To give the public the magic they are craving, we as Botmakers need to do a lot better.
We need to build bots that actually solve our users’ problems. Bots that work quickly, effectively, and don’t force our users to learn complex commands or memorize phrases.
We need bots that seem like magic.
To up us get there, I propose this challenge to the Botmaker community:
Make chatbots that perform just as well as humans.
Humans default to conversation because it is the primary way we have learned operate in the world. However, we only use conversation with other entities (humans mostly) of similar capabilities, who can understand what we are communicating quickly and without explicitly stating context.
When the entity you are conversing with does not understand you as well as most of the other entities you converse with, you get frustrated or even angry.
Right now, humans turn to a conversational channel because dealing with machines is confusing, complex or broken. When the machine does not understand them, or does not provide what they are looking for, they just want to ask a human for help.
If you send users from a GUI based machine that confuses them to a conversational machine that barely understands them, you just frustrate your users even more.
But when your chatbot can match or even outperform humans, it will be magical to the user.
This does not necessarily mean that a bot needs to pass the Turing Test; the bot only needs to perform well in the bot’s area of expertise.
If a human can answer a user query with no follow up question, so should your bot. If a human needs further information, then your bot is allowed to request more.
It is a simple concept, but hard to implement.
So I’ve created a practical set of guidelines to help you build chatbots that perform less like machines, and more like humans.