Event-Driven Tasks, Skills and Browser Use in Copilot Cowork
Microsoft 365 Copilot has passed 30 million paid seats. The product direction says more than the seat count: Copilot Cowork now supports event-driven tasks, custom skills, plugins, browser use in Microsoft Edge and usage-based billing. Work can therefore start when an email, a Teams message or an @mention arrives, not only when a user opens a chat. In this article, I explain what event-driven and composable AI change for enterprises and how to measure value at the workflow level.
According to the Microsoft 365 Copilot news page, Microsoft 365 Copilot has passed 30 million paid seats, with net seat adds more than doubling quarter over quarter. For me, the product direction says more than the number. With Copilot Cowork, Microsoft builds an execution layer for work that carries out longer-running, multi-step tasks. The first phase of AI adoption focused on access and individual productivity. The next phase shows whether AI contributes to business execution without creating governance gaps or new risks.
From assistance to execution
Enterprise work rarely consists of a single interaction. It runs across several systems, people and approvals, with changing inputs and follow-up actions that happen later. Cowork is now generally available to Microsoft 365 Copilot tenants and has added event-driven tasks, custom skills, plugins, browser-based execution, usage-based billing and admin controls. Users increasingly ask AI to monitor, coordinate, trigger and continue work instead of delivering one result.
Event-driven tasks
With event-driven tasks, work starts when something happens: an email arrives, a Teams message comes in or someone is @mentioned. Many business processes begin with such signals, for example a customer escalation, a supplier message that shifts a timeline or a document that needs review. If AI responds to these signals within permissions and governance, it becomes operationally relevant. The question then changes from whether the model generates something useful to whether the system helps the organization respond better when conditions change.
Composable enterprise AI
Organizations customize Cowork by combining Microsoft 365 context, connected applications, reusable skills, model choice and admin governance. The closer AI gets to execution, the more it has to reflect the business environment: its systems, terminology, approvals, data boundaries and quality expectations. Value then comes from how well the organization assembles context, actions, rules and interfaces around the model, not only from the model itself.
Browser use
Local browser use in Microsoft Edge lets Cowork act in web-based systems that were not built for AI. Browser-based execution bridges the gap between intent and action across portals and web interfaces. At the same time, familiar questions arise: What may the AI do? Which approvals apply? How does IT observe activity and contain errors? How do organizational policies carry over to these actions?
Governance grows with capability
Microsoft names admin governance, model controls, consumption visibility, usage-based billing and alignment with its broader governance capabilities. When AI does more than answer, leaders need to know who uses it, which actions it takes, which systems it accesses, how usage is measured and where human oversight remains mandatory.
What organizations should consider
Suitable workflows Good candidates are repetitive but not trivial, cross-system, time-sensitive, easy to monitor and clearly bounded by policy and human review.
Explicit human control People direct higher-value work, and AI moves structured execution forward. Organizations have to define where approval, review or intervention remains mandatory.
Governance in step with capability As skills, plugins and event-driven actions are enabled, permissions, observability, billing transparency and policy enforcement have to scale too.
Value at the workflow level Prompts, sessions and satisfaction are easy to measure. More useful are shorter cycle times, fewer handoff delays, fewer missed follow-ups, more consistent processes and higher throughput without loss of control.
Copilot Cowork makes AI operationally usable. This raises the importance of trust and puts governance at the center. Organizations do not have to automate everything, but they have to decide which work AI should carry forward and under which controls.