The AI reimbursement puzzle: Who pays, and how much?

AI is rapidly reshaping clinical care, but the payment infrastructure hasn't kept pace. Federal officials are weighing a proposal to reimburse AI-driven diagnostic or clinical services at 60%–80% of the human rate — a move that experts say could significantly drive up healthcare spending while leaving physicians undercompensated. The core tension: Medicare ties payments to clinician time, but AI has an inverted cost structure — expensive to build, nearly free to run at scale.

Despite the FDA authorizing over 1,000 AI-enabled medical devices, only a handful have secured dedicated CPT codes. Currently, Medicare reimburses AI use in diabetic retinopathy screening and FFR-CT cardiac imaging, but the path from FDA clearance to routine reimbursement can take 7–10 years.

Key Takeaways:

• Experts propose three payment tiers: overhead tools (e.g., ambient scribes), add-on decision-support AI, and autonomous AI with its own per-use code
• CMS requires evidence that AI improves patient outcomes — not just detection rates — before granting coverage
• Adding AI to the budget-neutral physician fee schedule could inadvertently reduce physician payments

Why it matters: Getting AI reimbursement right is critical — overpay, and healthcare costs spiral; underpay, and innovation stalls. The answer likely isn't one number, but a nuanced framework tied to what the AI does, and whether it works with or instead of a clinician.