Chat Intents

Available since FlowX.AI v5.5Chat Intents are available starting with FlowX.AI version 5.5.

Overview

Chat Intents enables intent-based navigation and routing within Chat components to support sophisticated conversational AI workflows. This feature allows the system to classify user intents and route conversations accordingly, enabling dynamic workflow routing, UI Flow navigation from chat interactions, and context preservation during intent transitions.

Intent classification

AI-powered detection of user intents from messages with multi-intent and context-aware support

Dynamic routing

Route conversations to appropriate workflows based on detected intents

UI Flow navigation

Navigate to UI Flow screens directly from chat interactions

Context preservation

Maintain conversation context during intent transitions and thread management

Chat Intents builds on and integrates with:

Chat Component (released in v5.4.0)
Conversational Workflow & Memory
Routing Agent / Intent Classification workflow nodes

How intents work

Chat Intents provide a structured way to handle user inputs that match known patterns, while falling back to AI for unmatched queries.

Intent processing flow

User sends message

The user types a message in the chat interface.

Intent matching

The system analyzes the message against configured intents using:

Keyword matching
Pattern recognition
Semantic similarity

Action execution

If an intent matches:

Execute the configured action (workflow, response, or custom logic)
Display any quick reply options

If no intent matches:

Forward to the AI agent for natural language processing

Response delivery

The response is displayed in the chat interface.

Configuring intents

Accessing intent configuration

Open Chat component settings

In UI Flows Designer, select your Chat component and open the Settings panel.

Navigate to Intents

Click on the Intents tab to access intent configuration.

Add new intent

Click Add Intent to create a new intent configuration.

Intent properties

Property	Description	Required
Name	Unique identifier for the intent	Yes
Display name	Human-readable name shown in UI	Yes
Trigger patterns	Patterns that activate this intent	Yes
Action type	What happens when intent is triggered	Yes
Response	Message to display (for response actions)	Conditional
Workflow	Workflow to execute (for workflow actions)	Conditional
Quick replies	Suggested follow-up options	No
Priority	Order of evaluation when multiple intents match	No

Trigger patterns

Define how user messages are matched to intents.

Pattern types

Keyword matching

Match messages containing specific keywords:

{
  "type": "keywords",
  "values": ["account balance", "check balance", "balance inquiry"]
}

Matching behavior:

Case-insensitive by default
Matches if any keyword is present
Supports partial matching with wildcards

Exact match

Match messages that exactly match specified phrases:

{
  "type": "exact",
  "values": ["help", "start over", "main menu"]
}

Matching behavior:

Case-insensitive
Requires exact phrase match
Trims whitespace

Regular expressions

Match messages using regex patterns:

{
  "type": "regex",
  "pattern": "^(show|display|get)\\s+(my\\s+)?account"
}

Matching behavior:

Full regex support
Useful for complex patterns
Performance consideration for complex patterns

Semantic matching

Match messages with similar meaning:

{
  "type": "semantic",
  "examples": [
    "What is my account balance?",
    "How much money do I have?",
    "Show me my balance"
  ],
  "threshold": 0.8
}

Matching behavior:

Uses AI to understand intent
Matches semantically similar phrases
Configurable similarity threshold

Combining patterns

Combine multiple pattern types for comprehensive matching:

{
  "patterns": [
    {
      "type": "keywords",
      "values": ["balance", "account"]
    },
    {
      "type": "semantic",
      "examples": ["How much money do I have?"],
      "threshold": 0.75
    }
  ],
  "matchMode": "any"
}

Match mode	Description
any	Triggers if any pattern matches
all	Triggers only if all patterns match

Intent actions

Configure what happens when an intent is triggered.

Action types

Response action
Workflow action
Navigation action
Process action

Display a predefined response message:

{
  "actionType": "response",
  "response": {
    "message": "Your current account balance is ${accountBalance}.",
    "format": "text"
  }
}

Options:

Static text messages
Dynamic content with variables
Markdown formatting support

Execute a workflow and return the result:

{
  "actionType": "workflow",
  "workflow": {
    "name": "get-account-balance",
    "inputMapping": {
      "userId": "${session.userId}"
    },
    "outputMapping": {
      "balance": "accountBalance"
    }
  }
}

Options:

Select target workflow
Map input parameters from chat context
Map output to chat response

Navigate to a different screen or open a modal:

{
  "actionType": "navigate",
  "navigation": {
    "target": "account-details",
    "params": {
      "tab": "transactions"
    }
  }
}

Options:

Navigate to UI Flow screens
Open modals
Pass navigation parameters

Start a FlowX process:

{
  "actionType": "startProcess",
  "process": {
    "name": "loan-application",
    "inputData": {
      "customerId": "${session.customerId}"
    }
  }
}

Options:

Start new process instances
Pass initial data
Handle process completion

Quick replies

Configure quick reply buttons that appear after an intent response, guiding users to common follow-up actions.

Quick reply configuration

{
  "quickReplies": [
    {
      "label": "View transactions",
      "value": "show my recent transactions",
      "icon": "list"
    },
    {
      "label": "Transfer money",
      "value": "I want to transfer money",
      "icon": "arrow-right"
    },
    {
      "label": "Talk to agent",
      "value": "connect me to a human agent",
      "icon": "user"
    }
  ]
}

Quick reply properties

Property	Description	Required
label	Text displayed on the button	Yes
value	Message sent when clicked	Yes
icon	Icon displayed with the button	No
style	Button styling variant	No

Styling quick replies

Quick reply buttons can be styled through Theme Admin:

Property	Description
Background color	Button background
Text color	Button label color
Border	Border style and color
Border radius	Corner rounding
Padding	Internal spacing

Fallback handling

Configure behavior when no intent matches the user’s message.

Fallback options

AI processing

Forward unmatched messages to the AI agent:

{
  "fallback": {
    "type": "ai",
    "agentId": "default-agent"
  }
}

This is the default behavior - unmatched messages are processed by the configured AI agent.

Default response

Display a default response:

{
  "fallback": {
    "type": "response",
    "message": "I'm not sure I understand. Please try one of these options:",
    "showQuickReplies": true,
    "quickReplies": [
      {"label": "Account balance", "value": "check balance"},
      {"label": "Help", "value": "help"}
    ]
  }
}

Escalation

Escalate to a human agent or different workflow:

{
  "fallback": {
    "type": "escalate",
    "workflow": "human-handoff",
    "message": "Let me connect you with someone who can help."
  }
}

Example: Banking assistant intents

Here’s a complete example of intent configuration for a banking assistant:

{
  "intents": [
    {
      "name": "check-balance",
      "displayName": "Check Balance",
      "patterns": [
        {
          "type": "keywords",
          "values": ["balance", "how much", "account"]
        },
        {
          "type": "semantic",
          "examples": ["What's my balance?", "How much money do I have?"],
          "threshold": 0.8
        }
      ],
      "action": {
        "type": "workflow",
        "workflow": "get-account-balance"
      },
      "quickReplies": [
        {"label": "View transactions", "value": "show transactions"},
        {"label": "Transfer funds", "value": "transfer money"}
      ]
    },
    {
      "name": "view-transactions",
      "displayName": "View Transactions",
      "patterns": [
        {
          "type": "keywords",
          "values": ["transactions", "history", "statement"]
        }
      ],
      "action": {
        "type": "navigate",
        "target": "transactions-screen"
      }
    },
    {
      "name": "help",
      "displayName": "Help",
      "patterns": [
        {
          "type": "exact",
          "values": ["help", "?", "what can you do"]
        }
      ],
      "action": {
        "type": "response",
        "message": "I can help you with:\n- Checking your account balance\n- Viewing transactions\n- Transferring money\n- Finding branch locations\n\nWhat would you like to do?"
      },
      "quickReplies": [
        {"label": "Check balance", "value": "check balance"},
        {"label": "Transactions", "value": "view transactions"},
        {"label": "Transfer", "value": "transfer money"}
      ]
    }
  ],
  "fallback": {
    "type": "ai",
    "message": "Let me think about that..."
  }
}

Best practices

Intent design

Keep intents focused

Each intent should handle one specific user need

Use clear patterns

Make patterns specific enough to avoid false matches

Provide quick replies

Guide users with suggested actions after responses

Test thoroughly

Test with various phrasings to ensure reliable matching

Pattern configuration

Start with keyword matching for common phrases
Add semantic matching for more natural interactions
Use regex sparingly for complex patterns
Set appropriate thresholds for semantic matching

Fallback handling

Always configure a fallback behavior
Consider user experience when no intent matches
Use AI fallback for natural conversation flow
Provide escape routes (e.g., “talk to human”)

Troubleshooting

Intent not triggering

Possible causes:

Pattern not matching user input
Priority conflict with another intent
Semantic threshold too high

Solutions:

Review and test patterns with actual user inputs
Check intent priorities
Lower semantic matching threshold
Add more keyword variations

Wrong intent triggered

Possible causes:

Overlapping patterns between intents
Semantic matching too broad
Priority configuration incorrect

Solutions:

Make patterns more specific
Increase semantic matching threshold
Adjust intent priorities
Use exact matching for specific commands

Quick replies not showing

Possible causes:

Quick replies not configured
Display mode not supporting quick replies
Theme configuration hiding quick replies

Solutions:

Verify quick reply configuration
Check display mode compatibility
Review theme settings for quick reply visibility

Integration points

Chat Intents integrates with multiple FlowX.AI components to enable sophisticated conversational experiences.

With Chat Component

Integration	Description
Intent detection	Analyze incoming messages for intent classification
Response routing	Route responses based on classification results
Context passing	Pass conversation context to target workflows

With UI Flows

Integration	Description
Intent-triggered navigation	Navigate to specific screens based on detected intent
State preservation	Maintain UI state during navigation transitions
Summary prompt support	Display conversation summaries during navigation

With Workflows

Integration	Description
Workflow initiation	Start workflows based on detected intents
Data mapping	Map chat context data to workflow input parameters
Response handling	Return workflow results back to the chat interface

Use cases

Customer support routing

Route customer inquiries to appropriate support workflows:

User describes their issue in chat
Intent classification identifies the support category
Conversation routed to specialized support workflow
Escalation triggered based on intent confidence or keywords

Multi-step process guidance

Guide users through complex workflows conversationally:

User expresses intent to complete a task (e.g., “I want to apply for a loan”)
Intent triggers the appropriate onboarding workflow
Chat guides user through each step with context-aware prompts
Navigation to relevant UI screens as needed

Conversational forms

Complete forms through natural conversation:

Intent classification identifies the form type needed
Chat collects required fields through conversation
Smart suggestions based on user context
Validation and confirmation through chat

Chat Component

Learn about the Chat component

Integration Workflows

Configure workflows for intent actions

UI Actions

Configure UI actions

Knowledge Base

Ground AI responses with knowledge bases

Config-time agents

Agent Builder

Using agents

​Overview

Intent classification

Dynamic routing

UI Flow navigation

Context preservation

​Related features

​How intents work

​Intent processing flow

​Configuring intents

​Accessing intent configuration

​Intent properties

​Trigger patterns

​Pattern types

​Combining patterns

​Intent actions

​Action types

​Quick replies

​Quick reply configuration

​Quick reply properties

​Styling quick replies

​Fallback handling

​Fallback options

​Example: Banking assistant intents

​Best practices

​Intent design

Keep intents focused

Use clear patterns

Provide quick replies

Test thoroughly

​Pattern configuration

​Fallback handling

​Troubleshooting

​Integration points

​With Chat Component

​With UI Flows

​With Workflows

​Use cases

​Customer support routing

​Multi-step process guidance

​Conversational forms

​Related resources

Chat Component

Integration Workflows

UI Actions

Knowledge Base

Overview

Related features

How intents work

Intent processing flow

Configuring intents

Accessing intent configuration

Intent properties

Trigger patterns

Pattern types

Combining patterns

Intent actions

Action types

Quick replies

Quick reply configuration

Quick reply properties

Styling quick replies

Fallback handling

Fallback options

Example: Banking assistant intents

Best practices

Intent design

Pattern configuration

Fallback handling

Troubleshooting

Integration points

With Chat Component

With UI Flows

With Workflows

Use cases

Customer support routing

Multi-step process guidance

Conversational forms

Related resources