Turn your website into a chatbot with IBM Watsonx.ai Agent Lab

Introduction

4 min readFeb 11, 2025

Websites are filled with valuable information, but finding the right answers quickly can be a challenge for visitors. Instead of making users search through pages of content, what if your website could instantly respond to their questions like a smart assistant? With Watsonx.ai Agent Builder, you can easily turn your website into an interactive chatbot that understands queries and provides relevant answers. This means visitors can get the information they need in seconds without scrolling endlessly or clicking through multiple links.

Traditionally, building such a system would require complex tools for web crawling, indexing, and creating a retrieval-augmented generation (RAG) application. However, Watsonx.ai simplifies this process by allowing you to convert your website’s existing content into a chatbot without extra development work. By leveraging AI, your website can engage users, improve customer experience, and maximize the value of your online information — all while saving time and effort in building custom solutions from scratch. In this article, we’ll explore how Watsonx.ai Agent Builder makes this transformation easy and beneficial for businesses of all sizes.

Step 1

Agent builder is integrated within the watsonx.ai runtime environment. The user can easier create a new asset and choose to create a new agent. The user inside the project clicks on “New Asset” button and then will choose the option to “Build an AI agent to automate tasks”; This will bring the user to the “Agent Builder User Interface”

Step 2

The user Selects a Large Language Model amongst supported by the platform or can bring their own model. Enter a name and description of the agent, and may be couple of sample questions that the agent can answer to guide users who are not familiar with the agent’s functionality.

STEP 3

Then we Add instructions, in this case we want to turn a website into a chatbot. For the sake of demo I am going to use my favourite website https://www.bcferries.com/ , and my instruction will be

You are a helpful assistant that uses tools to answer questions in detail. your job is to crawl bc ferries website : https://www.bcferries.com and answer questions regarding content in this website
When greeted, say “Hi, I am bc ferries agent. How can I help you?”

STEP 4

Next we will add a a tool; there are various tools available as well as custom tools. For this use case I will click on “Add a tool” and then select the “Web crawler”; if I want I can create my owns custom tool by providing the python code, input signature by clicking on the “create custom tool” button;

Some of the available out of the box tools are :

Google Search
Document Search
Wikipedia Search
DuckDuckGo Search
WebCrawler
Python Interpreter
Weather

Step 5

As simple as that, I am done. I can now test my agent on “Agent Preview” on the right side of the page, select one of the questions I added earlier when I was creating the agent. As shows in the image below we can see the right answer as highlighted in the response panel.

I can take this further. I can make this available easily without writing a single line of code.

Step 6

I can save my agent and choose to deploy it in deployment space.

Step 7

User clicks on deploy button and can deploy the agent as an endpoint or AI Service

Step 8

End points are ready; we have private, public end points as well as sample code in various programming language to integrate this service within other tools or applications.