare beneficiaries. Meanwhile, the FDA just installed a former AI company executive to lead its digital health oversight. These two moves, happening within the same week, signal a fundamental shift in how healthcare coverage decisions get made — and who (or what) makes them.

What Happened

CMS has launched what it calls an AI modernization project aimed at streamlining Medicare coverage determinations. The initiative targets the prior authorization process — the bureaucratic chokepoint that already delays care, burns staff hours, and costs the U.S. healthcare system an estimated $35 billion annually in administrative overhead. The project envisions machine learning models evaluating clinical documentation against coverage criteria in near real-time, replacing a patchwork of manual reviews, fax-based workflows, and inconsistent human decision-making. CMS frames it as efficiency. Providers should read it as a seismic change in the rules of engagement.

Simultaneously, the FDA appointed Rick Abramson — former Chief Medical Officer at Harrison.ai, an AI-driven radiology and pathology company — as the new director of its Digital Health Center of Excellence. Abramson's appointment is not symbolic. It puts someone with deep commercial AI experience at the helm of the agency that greenlights AI-powered clinical tools. The FDA has already cleared over 900 AI-enabled medical devices, and that pace is about to accelerate under leadership that understands the technology from the builder's side, not just the regulator's.

Taken together, these developments form a clear picture: the federal government is moving aggressively toward AI-mediated healthcare decision-making, from the coverage determination layer (CMS) to the clinical tool approval layer (FDA). The infrastructure for algorithmic medicine is being built right now, and it will not wait for providers to catch up.

The Risks

Algorithmic denials will be harder to fight. Today, when a prior auth is denied, your billing team can call a medical director, present clinical context, and argue the case. An AI system doesn't take phone calls. If CMS deploys machine learning models trained on claims data and coverage criteria, denials could become faster, more consistent — and far more difficult to overturn. The appeal process, already adversarial, risks becoming a battle against a black box where the logic behind a denial is opaque, proprietary, or simply unexplainable in human terms. Medicare Advantage plans, which already deny prior authorizations at alarming rates (a 2022 OIG report found 13% of MA prior auth denials would have been covered under traditional Medicare), will have even more sophisticated tools to say no.

Revenue cycle disruption is coming. Practices that rely on predictable prior auth timelines — especially in specialties like orthopedics, cardiology, oncology, and oral surgery — need to prepare for a period of volatility. As CMS phases in algorithmic review, the rules about what documentation triggers approval versus denial will shift. Models trained on historical claims data will encode existing biases, and providers who don't understand what the algorithm is looking for will see approval rates drop. The practices that suffer most will be the ones still submitting narrative-heavy clinical notes when the AI is scanning for structured data points, CPT-diagnosis pairings, and quantitative clinical thresholds.

Smaller practices face a structural disadvantage. Large health systems and DSOs can invest in AI-powered revenue cycle tools, hire data analysts, and build internal teams to optimize documentation for algorithmic review. A five-provider dental practice or a solo specialty clinic cannot. The compliance burden of understanding, adapting to, and monitoring AI-driven coverage decisions will fall disproportionately on smaller operators — the same providers who already lose 15-20% of revenue to administrative friction. Without intervention, algorithmic prior auth could accelerate consolidation across every healthcare vertical.

The transparency problem is real. CMS has not yet clarified how much visibility providers will have into the algorithmic decision-making process. Will practices be told which data elements triggered a denial? Will the models be auditable? Will there be a public comment period before deployment? These are not theoretical concerns — they determine whether algorithmic prior auth becomes a tool for efficiency or an instrument of opaque cost control. The regulatory framework has not caught up to the technology, and providers will be operating in the gap.

The Opportunity

Here is the reality that forward-thinking operators need to internalize: AI-driven prior auth is not a question of if. The practices that win in this environment will be the ones that start optimizing now, before the models are fully deployed and the rules are locked in. That means investing in structured clinical documentation, adopting AI-assisted coding and billing tools, and building internal workflows that produce the kind of clean, data-rich submissions that algorithms reward. Early movers will see faster approvals, fewer denials, and lower administrative costs — a compounding advantage that widens over time.

This is also a strategic moment for payer contract negotiations. Medicare Advantage plans covering over 30 million enrollees will leverage AI to tighten medical necessity criteria. Providers who understand algorithmic review logic can negotiate contracts with explicit guardrails — denial rate caps, transparency requirements, mandatory human review thresholds. The leverage exists now, while plans are still rolling out these systems. In 18 months, the leverage shifts permanently to the payer side.

The FDA's new leadership posture toward AI devices also creates opportunity. As more AI-powered clinical and administrative tools gain clearance, practices that adopt early will have better data, better documentation, and better outcomes — all of which feed back into stronger prior auth performance. The practices treating AI as a workflow partner rather than a threat will outperform their peers by measurable margins within two to three years.

Action Items

1. Audit your prior authorization denial rate today. Pull 90 days of data. Identify which procedures, payers, and diagnosis codes are generating the most denials. This is your baseline — you cannot optimize what you have not measured.

2. Shift clinical documentation from narrative to structured. Train providers to include quantitative clinical thresholds, standardized terminology, and explicit medical necessity language tied to specific coverage criteria. Algorithms parse structured data; they struggle with prose. This single change can improve approval rates by 10-15%.

3. Evaluate AI-powered revenue cycle and prior auth tools now. Solutions from companies like Infinitus, Cohere Health, and Rhyme are already automating prior auth submission and tracking. Practices that integrate these tools before algorithmic review becomes standard will have a significant head start. Budget $200-500/month per provider — it pays for itself in reduced denials within the first quarter.

4. Renegotiate Medicare Advantage contracts with AI transparency clauses. Demand language that requires payers to disclose when AI is used in coverage determinations, to provide specific denial rationale beyond boilerplate, and to maintain human review for all algorithmically generated denials above a dollar threshold you define. If your current contracts are silent on AI, you are exposed.

5. Designate a compliance point person to monitor CMS rulemaking on AI. Subscribe to CMS Federal Register notices, join your specialty association's regulatory affairs committee, and track proposed rules related to the AI modernization project. The comment periods are where providers shape the final policy. Missing them means accepting whatever CMS deploys.

Bottom Line

The algorithm is coming for prior auth whether you are ready or not. The practices that treat this as a six-month preparation window — not a someday problem — will be the ones still standing when the dust settles.

Built by the team behind Praxis AI — intelligence for modern healthcare. Read more on our blog.