Note: This is a Chatbot I developed as a demo. Neither the Chatbot nor I have any official connection with Haloview and its fine products.

The Demo

Demo Guide

Step 1 (Optional): In the small box, click “Generate Scenario” to generate your fictional use cases for the chatbot.

Step 2: In the large box, type “hi” to activate the Chatbot.

The Problem

While explaining my work with conversational AI to my dear and lovely mother, she said she wishes Haloview had a chatbot because she’s been trying to troubleshoot a problem with one of their products. Haloview makes rear view camera systems for RVs, trucks, and other large vehicles.

The Solution

I looked up the Haloview website and found they have the technical documentation for all their products available for download. This made them fertile ground for me to design a demo Chatbot based on their product line.

The Process

I used BotPress because it hit the sweet spot for me. It’s easy to start, but allows plenty of advanced features for future growth.

The bot I developed for Haloview uses seamless mix of human-written text and Generative AI. I use customer-facing generative AI to converse with the user, and also use generative AI to read the conversation and intelligently transition into correct workflows.

In addition, it relies heavily on retrieval-augmented generation (RAG). I created two databases. The sales flow (“I want to learn about a product”) uses an excel sheet I created of Haloview product names, prices, features, and likely use cases.

The second database is for the tech support flow (“I have a problem”). This database is full of every Haloview manual available for download on their website.

The Limitations

I have no official relation with Haloview. Therefore, the sales flow is unable to connect via API to their purchasing system. Instead, I just have placeholder text where the final purchasing flow would be.

In addition, Haloview has email support but no chat support. Therefore, at the moment my Chatbot links to their support page when a user wants to escalate. In the future if I can connect with Haloview, I plan to implement direct email messaging from within the Chatbot and–ideally–eventually direct chat escalation to human support.

Some of the data in the sales flow is dummy data. My goal was to show the chatbot’s functionality as a proof of concept, not for every piece of data to be accurate. Therefore some information in my chatbot such a as price, functionality, and product use cases may not accurately reflect Haloview products. At least, not yet.

Finally, BotPress and my WordPress website don’t seem to be playing along very nice. You must manually scroll down when the conversation gets too long. This is a WordPress limitation, not a limitation of the Chatbot. Strangely, this actually seems to work better on mobile than on desktop. I don’t know, I’m a conversational AI designer, not a frontend web developer 😅

The Design

Often new technology is difficult to learn. Generative AI, when used well, can be different. My guiding philosophy with this chatbot was to use Generative AI to make a chatbot experience both more user friendly while also being more helpful than would ever be possible without this technology.