Built-in Tools
Your AI agent comes with built-in capabilities that allow it to take actions during conversations. You don't need to configure anything — simply describe what you want the agent to do in your Custom Instructions.
Overview
Built-in tools are actions your agent can perform automatically based on your instructions. Write in natural language — the agent figures out which tool to use.
| Capability | What It Does | Requires |
|---|---|---|
| Search knowledge base | Find answers from your FAQs and content | Knowledge base content |
| Search products | Find products matching customer criteria | Product integration |
| Show product carousel | Display products with images and links | Product integration |
| Show image carousel | Display custom cards with images and buttons | — |
| Send media | Share images, videos, documents | — |
| Save customer info | Update customer profile and tags | — |
| Handover | Transfer to human or another agent | Handover reasons configured |
| Resolve | Mark conversation as complete | — |
Searching Information
Knowledge Base Search
Your agent can search through your knowledge base to find answers to customer questions.
How to instruct your agent:
Before answering questions about our policies, search the knowledge base first.
When customers ask about shipping or returns, look up the information before responding.
If you don't know the answer, search our FAQ before saying you don't know.
Product Search
This capability requires products in your catalog — either manual products, Shopify, or Zid.
Your agent can search your product catalog to find items matching customer requests. It can filter by:
- Text descriptions (semantic search)
- Categories
- Brands/vendors
- Price ranges
- Product attributes (size, color, etc.)
How to instruct your agent:
When customers ask about products, search our catalog to find matching items.
If a customer mentions a budget, filter products within their price range.
Help customers find products by asking about their preferences, then search accordingly.
Sending Rich Content
Product Carousels
This capability requires a product catalog.
After searching for products, your agent can display them in an interactive carousel. Each card shows the product image, a caption you specify, and a button linking to the product page.
How to instruct your agent:
After finding products, show the top 3-5 matches as a carousel.
When recommending products, display them visually so customers can see the images.
Present product options in a carousel format with prices in the captions.
Limitations:
- Maximum 10 products per carousel
- Products must have images to display properly
Image Carousels (Custom)
Your agent can send custom image carousels with links — perfect for promotions, menus, or any content where you want to show images with clickable buttons.
Each card can include:
- Header text (title)
- Footer text (subtitle/description)
- Image
- Button with link
How to instruct your agent:
When showing our service packages, display them as a carousel with images and "Learn More" buttons.
Present menu options visually with images and links to each section.
Best for:
- Service packages or pricing tiers
- Menu items or categories
- Promotional content with external links
- Any content where your API tool returns image URLs and links
Example scenario: Your external API returns data like:
{
"packages": [
{ "name": "Basic", "image": "https://...", "link": "https://...", "price": "$99" }
]
}
Instruct your agent:
Display packages as a carousel showing the package name, price, and image with a button to view details.
Media Messages
Your agent can send images, videos, audio files, and documents directly in the conversation.
How to instruct your agent:
If you have a relevant image or document, send it to the customer.
When explaining installation, send the instruction manual PDF.
Share product images when discussing specific items.
Platform differences:
| Platform | Behavior |
|---|---|
| Each file sent as a separate message | |
| Multiple images may be grouped together |
See WhatsApp Features and Instagram Features for platform-specific details.
Managing Customer Information
Saving Customer Details
Your agent can save information customers share during the conversation to their customer profile:
- Name
- Email address
- Phone number (with country code)
- Address
- Country
- Language preference
- Profile notes (temperament, preferences, interests)
- Custom tags
- Any custom attributes you've defined
How to instruct your agent:
When customers share their contact information, save it to their profile.
After learning the customer's name, update their profile.
Tag customers based on their interests or the products they're asking about.
Note the customer's preferences in their profile for future reference.
Notes:
- The agent will only save information explicitly shared by the customer
- Phone numbers should include country codes
- Language is detected automatically from the conversation
Customer Tags
You can define tags in your system, and the agent can apply them based on the conversation.
How to instruct your agent:
If a customer is interested in bulk orders, tag them as a wholesale lead.
Tag VIP customers when they mention they're repeat buyers.
Apply the "needs follow-up" tag when a customer says they'll think about it.
Conversation Management
Handover to Human
Your agent can transfer conversations to human team members when needed. You define the handover reasons (e.g., "Technical Support", "Billing Issue", "Complaint").
How to instruct your agent:
If the customer explicitly asks to speak with a human, confirm first then hand over.
Transfer to technical support if the issue requires troubleshooting you can't handle.
Hand over billing-related complaints to the billing team.
You can also route handovers directly to another AI agent by linking a handover reason to a specific agent. See Agent-to-Agent Handover.
Important: Configure handover reasons in your agent settings. The agent will include a summary for the human agent receiving the chat.
Resolving Conversations
Your agent can mark conversations as resolved when they're complete.
How to instruct your agent:
Resolve the conversation when the customer says thank you and goodbye.
If the customer confirms they have no more questions, close the chat.
Don't resolve if the customer seems unsatisfied or has pending questions.
Writing Effective Instructions
Add these instructions to your agent's Custom Instructions.
Be Specific About Triggers
Instead of vague instructions, be clear about when actions should happen:
| Less Effective | More Effective |
|---|---|
| "Help with products" | "When customers ask about products, search the catalog and show matching items" |
| "Use carousels" | "Display product recommendations as a carousel after searching" |
| "Save information" | "When customers share their email or phone, save it to their profile" |
Describe the Flow
Explain the sequence of actions:
When a customer asks about a product:
1. Search for matching products in our catalog
2. If found, show the top 3 options as a carousel
3. Ask if they'd like more options or different filters
4. If nothing matches, apologize and offer to help them find alternatives
Set Conditions
Tell the agent when NOT to do something:
Only hand over to a human after:
- Confirming the customer wants to speak with someone
- Attempting to resolve their issue first
Don't resolve conversations if:
- The customer seems frustrated
- There are unanswered questions
- The customer said they'll "be back"
Use Natural Language
You don't need technical terms. Write instructions as if you're training a human employee:
When customers ask about our return policy, look it up in our knowledge base
and share the relevant information. If they want to start a return, collect
their order number and email, then hand over to the returns team.
Common Scenarios
E-commerce Agent
Help customers find products by asking about their needs. Search our catalog
and show matching items as a carousel. Save their preferences to their profile.
If they want to place an order and we can't process it through chat, hand over
to sales.
Customer Support Agent
When customers have questions, search our knowledge base first. If you can't
find the answer or the issue is complex, hand over to support. Always save
any contact information they share. Resolve when their issue is addressed.
Lead Generation Agent
Collect customer information during the conversation - name, email, phone,
and their interests. Tag them based on what they're looking for. If they're
a serious buyer, hand over to sales. Otherwise, make sure their profile is
complete before resolving.
Next Steps
- Creating API Tools — Connect to external APIs for custom capabilities
- Agent Personality — Configure Custom Instructions
- Handover Settings — Set up handover reasons and routing