Conditions let your agent make smart decisions about which path to take in a workflow. Instead of rigid rules like “if field equals X”, you write natural language prompts like “if the customer seems angry” or “if this email needs a response”.Your AI agent evaluates these conditions by understanding context, sentiment, and meaning — not just exact matches. This means you can handle complex scenarios that would be impossible with traditional automation tools.Use conditions to route different types of requests, filter content intelligently, or handle multiple scenarios within a single workflow. Your agent chooses the best path based on what it actually understands from the data.
By default, if no conditions are met, your agent will stop the task. You can change this behavior by enabling “Force the agent to select a branch”.
When enabled, your agent must choose one of the defined paths even if none of the conditions perfectly match. This is useful when you want to ensure the workflow always continues.
Be explicit about what you’re looking for. Instead of “important email”, use “email from a C-level executive or marked as high priority”.
Test Edge Cases
Consider what happens when conditions overlap or when none match. Use the test panel to verify behavior.
Add Examples
Include examples in your condition prompts to improve accuracy: “the email asks about pricing (e.g., ‘how much does it cost’, ‘what are your rates’)”.
For conditions that require understanding context, sentiment, or complex logic, use more advanced models. For simple keyword matching or basic categorization, faster models work well.
Enable “Force agent to select branch” or add a catch-all condition
Wrong path is chosen
Review and refine your condition prompts
Add more specific examples
Conditions overlap
Make each condition mutually exclusive
Order from most specific to least specific
If your workflow stops unexpectedly, check that each condition branch has at least one action following it. Empty branches will cause the execution to end.