Microsoft explores bringing OpenClaw-style agents to 365 Copilot for the enterprise market

Microsoft is testing the integration of features inspired by the OpenClaw platform into its enterprise AI assistant, 365 Copilot. The Information reported this, citing sources inside the company. According to the journalists, the effort aims to let Copilot work autonomously around the clock, carrying out tasks on a user's behalf without constant manual control. Microsoft corporate vice president Omar Shahin confirmed the company's interest in this direction, saying Microsoft is exploring the potential of OpenClaw-like technologies in an enterprise context. This is the company's first official acknowledgment of specific interest in the platform. Previously, Microsoft had spoken about agentic AI only in general terms, without naming specific open-source projects as a reference point.

OpenClaw is an open-source platform for building AI agents that run locally on the user's device. Unlike cloud-based AI assistants, OpenClaw agents have direct access to the computer's file system, applications, browser and local data, without necessarily sending information to external servers. That makes the platform especially attractive to the enterprise market, where data security and regulatory compliance are critical. Companies are wary of handing confidential documents and business correspondence to third-party cloud services, and local agents could solve that problem.

OpenClaw's popularity surged in early 2026. Developers around the world began building agents on it to automate routine tasks: managing schedules, processing incoming email, analyzing documents, filling out forms and much more. In March 2026 the community held its first major conference, ClawCon in New York. It was a landmark moment: OpenClaw had grown from a niche developer tool into a full-fledged technology movement with thousands of participants.

Today, 365 Copilot works on a request-response model: the user asks a question or gives a command, and the assistant reacts. This differs fundamentally from the agentic model, in which the AI acts proactively. Integrating an OpenClaw-like architecture would let Copilot process incoming mail on its own and reply to standard requests, create tasks in the planner after meetings, prepare weekly reports by a set time, and track deadlines and warn about risks, all without an employee launching it manually. This is the scenario Microsoft describes as autonomous 24/7 operation.

The race among agentic AI systems for the enterprise market is already in full swing. Google introduced Agentspace, a platform for enterprise agents integrated with Workspace. Salesforce is developing Agentforce. Amazon and Oracle have announced their own solutions in this space. With one of the largest corporate customer bases in the world through the Microsoft 365 suite, Microsoft is well positioned to scale agentic technologies. If internal tests succeed, the updated Copilot capabilities could appear in upcoming releases. The company does not specify a timeline, but the market is not waiting.

For now, this is research, not a finished product. Still, Microsoft's confirmation of interest in local AI agent technologies points to an important shift: enterprise AI is moving from the role of an assistant that answers questions to that of an independent executor. For business, this promises to free up working time previously lost to routine. At the same time, it raises new questions: how to control the autonomous actions of AI, how to set limits on its authority, and how to ensure transparency of its decisions in a corporate environment.