Understanding the User

There are two core ways AI agent understands what the user is saying, depending on whether you use static logic (rule-based) or generative replies.

Rule-Based Understanding (Intent + Entity Detection)

With static logic, we first try to detect the user’s intent and optionally extract relevant entities (like names, numbers, or product types). Then, based on what was detected, specific dialog flow is started.

You define possible intents (e.g., Check order status, Cancel subscription)
You can optionally extract entities (e.g., order ID, product name)
Then, route the conversation to the appropriate dialog flow you created

Example:

Intent: Check delivery status
Entity: Order ID = 12345
→ Start the "Check Order Status" dialog flow, fetch delivery info, and respond

Summary:

✅ Fully controllable
✅ Useful when flows are structured
⚠️ Requires setup of intents and routing rules

Generative Understanding (AI Knowledge or Free Prompt)

When using AI Knowledge or a Free Prompt module, the intent detection step is skipped. The model reads the user’s message and immediately generates a response based on:

A user prompt you defined
Conversation history
Available documents or websites (RAG like architecture)

No need to match an intent or start a pre-built flow — the AI replies in one go.

User: "How long does delivery usually take?"
→ AI Knowledge scans your shipping policy, FAQs, and returns page, then replies:
“Standard delivery takes 3–5 business days. Express options are also available at checkout.”

✅ Faster to set up
✅ More natural replies to unpredictable, repeated or follow-up questions
⚠️ Less structure and control
⚠️ May vary in output depending on prompt/knowledge base quality