Microsoft is preparing to take its Copilot AI assistant beyond conventional question-and-answer interactions, with CEO Satya Nadella signalling the integration of newer AI models designed to handle complex and long-running tasks.

Nadella said advances across the artificial intelligence model ecosystem would be brought into Copilot to help users manage increasingly sophisticated workloads. The broader objective is to turn Copilot from a tool that primarily assists users into software capable of taking responsibility for completing multi-step tasks that can run for hours or even days.

One of the technologies at the centre of this shift is Project Opal, Microsoft's AI capability for task-oriented work in Microsoft 365 environments.

Copilot Moves Beyond Assistance

Traditional AI assistants generally respond to individual prompts, generate content or provide information. Microsoft's latest direction is focused on what the company describes as more autonomous, task-based computing.

Under this model, a user can give Copilot a broader objective rather than specifying every individual step. The AI can then reason through the request, create a plan, select appropriate tools and adapt its actions as the task progresses.

The approach could allow users to delegate workflows that previously required repeated manual intervention.

Microsoft's strategy places Copilot within the growing industry shift towards AI agents, systems designed not merely to answer questions but to perform sequences of actions on behalf of users.

Project Opal Targets Repetitive Work

Project Opal is designed for repetitive and manual tasks carried out within Microsoft 365 and related business environments.

Potential applications include compliance audits, employee onboarding and equipment ordering. Such workflows often involve several steps across different applications and can consume significant amounts of employee time.

Instead of requiring a worker to repeatedly instruct an AI system, Opal is designed to execute the workflow itself after receiving an initial task.

The technology is intended to operate within a controlled environment rather than having unrestricted access to a user's computer or the wider internet.

AI Builds Dynamic Task Plans

A central feature of Project Opal is its task-first operating model.

When a user starts a job, the underlying reasoning model converts the request into a dynamic plan. It can determine which tools are needed, sequence actions and modify the plan when circumstances change during execution.

This differs from a simple automated script, which normally follows a predetermined series of instructions.

The ability to adapt during execution is important for complex business workflows because unexpected information or a failed step can require the system to change its approach.

Microsoft's objective is to make such processes more manageable while retaining human oversight.

Users Can Watch AI Work

Project Opal also places emphasis on observability.

Users can monitor a task while it is running or review the process after completion. The job interface provides a view of the AI's plan alongside a live computer environment showing its activity.

Microsoft also describes a replay capability that allows users to examine what happened during a completed task.

This visibility is intended to address one of the major challenges associated with increasingly autonomous AI systems: knowing what the system did, why it took particular actions and where a workflow may have gone wrong.

Humans Can Take Control

Despite its autonomous capabilities, Project Opal is designed to keep users involved when necessary.

If the system reaches a step requiring human input, it can flag the issue through its activity timeline or a browser notification. The user can then intervene and complete the required action.

Users can also pause a task or take control of the underlying computing environment.

This human-in-the-loop approach is particularly relevant for enterprise workflows where certain actions may require approval, authentication or information that an AI system should not independently provide.

Enterprise Controls Add Guardrails

Microsoft says Project Opal has been designed with enterprise security and administrative controls.

Each task runs on a dedicated Windows 365 Cloud PC, while a separate supervisory system monitors the AI's activity and enforces defined safeguards.

Organisations can use allow lists to restrict the websites and resources the system can access. Administrators can also determine which scenarios users are permitted to initiate.

The system is opt-in, giving organisations greater control over where and how autonomous task execution is deployed.

These controls are intended to reduce the risks associated with AI systems that can interact directly with software and websites.

Copilot Enters Agentic Era

Microsoft's latest Copilot direction reflects a wider transformation taking place across the AI industry.

Companies are increasingly developing systems that can plan, use tools, browse software interfaces and complete multi-step workflows instead of simply generating text or answering questions.

For Microsoft, Copilot provides a large existing user base through products such as Microsoft 365, Windows and other enterprise services. Adding agentic capabilities could therefore change how users interact with workplace software.

The company is effectively seeking to move from “AI that helps you do a task” to “AI that can do the task under your supervision.”

Security Will Shape Adoption

The move towards autonomous AI also creates new questions around security, permissions and accountability.

An AI system that can interact with websites, applications and business information has greater potential to make mistakes or take unintended actions than a system that only produces text.

Project Opal's isolated Cloud PC environment, supervisory layer, activity logs and administrative controls are designed to address some of these concerns.

The success of Microsoft's approach will ultimately depend on whether businesses trust AI agents with increasingly important workflows while retaining sufficient human oversight.

For now, Project Opal represents Microsoft's broader effort to make Copilot more autonomous, observable and capable of handling complex work over extended periods.