In this episode, we look at the actuarial principles that make models safer: parallel modeling, small data with provenance, and real-time human supervision.
To help us, long-time insurtech and startup advisor David Sandberg, FSA, MAAA, CERA, joins us to share more about his actuarial expertise in data management and AI.
We also challenge the hype around AI by reframing it as a prediction machine and putting human judgment at the beginning, middle, and end. By the end, you might think about “human-in-the-loop” in a whole new way.
What if the real power of AI isn’t to decide for us, but to help us decide better? David Sandberg, a well-known actuary, talks about a simple, important truth: AI systems are machines that make predictions; humans make decisions and take responsibility. That lens changes how we value models, govern risk, and build durable systems in insurance and beyond.
David takes us back to the 1990s valuation “wars” between real‑world and risk‑free frameworks and shows why running both models in parallel created more resilient decisions. We bring that mindset to today’s AI: pair perspectives, reconcile differences, and expose hidden risks before they bite. Along the way, we explore the ethics that actuaries live by—clear assumptions, stated limits, and consequences for noncompliance—and why data science needs equally enforceable standards to keep hype from outrunning safety.
We discuss small data versus big data with practical examples: sensors that flag electrical arcing and water leaks, curated corpora that trace every fact to source, and knowledge systems that prefer provenance over scale. You’ll hear why interpretable, traceable data enables actual‑to‑expected learning, combats drift, and survives regime shifts that break pure regression or LLM-only approaches. We also tackle agentic AI: when to contain complexity with validated, end‑to‑end systems, and how to embed agents inside strong guardrails, thresholds, and audit trails.
This conversation offers basic advice for leaders who want trustworthy AI: define risk appetite upfront, map the decisions that truly matter, keep humans at the beginning, middle, and end, and make validation a habit, not a one‑time phase. If you’re building in insuretech, managing enterprise risk, or pushing AI into production, you’ll leave with concrete ways to blend actuarial discipline with modern tools for safer, smarter outcomes.
For more information, see Actuarial and data science: Bridging the gap.
Connect with us to comment on your favorite topics:
