Get started
FAQThe AI AgentUnderstanding Your AI Agent

Understanding Your AI Agent

Your AI agent is the engine behind every customer conversation on WhatsApp. This guide explains how it works, what it can and can't do, and how to get the most out of it.

How does the AI agent understand what customers want?

The agent uses a large language model (LLM) to read and interpret natural language. When a customer says "do you have a medium blue hoodie?" the agent searches your product catalogue, finds the closest match, and responds with the relevant options — all without any manual input from you. It understands context across multiple messages, so a customer can refine what they want across several turns in the conversation.

What tools does the agent have access to?

Your agent can:

  • Search your product catalogue by name, description, category, or attribute
  • Add, remove, and adjust items in a customer's cart
  • Create and confirm orders
  • Send payment links
  • Process and track refund requests
  • Accept one-time and recurring donations (charity vendors)
  • Escalate a conversation to a human team member when needed

What happens if the agent can't answer a question?

If a customer asks something outside the agent's knowledge, such as a very specific delivery question you haven't configured yet, the agent will acknowledge that it can't help and will either ask the customer to contact you directly or escalate the conversation to your team for human follow-up.

Can the agent handle multiple customers at once?

Yes. The agent handles every incoming conversation independently and simultaneously, so there is no queue or delay for customers regardless of how busy your WhatsApp number gets.

What is human escalation and when does it trigger?

Human escalation is when the agent flags a conversation for a real team member to take over. It triggers when:

  • A customer explicitly asks to speak to a human
  • The agent encounters a query it cannot confidently resolve
  • A complaint or sensitive issue is detected

Escalated conversations appear in the Conversations part of your dashboard so you can follow up promptly.

Are there any guardrails on what the agent will say?

Yes. The agent is configured to stay focused on your business. It will not engage with off-topic requests, share information about competitors, or make commitments your business hasn't authorised. These guardrails are enforced at the model level and cannot be bypassed by customers.

Can I see what the agent said to a customer?

Yes. Every conversation is stored in your dashboard in the Conversations page. You can review the full message thread for any customer at any time, and pause the AI to reply yourself.

← Back to all FAQs

Still have questions?

Reach out and we'll get back to you — or just start selling and see how it works.