Understanding the Conversational Model in Flow Builder

Overview

Every Conversational App created with Flow Builder follows a specific conversational model, the logic that determines how your bot interprets and responds to each user message.

This model defines the order in which the system processes messages, starting with the most specific triggers (like keywords) and moving toward more general or AI-driven ones (like AI Agents).

Understanding this flow will help you design effective, predictable, and intelligent conversational experiences.

The Four-Step Conversational Model

When a user sends a message, Flow Builder processes it in this order:

Keywords

The bot first checks whether the message contains any keywords you’ve defined.

• Keywords are exact words or short phrases that act as direct triggers for a content block.
• They’re ideal for explicit or repetitive inputs such as “help”, “menu”, or “open hours”.
• You can also use regular expressions to detect specific patterns, like order numbers or emails.

If a keyword match is found, the associated content is shown immediately, and the system stops processing the next steps.

Smart Intents

If no keyword matches, the bot checks for Smart Intents.

• Smart Intents use AI to understand the meaning of a user’s message, not just the exact words.
• You define short intent descriptions (e.g. “Track an order”) and the AI detects when the user’s message fits that intent.
• This allows flexible recognition of user goals — for example, “Where’s my package?” or “Check my delivery” can both trigger the same intent.

When a Smart Intent matches with high confidence, the bot triggers the connected flow.

AI Agents

If neither Keywords nor Smart Intents apply, the system forwards the user’s message to AI Agents.

• AI Agents are powered by large language models that can understand natural language and respond intelligently.
• They’re designed to handle open-ended or complex queries that aren’t explicitly covered by flows.
• AI Agents can pull from your Conversational App’s context and knowledge sources to give coherent answers or route users appropriately.

If the AI Agent provides a relevant response, it’s returned to the user in real time.

Knowledge Base

If no match is found through the steps above, Flow Builder checks the Knowledge Base.

• The Knowledge Base acts as a library of frequently asked questions and answers.
• This helps the bot handle generic informational requests without specific flows.
  

If a Knowledge Base article matches the user’s query, its answer is displayed.

When No Matches Are Found: The Fallback Flow

If no results are found from any of the four steps, the bot triggers the Fallback Flow.

This flow ensures users never reach a dead end.

You can customize it to:

• Ask users to rephrase their message, or
• Offer a handoff to a human agent for live support.
  

Why This Order Matters

The processing order ensures:

• Specific rules (like keywords) take priority over broader AI responses.
• You maintain full control over how your bot behaves in predictable scenarios.
• AI Agents and Knowledge Base answers act as intelligent safety nets rather than replacements for well-defined flows.
  

Best Practices

• Use Keywords for fixed triggers like commands or menu shortcuts.
• Use Smart Intents for user goals that can be described simply.
• Enable AI Agents for open-ended or flexible conversations.
• Keep your Fallback Flow friendly, helpful, and human-like.