Categorize with AI

Categorize with AI lets you drop an AI classification step into any automation. Give it text — a chat summary, a form submission, a webhook payload — and define your categories. The AI reads the input, picks the best match, and the automation branches accordingly. No code, no rules engine, no manual routing. Set it up in minutes and every inbound message, lead, or support request goes to the right team automatically.

How it works

1. You provide input text — the content the AI should read (a message, form response, note, etc.)
2. You define categories — each with a name and an optional description that helps the AI decide when to pick it
3. When the automation runs, the AI reads the input and picks the best matching category
4. The automation continues down that category's branch

If the AI determines that none of your categories match, it returns an empty result and the automation follows the default path.

Set up the action

1. Open the automation builder and add a new step
2. Under Advanced, select Categorize with AI
3. Configure the two sections:

1. Choose what to categorize

In the Input text field, enter or insert the text the AI should classify. Select Add dynamic content to insert data from a previous step — for example, a customer message from a trigger, a form response, or a webhook payload.

tip

Dynamic content values are resolved before the AI reads the text, so the AI always sees the actual data — not the placeholder.

2. Define the categories

Add at least one category. For each category, provide:

Field	Required	Description
Category display name	Yes	A short label that identifies this category (e.g., "Billing", "Support", "Sales"). This becomes the branch name in your workflow.
Description	No	A sentence explaining when the AI should choose this category. The more specific, the better.

• Click Add another category to add more (up to 20)
• Drag categories to reorder them
• Click the x button to remove a category

info

Each category you define creates a separate branch in the automation. After classification, the automation continues down the matching branch only.

Example configuration

Scenario: Route Web Chat messages by intent so the right team follows up.

Trigger: Web Chat captures a lead

Input text: The lead's message, inserted via Add dynamic content → From a previous step → Trigger

Category display name	Description
Pricing	The visitor is asking about cost, plans, or a quote
Support	The visitor needs help with a product issue or something that is broken
General inquiry	The visitor has a general question that does not fit the other categories

When a Web Chat message comes in, the AI reads it and picks the best match. A message like "How much does your SEO package cost?" goes down the "Pricing" branch. "My website dashboard won't load" goes down "Support." Everything else follows "General inquiry."

Outputs

After the step runs, two values are available to downstream steps:

Output	Description
Classification key	The name of the matched category. Empty if none matched.
Reason	A one-sentence explanation of why the AI chose that category.

You can reference these outputs in later steps using Add dynamic content → From a previous step.

Tips for writing good categories

• Be specific in descriptions — "Customer is asking about an invoice or payment" works better than "Money stuff"
• Keep categories distinct — Overlapping descriptions make it harder for the AI to choose. If two categories sound similar, combine them or sharpen the descriptions.
• Use 2 to 10 categories — The action supports up to 20, but fewer well-defined categories produce more consistent results
• Test with real text — Run your automation with sample data to verify the AI categorizes as expected before going live

Use case examples

Route AI employee responses

Automatically categorize AI employee conversation outcomes so resolved chats get logged and unresolved ones reach a human.

Step 1 — Set the trigger

Select A contact communication summary is created (under Contacts). This fires whenever an AI employee finishes a conversation and a summary is generated.

Step 2 — Add the Categorize with AI step

• Under Advanced, select Categorize with AI
• Input text: select Add dynamic content → From a previous step → Trigger → Chat Summary
• Add three categories:

Category display name	Description
Resolved	The conversation was fully handled and the customer's question was answered
Needs human follow-up	The AI could not fully resolve the issue or the customer asked to speak with a person
Escalation required	The conversation involves a complaint, urgent issue, or potential churn risk

Step 3 — Configure each branch

• Resolved → Add a Create a CRM note step to log the resolution on the contact
• Needs human follow-up → Add a Create a CRM sales task for the contact step assigned to the contact's owner
• Escalation required → Add a Notify a salesperson step targeting the team lead

tip

The trigger also outputs a Chat Link Path. Include it in the task or notification body so the person following up can jump straight to the conversation.

Triage support requests

Route form submissions to the right team based on what the customer is asking about.

Step 1 — Set the trigger

Select A form is submitted for a contact (under Contacts). This fires when a contact submits a form linked to a company.

Step 2 — Add the Categorize with AI step

• Under Advanced, select Categorize with AI
• Input text: select Add dynamic content → From a previous step → Trigger and choose the form field that contains the customer's message
• Add four categories:

Category display name	Description
Billing	Customer is asking about an invoice, payment, or subscription
Technical support	Customer needs help with a product issue or something that is broken
Account access	Customer cannot log in or needs their credentials reset
Feature request	Customer is suggesting a new feature or improvement

Step 3 — Configure each branch

• Billing → Add a Create a CRM sales task for the contact step assigned to the billing team
• Technical support → Add a Start a campaign for the contact step that sends a troubleshooting guide
• Account access → Add a Create a CRM sales task for the contact step with a note to reset credentials
• Feature request → Add a Create a CRM note step to log the request on the contact

Qualify inbound leads

Categorize new contacts by intent level and respond accordingly — high-intent leads get a salesperson, everyone else enters a nurture flow.

Step 1 — Set the trigger

Select A contact is created or modified (under Contacts). Add a trigger condition to filter for newly created contacts only (e.g., lifecycle stage equals "Lead").

Step 2 — Add the Categorize with AI step

• Under Advanced, select Categorize with AI
• Input text: select Add dynamic content → From a previous step → Trigger and choose the field that contains the lead source or inquiry message
• Add three categories:

Category display name	Description
High intent	The lead is ready to buy, requesting a demo, or asking about pricing
Medium intent	The lead is researching options or comparing products
Low intent	The lead is browsing, signed up for a newsletter, or has no clear buying signal

Step 3 — Configure each branch

• High intent → Add an Assign a contact owner step to route to a salesperson, then a Notify a salesperson step so they follow up immediately
• Medium intent → Add a Start a campaign for the contact step that enrolls them in a nurture email sequence
• Low intent → Add a Start a campaign for the contact step that sends a single welcome email

Related resources

• Automation steps reference — Full list of available actions
• Advanced automation features — Logic steps, delays, and branching
• Creating and configuring automations — Build and configure workflows
• Data expressions in automations — Extract and transform data between steps