Microsoft Copilot and software beyond menus

Key insights

• The proposed shift is from navigating menus to describing a goal in natural language.
• Agents can combine information retrieval and tool use across multiple steps.
• Access to organizational data and connected tools is central to carrying out useful tasks.
• The material does not specify a Copilot setting, license, rollout stage or setup procedure.

Microsoft Copilot and AI agents point toward software that can act on a user's goal instead of requiring every task to be completed through menus. This approach matters to Microsoft 365 users, IT teams and developers considering how AI could work across apps and business data. Matthew Berman is the source associated with the short demonstration, but the supplied clip transcript contains no spoken or shown details to verify.

What's new

The broader idea is to make the interface less central for some work. Rather than opening several apps and manually moving information between them, a user could describe an outcome and have an AI agent interpret the request, retrieve relevant information and use connected tools to perform steps.

This is different from a chatbot that only responds with text. An agentic system is intended to pursue a goal over multiple steps, potentially making decisions within defined limits or asking for human approval. The research notes describe this as a direction across Microsoft products and services, not as one named feature with confirmed availability.

Who it is for and what you need

This direction is relevant to users who move between applications, administrators assessing AI controls, and developers building AI-powered experiences. Useful agents depend on access to relevant business information and the tools needed to act on it.

The supplied material does not identify required Microsoft licenses, tenant settings, prerequisites, regions or a rollout stage. It also does not establish that any particular Copilot experience or agent capability is available in a customer's environment. Treat the examples as a description of the broader approach, not deployment guidance.

How it works

The research notes describe a typical agent workflow rather than a verified product walkthrough:

1. A user states an intended outcome in natural language.
2. The AI interprets the request and identifies relevant information or steps.
3. Connected data sources and tools provide context or let the agent take actions.
4. The agent returns a result or, where configured, seeks human review for an action.

For organizations, the important consideration is that agents need governed data access and clear boundaries on what they may do. The supplied material does not provide menu paths or instructions for enabling these capabilities.

Limitations and open questions

Automating actions raises questions about permissions, oversight and the reliability of results. The research notes identify identity, security, compliance and governance as relevant controls, but the available source material does not explain how a specific Microsoft product implements them.

There is no confirmed feature name, setup process, license information or availability detail in the supplied transcript. No official Microsoft source was provided, so readers should not treat this article as confirmation of a current product change.

Sources

• Matthew Berman's YouTube Short
• Matthew Berman's original Short page

Frequently asked questions

What does it mean for AI to work beyond a software interface?

It means describing a goal in natural language and having an agent retrieve information or use connected tools to complete steps, rather than manually navigating every app.

Does Microsoft Copilot currently complete tasks across all my apps?

The supplied material describes a broad direction, not confirmed availability in specific products or tenants. It does not establish which apps or actions are supported.

What license or setup is required for Microsoft AI agents?

The material does not specify licenses, prerequisites, or setup steps.