Mission & Margin
Analytics Teams Keep Hiring Analysts. That’s the Problem | Bradley Riley
Overview
A study Bradley Riley referenced gave clinicians across 56 practices access to a clinical GPT that was accurate 96% of the time. Adoption of the good recommendations was low. Adoption of the 4% errors was disproportionately high. For Riley, VP of Analytics, Medical Economics, Reporting, and PMO at Longevity Health, that finding captures the deeper challenge: it’s not just about the AI, it’s about human behavior around AI adoption. His team operates with that same practical lens. Rather than leading...
Show Notes
A study Bradley Riley referenced gave clinicians across 56 practices access to a clinical GPT that was accurate 96% of the time. Adoption of the good recommendations was low. Adoption of the 4% errors was disproportionately high. For Riley, VP of Analytics, Medical Economics, Reporting, and PMO at Longevity Health, that finding captures the deeper challenge: it’s not just about the AI, it’s about human behavior around AI adoption. His team operates with that same practical lens. Rather than leading exclusively with AI-forward products, they focus on what he describes as “boring AI,” using AI tools to build non-AI solutions faster. A peer’s quote shifted his thinking on where strategies actually break down: your AI strategy is only as good as your compliance and IT team’s risk appetite.
Riley also tells Trey about what he calls an industry problem, not unique to any one organization. Analytics leaders hire analytics people, who hire more analytics people, until you have a team of technically strong analysts with no understanding of how the business actually runs. His fix was creating a role from scratch called the clinical delivery lead, pulling someone out of clinical operations and embedding them inside the analytics team as a permanent translator. He’s since added a second one for the market growth team. Brad is direct about the results: the team still makes mistakes, just far less often than they would without that conduit in place.
Topics discussed:
- Embedding clinical operators inside analytics teams
- Internal member movement alerts that bypass 60-90 day claims lag
- Small AI wins over single large pilots risking obsolescence
- Using AI to build non-AI tools faster
- Matching AI strategy to compliance and IT risk appetite
- Disproportionate clinician adoption of bad AI recommendations over good ones
- Building a custom managed care RAG agent for team training
- Agentic desktop automation for manual tasks where no APIs exist