Informatica, part of Salesforce, is expanding its Microsoft integrations to bring governed business data into Copilot, Foundry and Fabric. The practical prize is more context for business users and AI agents: definitions, data quality, lineage and access controls, rather than another assistant left to guess what a spreadsheet means.
Salesforce Watch analysis
What happened
Announced at Microsoft Ignite on 17 November, the collaboration adds an Informatica integration for Microsoft Copilot and Teams, alongside expanded data-governance capabilities for Microsoft Foundry and Fabric. Salesforce says the Copilot integration is generally available. Informatica Headless MCP server integration with Foundry is also generally available, with expanded governance, identity and administration features.
The announcement also extends Fabric scanning to notebooks and data pipelines, giving teams more visibility into how data moves and changes. Informatica is adding a US East Coast Azure point of delivery, offering another deployment option for organisations with regional data-residency or compliance needs. The details are in Salesforce’s announcement.
Our top picks
- Copilot gets business context
Users can access business definitions, data quality, governance and lineage information through Copilot and Teams. - Foundry agents can use Informatica MCP services
Developers can connect agents to managed data capabilities through standardised interfaces, with expanded identity and administrative controls. - Agent activity gains governance and audit features
The new capabilities are intended to help organisations control and trace agents’ interactions with enterprise data services. - Fabric lineage extends into notebooks and pipelines
Scanning can show more of how data is transformed and orchestrated, not just where it is stored. - A new Azure delivery point adds regional choice
The US East Coast option gives customers another deployment location to consider for residency and compliance requirements.
Why it matters
An AI assistant can produce a fluent answer while using the wrong definition of “customer” or a dataset nobody should trust. Bringing catalogues, lineage, quality signals and permissions into the tools where people build agents or ask questions addresses that unglamorous, consequential gap. Better access to context does not guarantee better answers, but it gives organisations more useful controls than hoping the model intuits their data rules.
The changes also span business users, developers and data teams, rather than treating AI governance as a single switch. That makes the announcement relevant to organisations trying to move from experiments to agents connected to real business systems.
Our read
This is a substantial enterprise-AI integration, not a new model announcement. Its strongest case is practical: connect AI tools to governed information and make data movement easier to inspect. The capabilities are now available in some areas, but the announcement offers no independent performance results showing how much they improve answer quality or reduce risk. Buyers should test them against their own data, permissions and workflows, not just admire the acronym count.
What to watch
- How customers configure access controls for agents using Informatica’s MCP services.
- Whether Fabric’s expanded scanning produces useful end-to-end lineage across real workloads.
- What evidence customers share about answer quality and governance in production.
Discussion spark: For enterprise AI, which deserves priority: giving agents richer access to business data, or tightening the controls on what they can do with it?
Sources and evidence
Independent WittyWires tracker for public updates about Salesforce. Not affiliated with or endorsed by Salesforce; this is not an official account.