Copilot Studio hooks: rebuild topic logic

Key insights

• The example rebuilds an existing custom topic for the Microsoft Copilot harness.
• The hook workflow uses the User Prompt Submitted event type.
• The described enhancement is identified as a Hooks Preview feature.
• The example uses Classify in place of AI Builder prompts.
• The description compares workflow use and model selection in the two approaches.

Microsoft Copilot Studio hooks provide a way to rebuild selected custom-topic logic as a workflow tied to an agent event. This approach matters to Copilot Studio makers and agent developers evaluating the Microsoft Copilot harness. Daniel Christian [MVP] demonstrates a test migration from a standard harness agent topic to a hook workflow.

What's new

The change explored here is a different place to implement some agent behavior: instead of leaving the example in a custom topic, the demonstration recreates it using hooks. The published description identifies the work as a test, not a claim that every topic should be replaced. It also labels the hook enhancement shown as Preview.

The example focuses on the User Prompt Submitted event type. In practical terms, this names a point associated with a submitted user prompt where the demonstrated hook workflow is configured. The available description does not specify the full event contract, inputs, outputs, or what happens for every possible prompt.

A second change in the example concerns AI Builder prompts. The description says these are replaced by Classify in the rebuilt workflow. It does not provide enough detail to infer a general product-wide deprecation or replacement of AI Builder, or to establish that Classify is suitable for every prompt scenario.

Who it is for and what you need

This walkthrough is most relevant to people building or maintaining Copilot Studio agents who want to compare an existing custom topic with a hook-based implementation. The specific scenario uses the Microsoft Copilot harness, and the published material calls out the GitHub Copilot harness agent as part of the transition in naming and context.

The hook capability is described as a Preview enhancement. The material does not state which licenses, plans, regions, or tenant settings are required, nor does it give a rollout schedule or confirm general availability. Check the current Copilot Studio interface and Microsoft documentation before planning a production migration.

The demonstration starts with an existing topic scenario and AI Builder prompts, then reviews the harness and rebuilds the workflow. It does not give a complete prerequisite checklist, so details such as required permissions, environment configuration, and any service dependencies remain unstated.

How it works

The useful takeaway is the comparison of two implementations of a particular scenario, rather than a universal recipe for converting topics. The described hook is associated with the User Prompt Submitted event. The workflow is then tested and compared with the original harness-based implementation.

1. Start with the standard harness agent scenario and identify the custom topic behavior you want to reproduce.
2. Review the topic and its AI Builder prompts so you can identify what the existing workflow does.
3. In the Copilot Studio harness experience, review the Hooks Preview enhancement and select the User Prompt Submitted event type for the demonstrated workflow.
4. Rebuild the scenario in the hook workflow, using Classify in place of the AI Builder prompts in this example.
5. Test the flow, then compare how the original and rebuilt versions handle the workflow and model selection.

The demonstration description does not specify exact screen labels, menu paths, field mappings, or configuration values. Treat the steps above as the sequence of tasks it describes, not as a complete click-by-click guide. For a larger migration, first separate the parts of a topic that are needed for the user conversation from the logic being tested in the hook.

The comparison notes that both implementations use a workflow and both offer an option to select a model for the agent and for prompt summary. It also says the standard harness allows use of AI Builder, which provides an option to leverage Microsoft Foundry models. Those are observations about the compared scenario; they do not establish that every hook workflow has the same options.

Limitations and open questions

The available material covers one example and calls it a test. It does not establish that hooks replace topics across Copilot Studio, explain which topic features transfer, or show how to migrate a complex conversational journey. Topics and hooks should not be treated as interchangeable based on this example alone.

Important implementation details are also missing from the published description: required access, supported event behavior, exact Classify setup, error handling, and the effect of a failed workflow. It does not say whether the rebuilt version behaves identically in all cases or provide measured results. Validate those points in your own environment before relying on the pattern.

Preview status is the clearest operational caveat. Preview behavior and configuration may differ from a later release, but the available material does not describe specific limitations or timelines. The Microsoft Copilot harness documentation and Copilot Studio agent experience documentation are useful places to check alongside the current product experience.

Sources

• Microsoft: The GitHub Copilot harness is now the Microsoft Copilot harness
• Microsoft Copilot Studio agents experience overview
• Daniel Christian [MVP] demonstration: Rebuilding topics in hooks, Part 1

Frequently asked questions

What does the Copilot Studio hooks example rebuild?

It tests rebuilding an existing custom-topic scenario for the Microsoft Copilot harness as a hook workflow associated with the User Prompt Submitted event.

Are Copilot Studio hooks generally available?

The described Hooks enhancement is identified as Preview. The material does not state a general availability date or rollout scope.

Does Classify replace AI Builder prompts in Copilot Studio?

Classify replaces AI Builder prompts in this demonstrated workflow. The available information does not establish a general replacement across Copilot Studio.

What licenses or prerequisites do Copilot Studio hooks require?

The material does not specify required licenses, plans, permissions, regions, or tenant settings.