Skip to main content
Most bots are menus: press 1 for this, press 2 for that. Paige’s AI bot template gives your bot something better — it reads what a customer actually wrote, works out what they want, and responds in their own words. This is the AI that runs inside your deployed bot, on your customer’s WhatsApp, around the clock. It runs on your own OpenAI API key (GPT-4o for chat, Whisper for voice notes) — a separate account and a separate bill that never touches your Paige credits or billing. It is not one of the builder agents you chat with while you’re making the bot. Ask the Code agent for an AI bot and it scaffolds this runtime AI for you.
This is not the Conversations agent. They’re easy to confuse and have nothing to do with each other:
  • The AI in your bot (this page) is a runtime bot feature. It talks to your customers on WhatsApp, is part of the bot you deploy, and runs on your own OpenAI key.
  • The Conversations agent is a builder agent. It talks to you, in the Paige sidebar, to help you work your inbox — finding contacts and drafting replies. Its usage draws on your Paige credits, like the other builder agents.

​
It’s your code, not a Paige feature

This matters, because it’s what makes the AI bot flexible. The AI conversation isn’t a service Paige runs on your behalf and lets you configure around the edges — it’s ordinary code in your project, which the Code agent wrote and which you can change at any time by asking. The prompts that decide your bot’s personality, the rules for what it will and won’t discuss, how it decides to escalate — all of it is yours.
Because the AI bot template calls OpenAI, it needs an OpenAI API key stored as a secret named OPENAI_API_KEY. That key is billed by OpenAI directly — it’s entirely separate from Paige’s credits, which pay for the builder agents that build your bot rather than your bot talking to customers.

​
What it does

Understands intent

Reads each incoming message, figures out what the user wants, and routes the conversation to the right next step — whether that’s answering a question, starting a flow, or escalating.

Stays in character

Replies in the tone and voice you’ve configured for your brand. Stays focused on the topics your bot is meant to handle and politely redirects off-topic messages.

Pulls from your knowledge

Answers questions using the knowledge base documents you’ve added to the project — product info, FAQs, policies, and anything else you want it to reference.

Collects information naturally

Asks for the details your bot needs (name, email, preferred date, etc.) without forcing users through rigid forms. Validates and stores answers as it goes.

Hands off cleanly

Knows when to step out of free-form chat and start a structured WhatsApp Flow, send a template, or hand the conversation to a human teammate.

Remembers context

Keeps track of where each user is in their conversation, so a user who returns hours or days later picks up exactly where they left off.

​
How to shape it

You don’t program the AI in your bot directly. You shape its behavior by talking to the Code agent — for example:
“Make the bot greet users in Afrikaans and English, and only answer questions about our restaurant menu and bookings. For anything else, suggest they call us instead.”
The Code agent updates the bot’s prompts, knowledge base, and routing so the AI in your bot behaves the way you described. You can iterate as often as you like — every conversation that happens after a deploy uses the new behavior.
Add a knowledge base document for any topic you want the AI in your bot to answer accurately. Bullet points and short FAQs work better than long marketing pages.

​
When to use it

The AI bot template shines whenever your bot needs to handle messages that aren’t predictable button taps:
  • Customer support — answering FAQs, troubleshooting, and routing tricky questions to a human.
  • Lead qualification — asking the right follow-up questions based on what the user just said.
  • Discovery and search — helping users find a product, service, or piece of information by describing it.
  • Onboarding — walking a new user through setup with conversational prompts instead of static menus.
If your bot only needs strict menu-driven logic, you can opt out of the AI in your bot entirely — see the simple bot template described on the Code agent page.