Skip to content
Lizely
Microsoft turns Office workflows into agent-driven projects

productivity · November 18, 2026

Microsoft turns Office workflows into agent-driven projects

What the sources reported

Office projects now span chat, documents, slides, and email

A question in an AI chat can become a Word draft, turn into a PowerPoint deck, and end as a decision in an email thread. That example shows Microsoft’s clearest change to everyday productivity: a project can move across multiple Office artifacts without requiring the user to recreate the work in each application.

For knowledge workers, this could reduce the handoffs that normally divide research, drafting, presentation, and approval. It also changes the practical role of an AI assistant. The assistant is no longer confined to generating text inside one interface; it becomes part of the path a project takes through Office.

Today becomes a central part of the Office AI experience

Microsoft is adding AI-powered capabilities to Office, including Today, as part of a broader effort to reshape how people work. The evidence does not specify Today’s individual functions or a release schedule, so users evaluating the feature should watch for details about what it surfaces and how it fits into existing projects.

The broader design is significant even without those granular details. Office is being presented as a coordinated work environment rather than a collection of separate applications. For teams, the practical question will be whether these connected capabilities can preserve context and produce reviewable artifacts across departmental boundaries.

AI agents become part of the Windows administration workload

IT teams are already managing more applications, device types, and workloads, and Microsoft is now adding AI agents to that environment. New Windows management capabilities are intended to automate more of the work involved in supporting these varied systems.

This extends the productivity shift beyond individual knowledge workers. Agents that can participate in application and device operations introduce additional questions about permissions, oversight, and failure handling. IT administrators will need to establish where an agent may act, how its changes can be reviewed, and how it fits with existing management processes.

Purview extends sensitivity labels to legacy Office files

Microsoft has updated the Purview auto-labeling workflow for sensitivity labels so that it supports files in legacy Microsoft Office formats, including DOC, XLS, and PPT. The change brings older document types into a more consistent labeling process.

This matters for organizations that retain substantial archives of legacy Office files or depend on them for active work. Teams should check whether older files receive the intended labels and whether the workflow covers every legacy format used in their environment. The evidence does not establish whether the update applies automatically to all tenants or requires configuration.

A connected project model changes what teams should test

The practical test for the connected Office workflow is whether a project can begin with a question and remain coherent as it becomes a Word draft, a PowerPoint deck, and an email decision. Reviewers should check whether context survives each transition and whether every generated artifact remains editable and accountable to a person.

Organizations introducing agents into Windows administration should set a separate review process for agent-driven changes. Security teams should also test how Purview’s expanded legacy-file coverage behaves against the document formats still present in their repositories. These checks connect the announcement to the controls practitioners can apply without assuming an unspecified release timetable.

Evidence

What this means for tooling

  • cross-application project tracker
  • agent permission and audit dashboard
  • legacy Office sensitivity-label scanner
  • Word-to-deck handoff checker
  • document-format conversion utility

Decision room queued — the team review of this signal has not started yet.

AI analysis by Lizely. Grounded in linked public evidence. Participants are fictional editorial roles, not real people or human authors.

More from other categories