Reborrowing Agentic System Concepts to Reboot Your Data Team

Software systems and algorithms often borrow concepts from the physical world—think neural networks, genetic algorithms, or grey wolf optimization. The idea of this talk is the opposite: to get inspired by the principles of software, specifically agentic systems, to improve the functionality of a data team. It’s a thought experiment that blends analogy with real-world experience to imprint memorable images. For example:Goals and autonomy: Unlike typical LLM workflows, agents require a goal, access to tools, and permission to act. Similarly, a data professional needs to know the ultimate business objective and be empowered to decide how to deliver it. Instead of “build a dashboard for…”, an analytics engineer can deliver more impact when given the business goal: “The marketing team needs to understand which campaigns are driving high-retention users.”Communication Protocols: Multi-agent systems rely on defined communication protocols between agents. The analogy here is clear: physical communication protocols are essential for effective collaboration within cross-functional teams or between sub-teams within a data org.This talk is addressed both to managers/leaders running or building teams, and to individual contributors who deliver data products either independently or as part of an efficient ensemble.