Omni vs Looker

Omni publishes a dated digest every week, each bundling a handful of unrelated changes. Across the current window the AI work is continuous rather than occasional: AI Hub and Markdown columns reached general availability in May, visualization annotations in June, semantic model generation in July, with AI Routines gaining Slack support and chat-based creation along the way. The non-AI half is connection and embedding plumbing — OAuth for database connections, GitHub App and HTTPS deploy-token authentication for dbt, AccessBoost for Apps, embed display and timezone controls.

Omni is putting AI underneath the modeling layer rather than beside the charts. Generating the semantic model is a different bet than generating a query: the semantic layer is where a BI tool encodes what its metrics mean, and automating it moves AI from answering questions to defining the vocabulary the answers use. The governance work is arriving in step — AI credit controls per embed entity group and per user, AI skills gated by required access grants, evals support — which is what a vendor builds when customers are embedding these features into products they resell.

Expect AI Routines to keep expanding their trigger surface after Slack and chat-based creation, and the credit controls to grow into fuller usage governance as embedded AI reaches more end users. The digest format means individually significant launches will keep arriving in the middle of a list of unrelated fixes.