# How to Add and Remove Customer Labels With ZazzyAgent AI Agents Labels are a simple way to keep track of what is happening with a customer. An AI Agent can automatically add or remove labels based on what a customer says or does. For example: > Customer says they are ready to buy. → Add **High Intent** Or: > Customer completes their purchase. → Remove **High Intent** This lets your customer database change automatically as conversations progress. ## Why use labels with AI? Without automation, your team may need to read conversations and manually classify customers. An AI Agent can recognize useful customer signals and update the customer record. Examples: * New Lead * Qualified Lead * High Intent * Pricing Interested * VIP * Support Required * Callback Required * Interested in Product A ## Add a label The action format is: ```text ##add_label## : 12 ``` The number represents the selected label's internal ID. You don't need to memorize or guess the number. Inside the System Prompt editor, type: ```text ## ``` Then select the label action and the label you want to use. The editor shows the readable label name while the saved prompt contains the internal ID. ## Example: Add a high-intent label Suppose a customer says: > I want to place the order today. You could instruct the agent: ```text When the customer clearly confirms that they are ready to purchase, add the High Intent label. ##add_label## : 12 ``` The exact ID will depend on your account. Use the autocomplete to select the correct label. ## Remove a label The removal action is: ```text ##remove_label## : 12 ``` Example: ```text When the customer completes their purchase, remove the High Intent label. ##remove_label## : 12 ``` ## Always define the condition Don't give the agent an action without explaining when it should happen. Weak: ```text ##add_label## : 12 ``` Better: ```text When the customer clearly expresses strong purchase intent, add the High Intent label. ##add_label## : 12 ``` The instruction provides the decision logic. The action performs the change. ## Example: Lead qualification Suppose your AI Agent is qualifying leads. It asks about: * Requirement * Budget * Timeline Once the customer meets your qualification criteria: ```text When the customer has provided all required qualification information and meets the criteria for a qualified lead, add the Qualified Lead label. ##add_label## : 12 ``` ## Example: Product interest A customer says: > I'm looking specifically for running shoes. Your agent could add: ```text ##add_label## : 25 ``` with an instruction such as: ```text When the customer clearly expresses interest in running shoes, add the Running Shoes Interest label. ##add_label## : 25 ``` ## Example: Support requirement Customer: > I need someone from your support team to call me. The agent could add: ```text ##add_label## : 31 ``` with: ```text When the customer requests human support or a callback, add the Callback Required label. ##add_label## : 31 ``` You can then use that label for staff workflows or targeted follow-up. ## Use labels to represent changing customer states A useful pattern is: **New Lead** ↓ **Qualified Lead** ↓ **High Intent** ↓ **Customer** Don't keep all of those labels forever. When the customer's state changes, remove outdated labels and add the new one. ## Example: Moving a lead through the funnel ```text When the customer becomes qualified, remove the New Lead label. ##remove_label## : 10 Add the Qualified Lead label. ##add_label## : 12 ``` Later: ```text When the customer confirms they are ready to buy, remove the Qualified Lead label. ##remove_label## : 12 Add the High Intent label. ##add_label## : 15 ``` ## Use labels with broadcasts Once customers are classified, labels can help create targeted audiences. For example: > Send the new-product campaign to customers labelled VIP. This gives your marketing team a much more useful audience than sending every campaign to everyone. ## Use labels with sequences A label can also be used as part of a follow-up process. Example: **Customer expresses interest** ↓ Add **Interested Lead** ↓ Start **Sales Follow-up** The label provides the customer state while the sequence handles the follow-up. ## Don't label based on assumptions This is especially important with AI. Don't instruct: > Add High Intent when the customer asks about the product. A product question does not necessarily mean the customer is ready to purchase. A better rule is: > Add High Intent when the customer clearly confirms a strong intention to purchase or asks to proceed with the purchase. ## Don't create too many labels Use labels only when they have a practical purpose. A label should help answer a useful question: > What is important about this customer right now? ## Test your label actions Try several conversations. ### Test 1 > Can you tell me about your products? No purchase-intent label should be added. ### Test 2 > This looks good. I want to order it. High Intent should be added. ### Test 3 > I've placed the order. The previous purchase-intent label may need to be removed and a customer label added. ## Common problems ### The wrong label is added Review the condition in the System Prompt. The AI may be interpreting a broad instruction too loosely. ### Label isn't changing Check that the action is configured with the correct label. Use `##` autocomplete rather than manually guessing the ID. ### Too many labels are being added Make your conditions more specific. For example: > Customer mentions the word "buy" is weaker than: > Customer clearly confirms they want to purchase. ## A good label strategy Think of labels as **customer state markers**, not as a transcript of everything the customer has ever said. Useful: > High Intent Less useful: > Asked About Price unless your business actually needs that information for a later workflow. Labels become much more powerful when they change automatically as the customer journey changes. [zazzyagent.com](http://zazzyagent.com)