Not because of the AI model. Most failures come down to missing context.

Agents need two layers of context to answer questions accurately. The first is business context: how metrics are defined, what KPIs matter, what questions people actually ask, and what good answers look like. The second is a metadata foundation: the data that describes your data — tables, columns, relationships, and how the data was generated. This is part of the broader data foundation, but specifically the layer the agent needs to understand what it's looking at.

Without business context, agents write technically correct queries that answer the wrong question. Without the metadata foundation, they guess at schema and join the wrong tables. A common example: someone asks "who are our top 10 customers?" and the agent returns lifetime value rankings including customers who churned two years ago. Technically right. Completely useless. That's a context problem, not a model problem.