The entry lists a November 2026 preview and January 2027 general availability, with development still underway. These are planned release windows—not evidence that the capability is available today. Microsoft explicitly describes its AI at Work roadmap dates as estimates and warns that the information can change.
What Microsoft is planning
The roadmap record identifies the following scope:
| Detail | Listed scope |
|---|---|
| Products | Work IQ and Microsoft Copilot (Microsoft 365) |
| Platform | Web |
| Development status | In development |
| Preview target | November 2026 |
| General availability target | January 2027 |
| Cloud instance | Worldwide (Standard Multi-Tenant) |
The central change is straightforward: an automated workflow would have a point where execution waits for a person’s response rather than continuing uninterrupted. Importantly, the description includes both input and approval. That leaves room for requesting information as well as asking permission, although the entry does not explain the interaction design.
The listed scope does not establish desktop or mobile support, sovereign-cloud availability, or licensing requirements for this particular feature.
Why a pause matters in Work IQ
Microsoft Learn describes Work IQ as an intelligence layer that lets agents reason over organizational data, context, and tools. Its documented tools can retrieve information and perform actions across Microsoft 365—not merely generate answers. That makes the distinction between preparing work and executing it particularly relevant.
Consider an illustrative workflow that prepares a customer communication, then waits for someone to review it before proceeding. Or imagine an agent encountering missing information and asking an employee to supply it. These are possible uses of pause-and-resume behavior, not confirmed scenarios for this roadmap item.
The practical appeal is selective automation: let software handle preparation while reserving a consequential decision for a person. The useful question is not simply “Can the agent finish?” but “Where should it stop?”
Approval is not the same as authorization
There is an important distinction for IT administrators. Microsoft’s existing Work IQ documentation says requests run in a specific user’s context and can access only what that user is permitted to see or do. It also says tool invocations are logged and evaluated for governance purposes. Those are platform-level controls, separate from the forthcoming human-input feature.
In practical terms, permission answers whether an action is allowed; approval asks whether a person wants that particular action to proceed. Neither should be treated as a substitute for the other.
The roadmap description does not establish whether approval will be mandatory, which agent types will support it, or how approval decisions will interact with existing logs and policies.
What administrators should look for next
Before relying on this capability in a production workflow, organizations should seek answers to several implementation questions:
- Placement: Who decides where an agent must pause?
- Routing: Who receives the request, and can it be reassigned?
- Failure handling: What happens after rejection, cancellation, or no response?
- Review context: Can the person see exactly what will happen next?
- Evidence: How are the decision and subsequent action recorded?
These are evaluation criteria, not announced features.
For now, the development to watch is a planned human checkpoint in Work IQ agent execution. Its eventual value will depend less on displaying an approval button than on making the decision clear, enforceable, and traceable. Until Microsoft documents those mechanics, this is a useful roadmap signal—not a ready-made guarantee of human oversight.