Microsoft integrates Anthropic tech into Copilot Cowork

Microsoft said in a blog post that it introduced Copilot Cowork in Microsoft 365 to let Copilot execute tasks across Outlook, Teams, Excel, and other apps.

Microsoft integrated the technology behind Anthropic’s Claude Cowork into Microsoft 365 Copilot, and that Copilot Cowork is being tested with a limited set of customers in Research Preview before broader availability in the Frontier program in late March 2026.

Cowork, powered by Work IQ, turns requests into plans by using signals from email, meetings, files, and data, and runs tasks in the background with checkpoints for approval, and checks in if it needs clarification.

It can triage calendars by flagging conflicts and proposing reschedules and focus blocks.

It can assemble meeting packets from emails, files, and calendar items, schedule prep time, and produce briefing notes, decks, and follow-up drafts such as a customer status email update that can attach the latest files.

It can gather web and work sources, including earnings reports, SEC filings, analyst commentary, and relevant news, and deliver an executive summary, a structured research memo, and a labeled Excel workbook with citations.

It can coordinate launches by creating competitive comparisons in Excel, value-proposition documents, customer pitch decks, and outlining milestones with owners and next steps.

Cowork runs inside Microsoft 365 security and governance boundaries, so identity, permissions, and compliance policies apply by default. It runs in a protected, sandboxed cloud environment, and actions and outputs are auditable.

🔗 Source: Microsoft

🧠 Food for thought

Implications, context, and why it matters.

Microsoft is selling a pricey fix for a problem tied to its own AI agents

New agent features ship with Agent 365, a management tool meant to handle security and governance issues from AI agents working across Microsoft 365 apps like Outlook, Teams, and Excel 1.
Microsoft research found 29% of AI agents in surveyed organizations run without IT or security approval, which raises risks the company calls “double agents” 1.
Agent 365 acts as a control plane (a centralized admin layer) and costs $15 per user per month. It extends Defender (Microsoft’s security suite) and Entra (Microsoft’s identity and access management service) to non-human entities 1.
Microsoft bundles this into Microsoft 365 Enterprise 7 (E7), a premium subscription that pairs Copilot with advanced security for $99 per user per month 1.

Microsoft tightens control over where enterprise AI runs

With Anthropic technology in the mix, Microsoft leans into a familiar play. It focuses on the platform where models get deployed, governed, and monetized, instead of betting on having the top model 2.
The same approach reduces the risk from standalone agents, which some investors worried could shake up established enterprise software providers 2.
This Copilot push follows Azure growth that fell short of investor expectations. Microsoft said it is shifting compute capacity inward to improve Copilot 2.
Agentic AI can cost a lot to run. One analysis of Anthropic’s original tool said a single Cowork session with complex file work may use quota equal to 50 to 100 standard messages 3.