Automation often comes down to choosing the right path. Do we approve or reject? Route to team A or team B? Run the lightweight AI model or the advanced one?

The AI Router block makes these choices simple. You give it:

1. A prompt – what to decide.
   
2. Decision Data – the context it needs (e.g., query results, API responses).
   
3. Expected Decisions – the possible outcomes (A, B, C…).
   
   

The router evaluates the data using AI, picks one of your expected decisions, and sends the flow down the matching branch.

Think of it as a decision junction powered by AI.

How It Works

• Prompt: Define what the router should determine (e.g., “Determine if the order value exceeds $1,000”).
  
• Decision Data: Attach larger datasets without cluttering the prompt. For example, database results or order records.
  
• Expected Decisions: Preconfigure outcomes like “High Value”, “Low Value”, “Review Needed”. Each creates a connector in your flow.
  
• Everything Else: A built-in fallback if none of the defined outcomes match, so your flow never breaks.
  

Real-World Uses

• Sales – route leads by score (Hot, Warm, Cold).
  
• Support – classify tickets into Billing, Technical, or Other.
  
• Content – direct posts into Publish, Flag for Review, or Reject.
  
• AI Model Selection – choose between fast/cheap models and slower/high-quality ones.
  

Why It Matters

The AI Router is:

• Compact – one block replaces long conditional chains.
  
• Performant – evaluates and routes in a single step.
  
• Flexible – adapts to structured data, unstructured text, or mixed inputs.
  

The Takeaway

The AI Router gives you a clean, high-performance way to make branching decisions without clutter. Instead of hard-coding logic for every scenario, you define the possible routes, and the AI picks the right one.

👉 Next time your flow needs to decide between multiple paths, drop in an AI Router

‍