Microsoft 365 Copilot Wave 3 introduces advanced agentic AI capabilities for document creation and editing, alongside the debut of an E7 pricing tier. This new tier combines AI tools with M365 applications for $99 per user monthly. However, analysts caution businesses to be mindful of the limitations and potential risks associated with using Copilot Cowork.
Microsoft has integrated agentic AI capabilities into Microsoft 365 Copilot, enhancing its utility for professionals. This includes the rollout of its own version of Anthropic’s Claude Cowork.
Despite being available for over two years, Microsoft’s AI assistant has struggled to gain substantial traction among businesses, especially when compared to its other offerings. Microsoft reported last month that only 3% of its Microsoft 365 customer base has subscribed to the paid AI agent, amounting to 15 million paid licenses in total.
New features unveiled on Monday are designed to boost adoption. Notable among them is the debut of Copilot Cowork, which is powered by Anthropic’s Claude Cowork — an AI agent known for its autonomous task completion, which previously impacted SaaS vendor stock values.
Charles Lamanna, Microsoft’s president for business apps and agents, stated in a pre-recorded message that Copilot Cowork “integrates the ability to complete multiple long-running tasks concurrently within Microsoft 365 Copilot.”
Microsoft says its Copilot Cowork offers users “a new way of getting work done.”
Microsoft
Lamanna explained that Copilot Cowork extends Microsoft 365 Copilot’s capabilities beyond simple chat interactions, enabling the agent to perform tasks in the background and freeing users to concentrate on other duties. He added, “This empowers you to enhance your productivity and efficiency by working on multiple tasks across various Microsoft 365 applications simultaneously.”
J.P. Gownder, vice president and principal analyst at Forrester, noted that Copilot Cowork “capitalizes on the growing excitement surrounding Anthropic’s Claude Cowork concept.” He further commented that it “significantly expands its reach by embedding the functionality directly within Microsoft 365 applications, moving beyond a desktop-only approach.”
Gartner analysts, however, pointed out certain limitations when compared to Anthropic’s original tool in a research note. They stated, “Unlike Claude Cowork, [Copilot Cowork] does not support local computer usage, cannot interact directly with local files or applications, and lacks native integrations with third-party tools and services.” These omissions, they added, “restrict Cowork’s autonomy and its capacity to manage end-to-end workloads outside of Microsoft 365.”
Gownder also highlighted the inherent risks involved in deploying an agentic tool like Copilot Cowork in a business environment. He recalled that “Anthropic openly advises against using sensitive information with Claude Cowork at present.” He elaborated, “Theoretically, integrating [Copilot Cowork] into Microsoft 365 makes it cloud-enabled and scalable with enhanced data access… assuming it functions as intended, which is not guaranteed given Copilot’s performance history.”
Furthermore, questions persist regarding the tool’s ability to deliver on its promise of genuine value for business users.
“Even though Copilot is powered by OpenAI’s models, enterprise leaders inform me it consistently underperforms ChatGPT and ChatGPT Enterprise within the Copilot ecosystem,” Gownder remarked. He added, “Microsoft’s history of overstating Copilot’s capabilities means there’s a trust deficit, so Copilot + Cowork faces a considerable challenge in demonstrating its practical value.”
Copilot Cowork is currently accessible as a “research preview” through Microsoft’s early access Frontier Program, with no pricing details released yet by the company.
Microsoft’s collaboration with Anthropic on Copilot Cowork’s development signals a further shift away from its exclusive partnership with OpenAI. Microsoft has also made Claude an optional AI model for use within the core Copilot chat interface.
Additionally, Microsoft 365 Copilot is gaining new “agentic experiences.” These allow the AI to execute user-directed actions within documents, such as generating a pivot table or drafting presentation slides. Users can also issue commands via the Copilot chat interface, like instructing Copilot to write and then send an email without needing to switch applications.
Zoe Hawtof, senior technical advisor for Microsoft 365 Copilot, stated, “Our goal is to evolve beyond simple prompts and responses, enabling Copilot to truly perform the work for you.”
Microsoft confirmed that these agent capabilities are presently available in Word, Excel, and Copilot Chat, with plans for a rapid rollout to Outlook and PowerPoint.
Microsoft also announced the general availability date for Agent 365, its agent management and governance platform initially unveiled at Ignite last year. It will launch on May 1, priced at $15 per user per month.
Internally, Microsoft has already implemented Agent 365, using it to oversee 500,000 agents for various functions including research, sales, and HR self-service. These agents have collectively produced 65,000 responses daily over the past four weeks, according to Microsoft.
Furthermore, Microsoft revealed its long-anticipated E7 payment tier. Named the “Frontier Suite,” E7 combines all functionalities of Microsoft 365 E5—encompassing Entra Suite, Defender, Intune, and Purview—with Microsoft 365 Copilot and Agent 365, all offered as a single SKU.
Microsoft E7 “Frontier Suite” bundles all the features of Microsoft 365 E5 with Microsoft 365 Copilot and Agent 365 in a single SKU.
Microsoft
Launching May 1 at $99 per user per month, the E7 tier will be Microsoft’s premium offering for business customers, surpassing E5 in cost. Microsoft claims this bundle will result in cost savings for customers compared to purchasing each product separately.
There have been recent suggestions that the E7 tier would incorporate consumption-based pricing. Computerworld sought clarification from Microsoft regarding this, but did not receive a reply.
According to Jack Gold, principal analyst at J. Gold Associates, businesses are still in the preliminary stages of deploying AI agents.
Gold observed that “large-scale agent deployment is currently proceeding cautiously, as enterprises, much like with many emerging technologies, remain reserved.” He added, “Nonetheless, extensive experimentation is underway, often occurring in ‘shadow AI’ environments.”
Gold emphasized that governance and security are “crucial considerations” influencing the selection and implementation of AI agents. He noted, “We are in the nascent stages of agent adoption, and many enterprises are still determining the standards to establish and how IT departments will manage potentially widespread deployments.”
For Microsoft customers, a primary hurdle will simply be navigating the multitude of agent products available, many of which offer overlapping functionalities.
“Microsoft’s ecosystem is rapidly expanding with various agents and solutions, and discerning which one to use for a particular task is already proving confusing,” Gownder stated.
He cited other AI agent tools with similar capabilities, such as Researcher, launched last year, and Copilot Tasks, which was announced only weeks ago. Meanwhile, he observed that “users are developing Copilot Agents that are leading to agent sprawl, yet these apparently haven’t resolved user issues…, prompting the introduction of Cowork.”
“It’s a complex landscape,” Gownder concluded.
