A warehouse people and agents can trust without a human in the middle. Governed metrics, a modelling standard we can defend, and self-serve that does not go around the warehouse. That is what lets the team move from answering “what happened” to analysis that is actually useful.

What we're building in Q3-2026

Same field as the other squads. Capabilities, not tickets.

• A governed semantic layer that agents and people can query the same way
• Models that agents build and maintain by default, with humans holding judgement and standards
• The conditions for deeper experimentation, insight extraction and predictive analytics.