Microsoft Copilot Studio

Copilot Chat lets users create agents

by HubSite 365 about Daniel Rohregger (365 Checkpoint)

Daniel Rohregger shows that users can create their own agents in free Copilot Chat, which lowers the barrier to tailored AI helpers.

Key insights

  • Users can create agents directly in free Copilot Chat.
  • The process starts with a natural-language description of the agent's job.
  • Agent Builder lets you set name, behavior, tone, and reply rules.
  • You can add files, websites, and other knowledge sources if allowed.
  • Microsoft points to Copilot Studio for more advanced scenarios.

Daniel Rohregger (365 Checkpoint) recently published a YouTube Short showing that users can create their own agents directly in free Copilot Chat. The takeaway is simple: Copilot Chat is moving beyond general prompts and into reusable assistants built for specific tasks.

The clip is short, so the details are limited, but it points to a workflow where a user describes what the agent should do and then refines it in the Agent Builder. That makes the feature relevant for people who want a focused assistant without building a full chatbot.

What's new

The main change is that Copilot Chat can now be used to create custom agents. Instead of starting from scratch with code, users describe the agent in natural language and then adjust its settings. Microsoft positions the Agent Builder as an embedded, simple way to create these agents, while Copilot Studio remains the option for more complex scenarios.

The material also indicates that an agent can have its own instructions, task scope, tone, and, depending on configuration and license, extra knowledge sources. In practice, that turns a chat experience into something closer to a specialist assistant.

Why it matters

This matters to makers, admins, and business users who need repeatable answers or structured output. The research notes mention common use cases such as document summaries, meeting prep, FAQ support, research, drafting, and repetitive administrative work.

For daily work, the value is consistency. A defined agent can follow the same rules each time instead of relying on ad hoc prompts. The notes also suggest a licensing or usage distinction: some agents may be based only on instructions and public websites, while others that use shared tenant data such as SharePoint or Graph connector content may be subject to usage-based billing.

How it works

  1. Choose Create agent in Copilot Chat.
  2. Describe the agent’s purpose in natural language.
  3. Set the name, behavior, tone, and response rules.
  4. Add knowledge sources such as files or websites if allowed.
  5. Test the agent in a chat before publishing it.
  6. Publish the agent and share it if organizational rules allow it.

The notes also describe the underlying structure: clear instructions, selected knowledge sources, a test area, and publication or sharing. That suggests the feature is designed for quick configuration first, with deeper customization reserved for Copilot Studio.

Limitations and open questions

The Short itself does not provide technical detail beyond the headline. It does not specify which tenants, licenses, or regions can use the feature, and it does not show the exact UI beyond the general workflow.

The research notes mention that some knowledge sources may depend on permissions and subscription, and that usage-based billing may apply in some cases. However, the source material does not confirm the full licensing model, so that should be treated as a boundary rather than a promise.

Sources

Frequently asked questions

Can users really create agents in free Copilot Chat?

The research notes say yes, users can create agents directly in free Copilot Chat. The Short itself does not provide extra licensing detail.

Do you need programming skills to build one?

No programming knowledge is described as necessary in the research notes. The workflow is based on natural-language instructions and later configuration.

What is the difference between Copilot Chat and Copilot Studio here?

Copilot Chat is presented as the simpler place to create agents, while Copilot Studio is mentioned for more complex scenarios. The material does not compare every feature side by side.

Can an agent use SharePoint or other tenant data?

The notes say some agents can use files, websites, SharePoint content, and connector data depending on permissions and subscription. They also note that tenant-data scenarios may be subject to usage-based billing.