Data teams face a critical inflection point: years of Modern Data Stack investment have created bloated, fragile architectures ill-suited for AI agents as the new primary data consumer. Six strategic shifts are proposed: (1) slim down the stack by consolidating onto cloud platform native features and deleting unused models; (2) adopt open table formats like Apache Iceberg for true storage/compute decoupling; (3) shift from shipping tables to owning data products with SLAs and clear ownership; (4) build a rich semantic 'context library' with documented business logic for AI agents to consume; (5) move from passive dashboards to automated feedback loops that trigger actions; and (6) evolve the data role toward governance, architecture, and business sense rather than SQL authorship. The overarching message is that data teams must prioritize business value and AI readiness over technical elegance.
Table of contents
1. Features as Products, No More: Putting the Stack on a Diet2. True Decoupling: Storage (and Data!) is Yours, Compute is Rented3. Stop Being a Service, Start Being a Product4. Foundations for Agents: The Context Library5. From “What Happened?” to “What Now?”6. The Evolving Data Persona: “Who Writes the SQL” Doesn’t MatterA final reality check: It’s all about the businessSort: