Databricks announced Governance Hub early this morning: an account-level surface that puts data health, AI consumption and cost on one screen across AWS, Azure and GCP. It has three pillars — Data (tag coverage, ownership, classification, with drill-down to tables missing tags or documentation), AI (token consumption, model activity, per-user spend, guardrail coverage), and Cost (30-day spend against the prior period, highlighting untagged spend that escapes chargeback).
What the common framing gets wrong
Three corrections, and the second is the substantive one.
It is Beta. The announcement's tone is GA-shaped, and any write-up that says "Databricks launches governance control plane" without that word is describing a different product.
More importantly, the AI-spend visibility is not universal. It covers what flows through Unity AI Gateway. The post says it spans "Databricks-hosted and externally-hosted models", which reads as everything — but a model call that never routes through the gateway is invisible to the hub. That is precisely the shadow AI spend enterprises are buying this category to find: the team with its own API key and its own credit card. A governance tool that sees only governed traffic reports on the part of the estate that was never the problem.
Third, "intelligent" governance today means recommendations plus a Genie chat surface — the example query in the post is "Why has there been a spike in costs". The announcement itself places autonomous policy configuration and alerting in the future tense. Coverage calling this an AI governance agent is describing a roadmap.
The gap worth naming
There is no pricing, no tier requirement and no quantitative limit anywhere in the announcement. On a product whose stated purpose is controlling cost, publishing no cost is itself notable — and it fits a land-grab rather than a product launch.
Why the ledger is worth owning
Every enterprise AI programme now carries an unmeasured line item: per-user token spend across in-house and third-party models. Whoever owns that ledger owns the renegotiation with every model vendor. Putting AI spend next to data-quality debt and chargeback gaps is a bid to be it — aimed squarely at Snowflake's equivalent surface.