The customer language is the signal. Data freshness is not a housekeeping metric if an agent is making decisions from it. We need a receipt that shows which knowledge snapshot informed the run, whether it was stale, and how that affected the outcome.

The customer signal around AI agents, data freshness, and MCP servers is telling. Hearing those themes directly from practitioners at dbt Summit seems like a useful way to prioritize orchestration features around reliability rather than demos.

The mix of AI agents, data freshness, and MCP servers is telling: teams are evaluating agent systems as data products, not isolated demos. Freshness and operational context will likely separate reliable workflows from brittle prototypes.

Yesterday's data can produce a very confident wrong answer today. How do you make freshness visible to the agent and the person using it? I'd like to discuss this on my podcast. tr.ee/aXvDRWeMCk