# AI Intent Detection in ZazzyAgent | Complete Guide

Customers don't always use the exact words your automation expects.

You might create a keyword trigger for:

> order status

but customers could say:

> Where is my package?

> Has my order shipped?

> Can you check my delivery?

These messages have the same underlying goal even though the wording is different.

That's where **AI Intent Detection** can help.

## What is AI Intent Detection?

AI Intent Detection identifies the customer's underlying intention from their message.

For example:

**Customer says:**

> I need to know where my order is.

**Detected intent:**

> Order Status

You can then connect that intent to the appropriate automation or action.

The current AI Assistant functionality includes Intent Detection as a separate configuration area. ([botsailor.com](http://botsailor.com))

## Intent vs keyword

A keyword looks for a word or phrase.

An intent looks for meaning.

### Keyword

> refund

Customer:

> refund

Trigger fires.

### Intent

**Refund Request**

Possible customer messages:

> I want a refund.

> Can I get my money back?

> I need to return this and get a refund.

> How do I request a refund?

The intent can capture different ways of expressing the same request.

## When should I use AI Intent Detection?

Use it when customers can ask for the same thing in many different ways.

Good examples:

*   Order tracking
    
*   Refund request
    
*   Product recommendation
    
*   Pricing enquiry
    
*   Appointment request
    
*   Human support
    
*   Product availability
    
*   Complaint
    
*   Cancellation request
    

For simple commands such as:

> start

a keyword trigger may be enough.

## Create an intent

Open the AI Intent Detection area in ZazzyAgent.

Create a new intent.

Give it a clear name.

Good:

> Order Status

> Product Recommendation

> Refund Request

> Human Support

Avoid:

> Intent 1

> New Intent

## Add examples

Give the intent examples of what customers might say.

For:

**Order Status**

Examples could include:

> Where is my order?

> Can you check my delivery?

> Has my order shipped?

> What's happening with my package?

The goal is to represent the ways real customers might express the intention.

## Connect the intent to an action

Once the intent is recognized, your automation can take the appropriate next step.

Examples:

**Order Status**

→ Order lookup

**Refund Request**

→ Support workflow

**Product Recommendation**

→ Product recommendation action

**Human Support**

→ Human assignment

## Use intents to trigger flows

An intent can be used to send a customer into an appropriate automation path where supported.

Example:

**Intent: Appointment Request**

↓

**Appointment Flow**

↓

Collect:

*   Service
    
*   Date
    
*   Time
    

## Use intents with labels

You can also use intents to organize customer information.

Example:

Customer expresses purchase interest.

↓

Add:

**High Intent**

label.

This makes later segmentation and follow-up easier.

## Use intents with sequences

An intent can also be used to start an appropriate follow-up sequence where supported.

Example:

**Intent: Interested in Demo**

↓

Subscribe to:

**Demo Follow-up**

## Intent design mistakes

### Mistake 1: Creating overlapping intents

Avoid:

> Order Problem

> Order Issue

> Order Trouble

These may describe almost the same thing.

Create distinct intentions.

For example:

**Track Order**

vs

**Cancel Order**

vs

**Return Order**

Those represent different actions.

### Mistake 2: Too few examples

Customers don't all talk the same way.

Add realistic variations.

### Mistake 3: Making intents too broad

Avoid an intent called:

> Customer Request

That's not useful.

Make the purpose specific.

## Intent detection vs AI Agent

Intent Detection and AI Agents serve different purposes.

Intent Detection can identify what the customer wants and route the conversation.

An AI Agent can conduct the actual conversation and take actions according to its instructions.

They can work together.

## Intent troubleshooting

### Wrong intent detected

Add more realistic examples to the intended category.

Make overlapping intents more distinct.

### Intent doesn't trigger

Check:

*   Intent is active/configured
    
*   Example phrases accurately represent the intent
    
*   Customer message actually expresses the intended meaning
    
*   Correct automation is connected
    

### Two intents overlap

Review both sets of examples.

Ask:

> Can I easily explain the difference between these two intents?

If not, they're probably too similar.

## A practical intent structure

For a typical ecommerce business:

**Product Recommendation**

→ Sales Agent

**Order Status**

→ Order Support

**Return Request**

→ Support

**Complaint**

→ Human Support

**Human Request**

→ Human Agent

This gives natural customer language a clear path into your automation.

## The key idea

Keyword automation answers:

> "Did the customer say this word?"

Intent detection answers:

> "What is the customer trying to do?"

That difference becomes very valuable as your conversations become more complex.

[zazzyagent.com](http://zazzyagent.com)