Microsoft Is Retiring Its Data Governance Tool. Are Your AI Agents Ready?
When the catalog moves, your agents' context can quietly break
Most organizations treat their data catalog as a tool decision. Pick a vendor, load the metadata, and the governance lives there for good.
The larger problem is what happens when the tool changes. This week Microsoft said that Microsoft Fabric is now its data governance solution and that support for Purview Data Governance will end on November 12, 2027. Customers are being moved to the OneLake catalog, and Microsoft's own reasoning is pointed: "Agents need to know what data means, who can use it, and whether it can be trusted."
That turns a vendor roadmap into a project on your data team's desk. Every glossary term, lineage map, steward assignment and certified dataset has to be found, checked and carried across, while the agents reading that context keep running.
And it gets worse as agents spread. Genpact reports that 79% of enterprises now have AI agents in production. When a definition or an owner goes missing in a move, nobody gets an error. The agent simply answers with less.
As AI agents take on more work, every agent needs to know what it can rely on, who stands behind it and which other agents it can trust. That is what we are building at OpenMap: The Agent Network, the connective tissue for the world's AI agents, starting with every person and every local business. We are raising a £5M seed. Meet us at openmap.ai/investors.
Why Does a Catalog Change Break Things Nobody Can See?
Picture the certified "active customer" dataset your service agent uses to decide who gets priority support. In the old catalog it carried an endorsement, a named steward and a lineage map back to three source systems.
In the new catalog, the table arrives. The endorsement does not. The steward field is blank. Half the lineage stops at a staging layer.
The agent still finds the table.
It no longer knows whether to trust it.
This is not a tooling problem. It is a context portability problem.
The Cost Isn't the Migration. It's the Rediscovery.
Microsoft argues that governance belongs close to the data. Its announcement says the OneLake catalog can govern data across Fabric, Azure and third-party sources without copying or moving it, and that Fabric works with governance tools such as Atlan, Alation, Collibra and Informatica.
That is useful. But a new home for metadata does not tell you which metadata was right to begin with. Most catalogs hold a mix of current entries, stale ones and entries nobody has opened in years.
Before an agent-facing catalog can be trusted after a move, someone usually has to:
Inventory which glossary terms, endorsements and classifications actually exist today
Rebuild lineage, end to end, for the datasets agents read most
Confirm a named owner or steward for every certified asset
Resolve duplicate and conflicting records the old tool quietly tolerated
Retire entries that describe pipelines that no longer run
None of these tasks are particularly difficult. They are slow, they cross team lines, and the knowledge usually sits with a few senior engineers who are already stretched.
Teams in regulated industries have even less room. Microsoft's own advice to customers with contractual obligations and in highly regulated industries is to engage with their account team early.
How to Make Your Governance Context Survive Any Platform Change in 30 Days?
You do not need to redesign your data platform, or wait for a vendor deadline, to make your governance context portable.
Within the first month, focus on the assets your AI agents and executives touch most: customer, product, revenue and risk data.
Start by discovering what metadata you actually have, where each of those assets comes from, and which reports, models and agents depend on them. Then confirm an owner for each one and record which version is certified.
Next, expose this context through a knowledge layer your analysts and agents can read, whichever catalog sits underneath.
DataManagement.AI accelerates this process by automatically discovering metadata, mapping lineage across the modern data stack, identifying downstream dependencies, and creating a centralized knowledge layer that stays continuously updated as your environment evolves. It resolves duplicate and conflicting records, and ownership and policy travel with every entity, so a platform change does not strip the context your agents rely on.
Instead of spending months documenting assets manually before every platform change, your team maintains the context once and every catalog and agent inherits it. When the tooling shifts, the knowledge stays.
Catalog Migration Is Actually a Knowledge Ownership Problem
Platform strategies will keep changing. Vendors consolidate, rename and retire products, and every change lands on the same small data team.
The organizations that feel it least are not the ones with the best catalog. They are the ones that know what their data means without having to open one.
Organizations that scale AI agents successfully do not store their context inside a tool. They own it, and they let every tool read it.
Warm regards,
Shen and Team