AI agents

Create and publish your first AI agent

Set an agent name, instructions and knowledge, test its limits, then publish it.

Published · Updated

A Message Pro AI agent follows the instructions, knowledge and actions configured for it. Start with one clear task instead of asking it to handle every type of request at once.

Before you start

You need AI-agent administration access and approved guidance on what the agent may say or do. Prepare real customer questions with sensitive details removed for testing.

Steps

  1. 1

    Open AI agents

    In the workspace, open AI agents and choose to create an agent or start from a suitable template if available.

  2. 2

    Name it and define its job

    Choose a clear name and one task, such as answering a defined set of service questions.

  3. 3

    Write instructions

    Define response style, when to say it does not know and when to hand off to a teammate. Do not ask it to invent prices or promises.

  4. 4

    Connect knowledge and actions

    Enable only the knowledge sources and permitted actions required for this task. Review each action before use.

  5. 5

    Test and publish

    Try clear questions, ambiguous questions and a handoff case. Fix the results, then publish when the behavior is acceptable.

Check the result

In testing, the agent answers from approved knowledge, recognizes limits and follows the configured handoff behavior before customer use.