Microsoft launches Copilot Cowork agents built on Anthropic’s AI and E7 AI suite to calm investor concerns over AI’s impact on SaaS

Microsoft has revealed a fresh lineup of products aimed at encouraging enterprise clients to develop AI agents on its platform. This includes Copilot Cowork—built atop Anthropic’s Claude Cowork AI product—and a new business productivity software bundle featuring Microsoft’s own AI offerings.

These new products, labeled by Microsoft as “Wave 3 of the Microsoft 365 Copilot journey,” arrive as the tech giant looks to fend off growing competition in the AI agent space. Rivals include business productivity firms like Salesforce, frontier AI companies such as OpenAI and Anthropic (which received billions in strategic Microsoft investments but are increasingly targeting the U.S. giant’s core customer base), and open-source options like OpenClaw.

Microsoft aims to ease investor worries that AI agents could reduce companies’ dependence on traditional software-as-a-service (SaaS) providers. The firm’s stock has dropped more than 14% since Anthropic launched Claude Cowork in mid-January.

At the center of the announcement is Copilot Cowork, a new feature developed in close partnership with Anthropic. Designed to handle long-running, multi-step tasks—such as prepping for a client meeting by creating a presentation, compiling financial data, emailing the team, and scheduling prep time—all from a single request.

“We truly believe this moment marks an inflection point,” Jared Spataro, Microsoft’s chief marketing officer for AI at Work, told . “For us, the inflection is Copilot adopting these agentic abilities—moving from providing assistance to actually taking action.”

Spataro noted Copilot Cowork leverages Anthropic’s Claude model for its reasoning and uses the same “agentic harness” (the system enabling the AI to use other tools and its operational guardrails) as Claude Cowork. However, Copilot Cowork includes features that simplify building the types of agents businesses need.

For example, Microsoft’s version operates in the cloud within a customer’s Microsoft 365 tenant—meaning it’s covered by the company’s enterprise data protection and integrated with “Work IQ,” a layer of intelligence drawn from a user’s emails, files, documents, meetings, and chats. In contrast, Anthropic’s Claude Cowork runs locally on a user’s device.

“We don’t run locally—and that’s a feature, not a flaw,” Spataro stated. He called Anthropic’s offering “a great tool” but noted it has “limitations” in corporate settings, citing lack of access to cloud-based enterprise data and security risks when deployed widely.

“Anthropic has demonstrated the value of these agentic capabilities and given us a practical look at what they could be,” he added. “Microsoft’s focus is on commercializing these technologies.”

The Copilot Cowork feature is currently being tested with a select group of customers and will be available as a research preview in March through Microsoft’s new Frontier Worker product suite.

The company also announced Anthropic’s Claude model is now accessible across the entire Copilot Chat experience—previously limited to Microsoft’s Researcher and Excel features.

While Microsoft initially built all Copilot offerings around OpenAI’s models, it has shifted to a flexible approach allowing customers to choose any model to power their AI assistants and agents. “At least every 60 days, there’s a new leading model,” Spataro said. “There’s strong demand for a platform that doesn’t require me to jump to another vendor.”

Microsoft revealed its Agent 365 product—described as a control plane or “orchestration platform” for AI agents—will be generally available starting May 1, priced at $15 per user per month. Agent 365 lets IT and security teams monitor, govern, and secure AI agents across an organization, including those built with other vendors’ software.

Spataro explained the key insight behind Agent 365 was recognizing that the same management infrastructure used for human employees—tools like Entra, Defender, Purview, and Intune—could be extended to manage AI agents. “AI agents are just as vulnerable to phishing attacks as people are,” he said. “As soon as an AI agent has an email address, it gets spam too—and it can respond to it.”

Microsoft reported that in just two months of preview, tens of millions of agents have been added to the Agent 365 registry. Internally, the company now has visibility into over 500,000 agents across its operations, with the most popular ones focused on research, coding, sales intelligence, customer triage, and HR self-service.

Finally, Microsoft announced the Microsoft 365 E7 Frontier Worker Suite, also launching May 1, priced at $99 per user per month. The bundle combines Microsoft 365 E5 (the company’s longstanding premium business productivity suite) with Microsoft 365 Copilot, Agent 365, the Microsoft Entra Suite, and advanced security capabilities from Defender, Intune, and Purview.

According to Microsoft, the $99 price is lower than what customers would pay if they bought these capabilities individually. The sum of component prices—E5 at $60, Entra Suite at $12, Copilot at $30, and Agent 365 at $15—totals $117 per user.

While many analysts have speculated AI agents will eventually push SaaS companies like Microsoft to move from per-user pricing to consumption-based models, Spataro said Microsoft’s customers aren’t demanding this change right now. “I think I have more data points than anyone in the industry—customers want per-user pricing right now,” he said. “That doesn’t mean it will always be the case, but it’s what they want currently.”

Microsoft shared that Copilot paid seats have grown by more than 160% year-over-year, with daily active usage increasing tenfold. The number of customers deploying Copilot at scale (more than 35,000 seats) has tripled year-over-year, and 90% of the Fortune 500 now use Copilot. The company also noted that 80% of the Fortune 500 are now using Microsoft AI agents in some capacity.