The AI Condition Rules node is a smart filter in your automations. Instead of manually specifying dozens of exact values like in regular rules, you describe the desired segments in plain language. Artificial intelligence understands the context and distributes contacts into the appropriate branches on its own.
What data the AI analyzes & pricing
To make decisions, the AI is provided with basic information from the contact's card:
-
What the AI sees: Name, current position, headline, about section, company name, location, and profile language. (The list of available fields will expand in future updates).
-
What the AI does NOT see: Message history, custom fields.
-
Pricing: This feature uses the platform's built-in AI model. Processing a single contact costs 1-3 LLM credits. The total number of available credits for your workspace depends on your current subscription plan.
-
If the credit limit is reached, the node will pause its operation (display a limit error).
How to set up the node: step by step
Setup consists of three simple blocks:
-
Classification instructions (Optional): Here you can provide general guidelines for the AI. For example, explain priorities: "Focus primarily on the position, not the headline" or "If a person is both a marketer and a business owner, classify them primarily as a business owner."
-
Branches: By default, the node includes one primary branch and one alternative branch. You can add as many extra branches as needed. In the description field, explain in plain text which contacts should enter this branch.
🟡 Note: Using variables in the
{{variable}}format is not supported in this node; they will not work. Describe criteria in plain text only. -
Alternative Branch: This branch always triggers last and does not require a description. All contacts that did not match any of the descriptions, or contacts with completely empty profiles, are automatically routed here.
Tips for better accuracy
-
Describe exceptions directly: AI tends to look for the closest match, so clear boundaries are very important.
-
One node - one criterion: Avoid sorting by region and position within the same node. Instead, place two AI Condition Rules nodes in sequence.
-
Provide examples in descriptions: Adding specific examples within branch descriptions significantly improves classification accuracy in edge cases.
-
Test on small batches: First add 10-20 contacts to the automation, check how the AI distributed them, and adjust branch descriptions if necessary.
-
Keep empty profiles in mind: Contacts lacking profile data (missing data about position, about section, etc) will always go to the Alternative branch. Make sure to design a logical next step for them in your flow.