What is a Knowledge Agent?
A Knowledge Agent helps users find answers from approved source material, such as website pages, FAQs, agendas, speaker profiles, exhibitor information and uploaded documents.
It can be used across channels such as website, email, WhatsApp or app-based experiences, depending on how it is configured.
Use this guide to create a Knowledge Agent, connect it to the right content sources, configure its behaviour, and prepare it for testing.
Understanding the three core components
Before creating a Knowledge Agent, it is essential to understand how its three core components work together:
Component | Role | Analogy |
|---|---|---|
Knowledge Agent (KA) | The agent that interacts with users across different channels. It takes questions, delivers answers, and manages conversations. | The Waiter – Serves the customer and manages the interaction. |
KA Engine | The "brain" that defines how the agent behaves—its tone, answer structure, intent detection, and fallback rules. | The Chef's Instructions – Dictates how to prepare and present the response. |
KA Collection | The "knowledge base" that defines what the agent knows—articles, PDFs, and web pages it reads to answer questions. | The Recipe Book – Contains all the ingredients and information. |
How They Work Together:
The Collection provides the raw information.
The Engine provides the rules and personality.
The Agent combines both to serve answers to users.
Multi-Channel Capability:
A single Knowledge Agent can be deployed across multiple channels simultaneously, ensuring consistent intelligence everywhere your audience interacts with you:
Website (chat widget)
Email
WhatsApp
Voice Caller (coming soon) 🚀
You can use the same Engine and Collection across all channels, or mix and match different combinations depending on the channel's specific needs.
Before You Begin
To create a Knowledge Agent, you will need:
A KA Engine – See the KA Engine Creation Guide
A KA Collection – See the Collection Creation Guide
Once both are ready, proceed with the channel-specific setup below.