Post by Thoughtworks

AI governance is architectural. As organizations move from AI experimentation to deploying agents in production, governance must extend across the entire AI stack—from data and models to agents and the tools they use. At #DataAISummit, Thoughtworks' Shayan Mohanty joined Databricks' Tim Lortz to discuss a data-centric approach to AI governance, built on consistent controls, observability and accountability across AI systems. Key themes included: ✅ Extending governance from data to models, agents and tools ✅ Applying consistent identity-based controls across the AI ecosystem ✅ Improving visibility with end-to-end observability ✅ Managing risk, compliance and cost without slowing innovation This is also the thinking behind Thoughtworks' Agent/works™, which integrates with Unity Catalog to help organizations apply governance, policy controls and oversight consistently across their AI ecosystem. As Shayan put it: "Governed agents aren't slow agents—they're the only agents that scale." The future of AI governance isn't a collection of disconnected controls. It's about governing the runtime and the substrate as one system, enabling organizations to scale AI with confidence.