Agentic actions in Automations

Bring AI agents into your automated workflows with Agentic actions. With this step in Automations, you can deploy an agent to do work triggered by any action already available in Automations.

Agentic actions let you add an AI agent to your automations. Instead of following fixed rules, the agent reads the information your automation receives, like a form submission or a new contact. It decides what needs to happen and does it in the apps you’ve connected, such as Slack, Jira, or Gmail.

What are agentic actions?

An agentic action is a step you add to an automation, like any other action. The difference is that you don’t set up each step yourself. You describe the goal in plain language, and the agent figures out how to get there.

When your automation runs:

  1. The automation’s trigger fires, for example, when someone submits a form or a contact is created.
  2. The agent reads the data from that trigger.
  3. Once you've added your goal and any context, it decides what needs to happen.
  4. It takes action in your connected apps, like posting in Slack or creating a Jira ticket.

For example, your goal might be: “If the feedback reports a problem, post it in our #product-feedback Slack channel and create a Jira ticket for the product team.” The agent reads each response, decides whether it reports a problem, and takes those steps when it does.

Agentic actions are most useful when responses are open-ended, like customer feedback, requests, or detailed explanations. For simple, predictable workflows, a standard action may be all you need.

What you need:

Here's what you need to know about adding Agentic actions:

  • You can only add one Agentic action per automation.
  • Your Agentic action must be placed at the bottom of the automation (or be the only action in the automation). 
  • Each Agentic action run consumes 10 action credits, including test runs.

Create Automations with Agentic actions

In the example below, we’ll create an automation that will be triggered when a new contact is created, then we’ll use the power of Agentic actions to draft a message with Typeform’s Gmail MCP connector that we can send to our new contacts to welcome them and share a link to our signup form.

1. From your Workspace, click Automations.

2. Click + Create automation.

3. Select what you would like to trigger the automation. Click Form submission, Contact activity or updates, or Scheduled. We'll select Contact activity or updates.

4. Give your automation a name by clicking the default name in the top left corner.

5. Type in the name of your automation and click Save. We’ll enter Qualify my leads.

6. Select what will trigger this automation from the dropdown list. We’ll select Contact is created.

7. Optionally, add a contact list and contact property filter if you want to filter this automation for a specific group of Contacts.

8. Click the + icon in the middle to add an action.

9. Then select Agentic action.

10. Select a goal from the list of available goal templates on the right, or click Create your own goal to set up a new one. 

11. When setting goals, you can tell the agent what you want to do with each new contact. We recommend starting with one of Typeform’s examples, then editing it to fit your needs. You can add up to 2,000 characters to the text field to describe your goal.

12. Optionally, click Add context to add some additional information about your business or audience. You can add up to 2,000 characters to the text field.

13. Click the Manage connectors icon to add an MCP connector to your agentic action.

14. Now you can select the AI connectors you want to use from the list of apps that appears.

15. For example, if your goal is to draft an email for your agentic action, you can select Gmail as your AI connector.

16. Now that your automation is set up, click Activate.

You can also click Run test to see what your Agentic action will look like in real life.

In the example below, the test run flagged that the task could not be completed because the link to the signup form mentioned in the goal description was missing.

After adding the missing information and running a new test, the task was completed successfully.

HC_BobRoss_ilu_cropped.png
Note! Test runs of agentic actions use 10 action credits.

When you’re happy with your setup, click Publish edits to make the changes made to the automation live.

You can click the three dots to reset or delete an Agentic action.

Or click Stop automation to turn it off.

Use Typeform AI to create Automations with Agentic actions

In the example below, we’ll create an automation that uses a feedback form as its trigger, and we’ll use the power of Agentic actions to notify our team via Slack messages to keep them up to date about new submissions and create Jira tickets so they can take action quickly when they receive negative feedback.

  1. Log in to your Typeform account and speak or type in the Typeform AI field to get started. 

You can describe the workflow you want performed by Typeform AI in a single prompt. 

For example, you can:

  • Ask Typeform AI to create a form for you, 
  • Then set up an automation based on responses to that form,
  • Then add tasks like sending automation actions to external tools.

After some thinking, Typeform AI will show you the steps it has performed to complete the task, and you’ll also see the form and the automation it created based on your prompt.

You can click the name of the form to review and edit it in the Content panel. Click Share to publish your form and make it live.

Click the name of the automation created to open it and review the workflow.

You’ll see that an Agentic action has also been added to the automation, and you can click on it to view its details.

You can also scroll down to see which AI connectors have been added to your Agentic action.

When you’re happy with your automation setup, click Activate.

You can also click Run test to see what your Agentic action will look like in real life.

HC_BobRoss_ilu_cropped.png
Note! Test runs of agentic actions use 10 action credits.

Now you’ll see the Activity log of your Agentic action with a Summary of the AI-generated test response and the completion status of the task. In the example below, the test run has been completed successfully.

You can also submit a test response to your form to test the full workflow. In the example below, the task failed because the Jira project hasn't been specified, and tickets couldn’t be created.

You can add the missing information to your Agentic action based on the explanation provided for failed tasks and have them up and running in no time. For a full list of supported apps, see Use MCP connectors with Typeform AI.

Was this article helpful?

Have more questions? Submit a request

Tap into our community knowledge

Find answers

Can’t find what you’re looking for?

Contact support