
Episode #42
Agentic MMM with Dr. Luca Fiaschi (PyMC Labs)
There's often a "human API" bottleneck when implementing marketing mix models in a business. What if AI agents could help? That's what we talked about with Dr. Luca Fiaschi, partner at PyMC Labs. Find out how agents can support the full MMM chain - data consolidation and validation, model building with priors and business knowledge, deployment and retraining pipelines, and interactive Q&A for scenario planning - shifting data scientists toward designing agent workflows and guardrails rather than manual iteration. Learn how Skills files can improve coding-agent correctness, token-cost and vendor lock-in concerns. And find out what PyMC Labs is cooking up with their open-source “Decision AI” stack (Decision Hub, Decision Lab, Decision Lens). Links from the show: Decision AI Discord Server discourse.pymc.io 00:00 Cold Open and Agents Joke 01:17 Meet Dr Luca Fiaschi 03:51 What Is Agentic MMM 06:11 Agents Across MMM Workflow 09:16 Do We Still Need Data Scientists 17:44 Skills vs Vanilla Coding Agents 25:27 Guardrails, Hard Checks and R-hat 28:50 Token Costs and Business Value 33:05 PyMC Labs Open Source Strategy 36:46 Decision AI Stack and Lens Demo 40:11 Agent to Agent Media Buying 45:34 Future Foundational MMM Models 49:31 Wrap Up Tips and Community

