Work IQ: how it supplies context to Copilot and agents

Key insights

• Work IQ brings together Microsoft 365 content such as files, email, meeting transcripts, recordings, and Teams chats.
• The interview describes Work IQ as a data backend for Copilot, Copilot Studio, and Foundry, as well as a platform available through APIs.
• Work IQ context can include permissions and information protections attached to the underlying data.
• Fabric IQ, Web IQ, and Work IQ can be combined as tools for an agent, depending on the task.
• The interview describes search and retrieval APIs and action tools, including sending email or Teams messages, as publicly available and generally available at that time.

Work IQ organizes workplace data and related context so Microsoft Copilot and AI agents can search information and, through available tools, take actions. It matters to Microsoft 365 administrators, data and AI architects, and developers building agents that need to work with organizational information. Reza Rad (RADACAD) [MVP] discusses the technology with Microsoft executive Charles Lamanna in a Fabric Insider interview.

What's new

Work IQ is described in the interview as a data backend for Copilot, Copilot Studio, and Foundry, as well as a platform that developers can access through APIs. Its purpose is to prepare Microsoft workplace data as context that AI systems can use, rather than treating each request as a simple search for a document or a short answer.

The sources named in the discussion include files in OneDrive and SharePoint, email, meeting transcripts and recordings, Teams chats, and data from Graph connectors. Lamanna also mentions Dataverse and Power BI semantic models as data that can connect through Work IQ. The interview does not provide a complete list of supported connectors or explain the configuration required for each one.

Context matters because an AI model cannot use all of an organization’s data at once. The interview describes Work IQ as helping select and organize relevant information for a model’s limited context window. Too much irrelevant material or too little useful information can lead to poor answers, Lamanna says.

Who it is for and what you need

The discussion is relevant to Microsoft 365 administrators, IT teams, data and AI architects, BI leaders, and makers or developers creating agents. For these roles, Work IQ raises practical questions about which organizational data an agent can use, how that data is protected, and what actions the agent may perform.

The interview describes Work IQ as part of the backend already used by Copilot and says developers can also use its APIs. It does not identify a required Microsoft 365 license, tenant setting, region, or setup procedure. It also does not establish that every Work IQ capability is available to every tenant or user.

For rollout planning, the clearest advice in the discussion is to review data access and protection before connecting AI tools. Lamanna recommends investing early in access control lists (ACLs), information labels, and information protection. He argues that an AI assistant may expose an existing permission problem rather than create a new one.

How it works

Work IQ can supply context for retrieval and search, and it can provide a route for agents to perform actions. The interview describes agents using it to find relevant information and, for example, send email or create Teams messages. Because agents may operate at higher volume than an individual user, Lamanna says the platform also provides an API pathway for higher-volume actions.

The discussion outlines two ways to combine Microsoft’s “IQ” capabilities. An agent can have separate tools and choose among them, or a Work IQ API can orchestrate retrieval across systems. The first approach gives an agent distinct capabilities; the second can bring multiple sources into a retrieval process through one API call, as described in the interview.

1. Identify the task the agent needs to complete and the workplace or business data it needs. The discussion names Microsoft 365 content, Dataverse, and Power BI semantic models as possible sources.
2. Choose the relevant intelligence layer or tools. Work IQ is for workplace information and collaboration; the discussion gives Web IQ for web information and Fabric IQ for Fabric data such as semantic models.
3. Check the permissions and protections on the source data. The interview emphasizes that context should inherit protections, so users and agents should not see information they lack permission to access.
4. Decide whether the agent needs retrieval only or actions as well. The interview mentions search and retrieval APIs and action tools such as sending email or Teams messages, but does not provide implementation steps.

These are planning steps drawn from the discussion, not a deployment guide. The interview does not give API names, code samples, portal paths, or a detailed method for configuring an agent.

Limitations and open questions

The discussion presents Work IQ’s role and architecture at a high level. It does not show a setup walkthrough or verify specific API behavior, usage limits, licensing costs, or tenant-level controls. Those details should be checked against current Microsoft documentation before an organization begins implementation.

Lamanna describes search and retrieval APIs and action tools as publicly available and generally available at the time of the interview. He also says another set of APIs was expected around the Ignite timeframe, but the discussion does not confirm what was later released. This interview alone therefore cannot establish current availability or rollout status.

Security still depends on how the organization has configured access to its underlying data. The interview’s advice is to apply least privilege, keep permissions with the data, and avoid copying data into systems with different permissions. It does not detail auditing, approval workflows, or how to configure each protection for a Work IQ implementation.

Sources

• Reza Rad’s Work IQ interview with Charles Lamanna

Frequently asked questions

What does Work IQ do?

It prepares organizational data and context for Copilot and agents to use for search, retrieval, and actions. The interview names Microsoft 365 content, Dataverse, and Power BI semantic models among the data sources discussed.

Does Work IQ respect existing permissions?

The interview says permissions and information protections should flow into the context, and that an AI system should not expose data a user cannot access. It advises organizations to review ACLs and information protection before enabling AI use.

What license do I need for Work IQ?

The interview does not specify a required license or plan. It also does not provide usage-based pricing details, so those requirements remain unconfirmed by this source.

Are Work IQ APIs generally available?

Lamanna describes search and retrieval APIs and action tools as publicly available and generally available at the time of the interview. The discussion does not verify their current availability, licensing, or tenant coverage.