# How AI Agents Save Customer Information in ZazzyAgent A useful AI Agent should not have to ask the same question repeatedly. When a customer provides information that will be useful later, you can save it into a custom field. Examples: * Name * Email * City * Product interest * Budget * Order number * Preferred date * Company name The AI Agent can be instructed to save that information during the conversation. ## The custom-field action The action format is: ```text ##save_custom_field## : 42 ``` The number represents the selected custom field. Inside the System Prompt editor, type: ```text ## ``` Then select the action and the correct custom field from the autocomplete list. The interface uses the field's name when you select it; the stored prompt uses its internal ID. ## Example: Save a customer's city Suppose the agent asks: > Which city are you based in? Customer: > Mumbai You can instruct the agent: ```text When the customer clearly provides their city, save that answer in the City custom field. ##save_custom_field## : 42 ``` Select your actual City field using autocomplete. ## Example: Save product interest ```text When the customer identifies the product category they are interested in, save it to the Product Interest custom field. ##save_custom_field## : 18 ``` Now the customer record can contain: **Product Interest = Running Shoes** ## Example: Save budget ```text When the customer clearly provides their approximate budget, save the amount to the Budget custom field. ##save_custom_field## : 27 ``` If the customer says: > Around ₹40,000. Your workflow can store that information according to the custom field's configuration. ## Tell the AI exactly what should be saved This is critical. Don't write: > Save the customer's information. That's too vague. Tell the agent: > Save the customer's preferred city to the City custom field. or: > Save the customer's approximate budget to the Budget custom field. The clearer the instruction, the easier it is to control the stored data. ## Don't save information before the customer provides it Bad instruction: > Save the customer's budget. The customer hasn't provided a budget yet. Better: ```text Ask the customer for their approximate budget. After the customer provides a clear answer, save the answer to the Budget custom field. ##save_custom_field## : 27 ``` ## Don't ask twice Your prompt should tell the AI to reuse information already provided. Example: ```text If the customer has already provided their city, do not ask for their city again. Use the existing customer information when appropriate. ``` This makes conversations feel much more natural. ## Save order numbers Order numbers are particularly useful for ecommerce support. Example: Customer: > My order number is 12345. Your agent can be instructed: ```text When the customer provides their order number, save it to the Order Number custom field. ##save_custom_field## : 35 ``` You can then use that information in other workflows where supported. ## Save lead information A lead-qualification agent might save: **Name** **Company** **Requirement** **Budget** **Timeline** Then your sales team can see the information without reading the entire conversation. ## Save appointment information An appointment agent could save: **Service** **Preferred Date** **Preferred Time** This allows the information to be reused later in the conversation or passed to another system where supported. ## Combine saving information with other actions A common AI workflow is: **Customer provides information** ↓ **Save custom field** ↓ **Add label** ↓ **Start sequence** For example: ```text When the customer confirms that they are interested in the Enterprise plan, save Enterprise as the Product Interest value. ##save_custom_field## : 18 Add the Qualified Lead label. ##add_label## : 12 Start the appropriate sales follow-up sequence. ##assign_sequence## : 8 ``` Each action has a different purpose. ## Custom field vs label Use a custom field when you need the actual value. > Budget = ₹40,000 Use a label when you need a simple classification. > High Intent Both can be used together. ## What if the customer changes their answer? Suppose the customer first says: > Mumbai and later says: > Actually, I moved to Pune. Your System Prompt should explain when the newer information should replace the old value. Example: ```text If the customer corrects or updates their city, save the new city value to the City custom field. ``` ## Don't store guesses The AI should never invent values. Use rules such as: ```text Only save a value when the customer clearly provides it. If the customer's answer is ambiguous, ask a follow-up question before saving it. ``` ## Use saved information later Once information is stored, supported ZazzyAgent features can use that information for: * Conditions * Personalization * Segmentation * Broadcast targeting * Other automation * External integrations ## Test your custom fields Test: ### Clear response > I'm based in Pune. Expected: **City = Pune** ### Ambiguous response > Somewhere near Mumbai. The agent should clarify if an exact city is required. ### Correction > Actually, I moved to Delhi. The field should be updated if your prompt says to accept corrections. ### Missing response > I don't want to share that. Don't save a made-up value. ## Common problems ### Field remains empty Check that: * The custom field exists. * The correct field was selected. * The customer actually provided the information. * The prompt tells the agent when to save it. ### Wrong value was saved Make the extraction rule more precise. ### Agent keeps asking the same question Tell the agent to use already-known customer information before asking again. ## A useful pattern The best instruction is usually: ```text Ask → Confirm → Save ``` Example: > What city are you based in? Customer: > Pune. > Just confirming, you're based in Pune? Then: ```text ##save_custom_field## : 42 ``` You don't always need the confirmation step, but it can be valuable for important information. The purpose is simple: **Turn useful conversation information into structured customer data.**