Build-A-Bot 101: The Basic Bot Type Distinctions

Written by adam.ahmad | Published 2018/08/01
Tech Story Tags: bots | build-a-bot | bot-type-distinctions | basic-bot-type | bot-101

TLDRvia the TL;DR App

It seems like we can’t get enough of bots and the current and future applications for them. Even though it’s cool to speculate about the future, we thought it would be useful to bring it back to the basics and discuss what it takes to create a bot from the ground-up. That’s exactly what we’re doing at kea: building a conversational artificial intelligence system, which is basically a fancy way of saying a voice-automated bot. We recently did a complete overhaul on our kea voice bot to create a new AI personality and let us tell you- it takes a lot of decision making. How will people use the voice? What feelings should the voice inspire? What problems should it solve? And so on.

It all starts with building out the high-level properties of the bot. These are basically the mechanics behind the AI, the foundation upon which everything else is built. Here are some of the decisions that need to be made:

Personal, Team or Public Bot

When deciding on the characteristics of your bot, you have to think about what level it will be used on. Is an individual interacting with the automated system, like a person ordering lunch with kea’s phone ordering bot or a scheduling bot issuing a meeting reminder on Slack? Or will it be on a team level like an AI phone conferencing service, or a group vacation-planning bot. Some programs can even be used on both the personal and team level- like Slack. Whether your bot does one or both, this is typically the first decision and is embedded in the initial plan for the system.

Super Bot or Domain-Specific Bot

This question is centered around the function of the bot. A domain specific bot is geared towards a specific service, like Stella, a job hunting bot that helps employers find qualified candidates for open positions based on matching skill sets. General purpose means that it’s used in a number of ways by a number of audiences, like Google Assistant. It is one bot, but it can tell you the weather, find the nearest ramen shop, and much more. kea is a domain-specific bot as it’s job is to take phone orders on behalf of restaurants, it won’t also give you driving directions or the score from today’s baseball game.

Consumer or Business Focused Bot

Just like an at-home system that automatically alerts you if you left the garage door open, kea’s system is consumer-focused. Consumer bots can serve a wide variety of functions, they can help us automatically find discounts on our favorite shopping websites, like Honey, remind us that it’s time to exercise, or alert of us of travel deals to our bucket list destinations. Business-focused bots, like expense filing or workflow automation software, have totally different goals than consumer bots. Business bots are designed to get straight to the point, to get a task completed, as efficiently as possible. A consumer bot, however, has room to show a little personality: humor, self-reflection, wisdom and more.

Voice or Text Bot

The kea system is an obvious example of voice interface, using conversational AI to take phone orders for customers. There are other popular examples like Apple’s Siri, Amazon’s Alexa and Google Home. Take, for example, Alexa, she’s a voice bot who can change the temperature in your home, play your “Classic 80’s” playlist, and set a reminder to remove the turkey from the oven. There are also text bots, like the chatbots that help you on e-commerce websites when you need a return label or the text bots that show up in communication apps, like Slack, Facebook Messenger, and WeChat.

Legacy Integration or Novel System

kea’s system works with existing POS systems in restaurants, meaning it is integrated or retrofitted with the ticketing technology already employed by the restaurant. This is one example of a legacy integration bot, others include the Streak bot for G-mail, which inputs important information from chat conversations into its data management tools. Some bots like Pypestream are built from scratch, employing proprietary technology from top to bottom.

That’s enough tech talk for this week. Next week we will dive into the details of building a personality for a robot and not too long after we will introduce our new kea voice bot, which we’ve been working hard on for months. We can’t wait- chat(bot) soon!


Published by HackerNoon on 2018/08/01