> ## Knowledge Base Index
> Fetch the complete knowledge base index at: https://help.textit.com/sitemap.xml
> Use this file to discover available pages before exploring further.
> Pure-Markdown content can be obtained by appending a '.md' suffix to the content URLs listed in the sitemap (without the trailing slash).

# Split by AI 

The **Split by AI** action uses LLMs to automatically categorize incoming messages and route contacts down different paths in your flow.

Instead of writing rigid keyword rules, you can define natural-language categories and let the AI handle the nuances of human language—including typos, slang, and context.

Use this action when you need to make a routing decision based on the meaning of a contact's response.

![](https://storage.crisp.chat/users/helpdesk/website/-/7/a/9/f/7a9fc4558a9de000/screenshot-2026-05-20-at-10303_1atjk7y.png =600x456)
### **Configuration**

1. **Split by AI**: Select the ‘Split by AI’ Action. Choose which LLM you will use to power the action. 
2. **Input:** Define the text to analyze (usually @input.text for the contact's last reply).
3. **Categories:** Enter the exact categories you want the AI to choose from.
4. **Save Result (Optional):** Enter a variable name to save the chosen category to the flow context (e.g., ticket\_type).

![](https://storage.crisp.chat/users/helpdesk/website/-/7/a/9/f/7a9fc4558a9de000/screenshot-2026-05-20-at-10104_m96spp.png)
### **Split by AI vs. Call AI**

**Split by AI** is for **routing**. It forces the AI to pick one category from a fixed list and automatically builds physical paths in your flow editor.
* *Use this to decide where a contact goes next.*
[Call AI](https://help.textit.com/en/article/call-ai-action-dbbjwo/) is for **generation**. It is a blank canvas where you write a custom prompt to have the AI generate text, answer questions, or summarize a conversation. It does not create branching paths.
* *Use this when you want AI to write something.*

### **Use Cases**

* **Support or Support Triage:** Classifying messages into Billing, Tech Support, or Sales to route them to the right team.
* **Sentiment Analysis:** Sorting feedback into Positive, Neutral, or Negative to trigger appropriate follow-ups.
* **Survey Classification:** Grouping open-ended answers into defined themes like Price, Features, or Usability.