Objective: Provide Automated Support and Escalate to a Human
This flow will use an AI node to understand a user's problem, attempt to answer it using trained data, and assign the chat to a human agent if the user needs to speak to someone.
Nodes we will use: `Incoming Message`, `LLM / OpenAI`, `Branch`, `Assign Agent`, `Message`.
Prerequisite: You should have some trained data (FAQs, Files, or Websites) in your AI Bot Training sidebar for the LLM to search.
Step 1: Configure the LLM Node
- Start with an `Incoming Message` trigger.
- Connect an `LLM / OpenAI` node.
- System Prompt: "You are a friendly and helpful customer support agent. First, try to answer the user's question using the provided context. If you cannot answer, or if the user asks to speak to a human, say 'Please hold on, I am connecting you to an agent.'"
- User Prompt: `{{last_message}}` (This passes the user's latest message to the AI).
- Enable Vector Search: Make sure this is checked so the AI can use your trained data.
- Variable Name: Set this to `ai_response`.
- Intentions: Add an intention with the name `needs_human` and the description "The user wants to speak to a human agent, is asking for a person, or is expressing frustration."
- Intent Trigger Keywords: For precise control, add phrases like "speak to an agent", "talk to a person", "connect me to a human".
Step 2: Create the Logic Branch
- Add a `Branch` Node: Connect the LLM node to a `Branch` node.
- Set the Condition: The condition should check if the AI detected the user's intent. Set it to: `{{ai_response_intent}}` equals `needs_human`.
- "False" Path (AI can answer): Connect the `false` handle to a `Message` node. Set the message to `{{ai_response}}`. This will send the AI's answer to the user. Connect this to an `End Flow` node.
- "True" Path (Needs Human): From the `true` handle, connect a `Message` node that sends the AI's holding message: `{{ai_response}}`.
- Assign to Agent: Connect this message to an `Assign Agent` node. Select the agent or team you want to handle the support request.
- End the Flow: Connect the `Assign Agent` node to an `End Flow` node.
Your support bot will now intelligently handle queries, answer what it can, and seamlessly escalate complex issues to your team!