Future capabilities: what is coming next in Microsoft 365 Copilot
Microsoft's Copilot and agent roadmap changes quickly. This page tracks capabilities that could become useful to awarding organisations, while keeping a clear distinction between what is available now, what is in preview and what is still planned.
Last reviewed 2026-09-07
Capabilities worth watching
| Capability | Current position | What it could enable |
|---|---|---|
| Interactive MCP Apps inside agent conversations | Preview planned from October 2026; rollout listed from November 2026 | An Agent could show richer interactive experiences such as sortable tables and drill-down views inside the conversation, rather than returning text only. For example, a centre-support Agent could present a filtered table of approved guidance or operational data without sending the user to another application. |
| Work IQ in Copilot Studio | Preview | Agents can use richer work context from Microsoft 365 and connected business systems. This could support internal helpers that understand relevant emails, meetings, files and business context rather than relying on a small static knowledge set. |
| Agents as workflow steps | Preview available; rollout listed from September 2026 | A workflow could call an Agent as one reasoning step, then pass the result to the next controlled process step. This could be useful where a normal flow handles the process but an AI step interprets an unstructured request, document or message. |
| Maker roles for shared Agent development | Preview listed from September 2026; rollout listed from October 2026 | Organisations could separate people who may view analytics/evaluations from people who may edit, configure, publish and share an Agent. This supports more mature team development and assurance. |
| Safer identity and sharing controls | Preview/rollout during 2026 | Copilot Studio can increasingly detect or block unsafe use of maker credentials and credential oversharing. This should make it easier to stop an Agent becoming an unintended route to somebody else's access. |
| Human approval for selected Agent tools | Rollout listed from September 2026 | Specific actions can be gated so a person approves them before they run. This makes action-taking Agents more practical for higher-impact processes where drafting or preparation can be automated but the consequential step remains human-controlled. |
How this could change awarding-organisation use cases
The important shift is not simply that Copilot will answer better questions. It is that Agents are becoming easier to connect to live work, structured processes and richer interfaces while Microsoft adds stronger controls around identity, approval and evaluation. That opens up use cases such as centre-support assistants, qualification-development helpers, controlled policy research, internal service desks, evidence-gathering workflows, project coordination and operational triage without requiring every process to become fully autonomous.
Microsoft sources
Related guidance
Sources
- AI at Work roadmap — Copilot Studio planned capabilities — checked 2026-09-07
- AI at Work roadmap — agent node and Copilot Studio roadmap items — checked 2026-09-07
- Work IQ in Microsoft Copilot Studio (preview) — checked 2026-09-07