Ai News
Ai News

Microsoft 365 Copilot Adds Python in Excel and a Unified Work IQ Agent API

Published Aug 25, 2026 Sources checked Aug 27, 2026

Microsoft's August 25 Copilot release-notes batch adds Python-assisted Excel editing, a unified Work IQ endpoint for agents, richer work grounding, image generation and model choice.

Microsoft's August Copilot batch is about deeper execution and richer work context

Microsoft's August 25, 2026 Microsoft 365 Copilot release notes cover changes released between August 11 and August 25. The batch is broader than a single feature launch: it pushes Copilot further into executable analysis, enterprise context, agent APIs, model selection and in-app media creation.

Microsoft also cautions that Copilot features use a gradual deployment model, rolling out to subsets of users before wider tenant availability. That means a feature appearing in the release notes does not guarantee that every eligible user sees it immediately.

Excel can now execute Python through Edit with Copilot

The most technically significant end-user change is Python support in Edit with Copilot in Excel on Windows, Mac and web.

Users can ask Copilot to perform tasks such as statistics, simulation, advanced visualization, cleaning and transformation. Copilot can execute Python and place results back into the workbook. Microsoft says existing security and execution controls continue to govern how code runs.

This is more consequential than simply generating a formula. Python gives Excel users access to a much larger analytical programming ecosystem while keeping the interaction grounded in workbook data and outputs.

The practical workflow is still reviewable: describe the analysis, inspect the generated work and verify the output. That review step matters because natural-language intent can be ambiguous and generated analysis can still encode bad assumptions.

Work IQ gets a unified endpoint for agents and workflows

For developers, Microsoft added a Work IQ unified REST endpoint for invoking agents and workflows.

Microsoft positions the endpoint as a consistent, scalable alternative to multiple legacy Copilot Chat endpoints. Standardizing invocation can reduce integration complexity for organizations that are building custom agents around Microsoft 365 data and workflows.

The release notes do not mean every Microsoft 365 data source becomes freely accessible. Work IQ and Copilot continue to operate within the user's organizational permissions and product controls. Developers should still design identity, authorization and audit behavior explicitly rather than treating an agent endpoint as a bypass around Microsoft 365 security.

Work grounding expands into private conversations, email and meetings

Several changes expand the material Copilot can use as context. Private Viva Engage communities and events can ground responses when the user already has permission to see that content. Outlook emails can open directly beside Copilot Chat and can also become references inside Copilot Notebooks. Teams meetings can be added as notebook references, including transcripts, notes, chats and shared content. Web and work chat are being unified with a dedicated Work IQ control that lets users turn access to work data on or off.

Together, these features make Copilot less dependent on manually uploaded files. They also raise the importance of source permissions, retention and information-governance policies because more day-to-day work context can participate in AI-assisted tasks.

Copilot also expands model and media controls

The same release batch adds user-facing model choice in Researcher, allowing users to select supported Researcher models and modes.

In Word, Microsoft lists Sonnet 5 in the models menu and says it uses a higher default reasoning level for drafting and edits. That is a product integration claim from Microsoft's release notes; it should not be interpreted as a universal benchmark showing one model is best for every writing task.

Copilot Cowork also gains image generation and editing using OpenAI's ChatGPT Images 2.0 model, bringing visual artifact creation into the Microsoft 365 collaboration experience.

Why this batch matters for enterprise AI

The common pattern is a shift from a standalone chat box toward a work execution layer.

Python in Excel lets Copilot perform programmatic analysis. Work IQ gives developers a more standardized route to invoke agents and workflows. Email, meetings and private community content expand permission-aware grounding. Model selectors expose more control over reasoning behavior, while Cowork can create visual artifacts without leaving the workspace.

For organizations, the key operational question is therefore not just whether Copilot can answer a prompt. It is which data it may ground on, which code it may execute, which agents it may invoke and how those actions are logged and governed.

Because Microsoft uses gradual rollout, administrators should validate feature availability in their own tenant and review licensing, platform and governance requirements before planning workflows around any one item in the August 25 batch.

Sources

This article is built from the source material below. Open the originals for full context and the latest updates.

More ways to save

Discover deals, coupons and free courses on our sister site.

Explore DealVorio
Save more with DealVorio: deals, coupons, free courses, apps and books