Hand a Task to an AI Agent from a Workflow

The Invoke Managed Agent action lets a workflow pass a contact and some context to one of your published AI agents, wait for it to finish, and then continue using whatever the agent returned. You keep the predictable structure of the workflow while letting the agent handle steps that need reasoning, such as summarizing or deciding.

Prepare the agent first

1. Open AI Agents > Agent Studio > Managed Agents.
2. Add the Invoked from a workflow trigger to the agent.
3. Check that its instructions, knowledge, and permissions cover the task.
4. Test it, then publish it. Only published agents appear in the workflow builder.

Add the action to a workflow

1. Open the workflow and click + at the spot where the agent should act.
2. Search for Invoke Managed Agent and add it.
3. Select your published agent.
4. Write the instruction, saying clearly what the agent should do and what format to return. You can insert dynamic values so each contact sends its own details.
5. Attach files if the agent needs them as context.

Use the answer

Later steps can reference the agent's response, for example to update a Contact field, choose an If/Else branch, or insert text into an email or SMS. Add a step after the agent that checks the output before sending anything to customers.

Good fits

• Summarizing a long conversation into a short note.
• Judging lead quality and choosing a follow-up.
• Drafting a personalized reply for review.

Good to know: Save, then test with a sample contact before publishing, and read the agent's real output to confirm the format your next steps expect.