John Bertrand on FDA Approval, CMS Reimbursement, and AI Healthcare Reality

Joe Grogan interviews John Bertrand, a veteran healthcare technology executive who led Digital Diagnostics to FDA approval and CMS reimbursement for an autonomous AI diagnostic device for diabetic retinopathy. Bertrand spent 13 years at Epic Systems before six years at Digital Diagnostics, and now works in supply chain automation. The episode is a candid post-mortem on bringing autonomous AI diagnostics to market. Bertrand reveals that regulatory and reimbursement frameworks—not clinical superiority—are the true gatekeepers for AI adoption. The FDA approval process, while well-intentioned, applies traditional medical device logic to autonomous AI, creating constant pivoting as guidance changes every 6-12 months. More damaging: CMS's initial reimbursement draft came in at $10 per test when the company needed $30+ to be viable, requiring four months of negotiation to shift the value argument. Bertrand emphasizes that workflow integration and financial incentives matter far more than being clinically better. He concludes that the first-mover disadvantage in regulated diagnostic AI is so severe he would not recommend it to new founders and has turned down multiple diagnostic AI opportunities since leaving Digital Diagnostics. The conversation pivots to where AI is actually gaining traction: patient-facing triage agents, administrative automation, and revenue cycle management—areas with far less regulatory friction. Bertrand explores the autonomous-versus-assistive AI regulatory gap, the nightmare of state licensing across 50 states, and the emerging concern of RCM automation becoming a zero-sum bot battle. The episode closes with a sobering assessment of AI healthcare reality versus hype, grounded in six years of hard-won experience navigating FDA and CMS processes. The show notes promise a 'tour de force on AI healthcare reality,' positioning this as insider wisdom for entrepreneurs and healthcare innovators.

Verdict: A sharp, insider-driven episode that delivers concrete lessons on regulatory and reimbursement barriers in healthcare AI, grounded in real experience; strong on substance and clarity, though balance could be deeper.

Government rubric

• Accuracy and sourcing 3/5 20%

  Guest credentials and company achievements (FDA approval, CMS reimbursement for Digital Diagnostics) are verifiable; specific claims (draft reimbursement at $10, needed $30+, four-month negotiation) are presented as first-hand account but l

• Balance and context 3/5 20%

  Episode presents Bertrand's perspective on regulatory friction and reimbursement barriers; acknowledges FDA's good intentions but heavily emphasizes the punitive aspects; limited indication of counterarguments or why regulators might defend

• Clarity on what it means 4/5 15%

  Bertrand clearly explains real-world consequences: reimbursement below cost kills adoption, regulatory uncertainty delays products, first-mover disadvantage punishes innovators; implications for founders and healthcare innovation are explic

• Substance 4/5 15%

  Dense with concrete examples (retinal imaging, $10 vs $30 reimbursement, six-month guidance changes), career narrative, and strategic insights; some tangential exploration of RCM automation and state licensing adds depth without major paddi

• Delivers on its title 5/5 15%

  Title promises FDA approval, CMS reimbursement, and AI healthcare reality; episode delivers detailed coverage of all three, with emphasis on the harsh realities and barriers that define the landscape.

• Value for time 4/5 15%

  48 minutes is well-paced with clear timestamps and topic progression; some exploration of tangential areas (state licensing, RCM bots) could be tighter, but overall earns its length with substantive insider perspective.

How we score Government

Worth it if you're an entrepreneur in healthcare AI, a healthcare innovation investor, or someone curious about why promising medical AI products struggle to reach patients despite clinical merit.

Based on the episode description. AI-generated; it can be wrong.