Human-Agent Collaboration in Data Science and AI

This 42-minute talk from the Data Science Festival explores the evolving relationship between humans and AI agents in data science workflows. Join Dr. Jason Hillary from Zerve as he examines how AI agents are transforming from simple accelerators into genuine collaborators. Discover real-world implementations and research findings that showcase both successful applications—including code generation, model debugging, and dynamic data exploration—and ongoing challenges such as agent misalignment, opaque reasoning, and context management limitations. Designed for technical practitioners, this presentation provides valuable insights for anyone interested in effectively building or working alongside AI agents as teammates rather than mere tools. The talk was featured at the Data Science Festival Game On event 2025, part of the UK's largest free-to-attend data community that brings together data professionals to share cutting-edge ideas and solve real-world problems.