Lock down your bot. Your users will thank you for it.
I’m a software engineer for Orderly Health. You can check out what we’re building by signing up and chatting with our bot Louie (it’s FREE). Please send your feedback and comments to jesse@orderlyhealth.com.
Interacting with an automated system is not a new concept. You’ve probably called a business before and had to navigate a phone menu by pressing digits on the keypad. Perhaps you were even brave enough to speak out loud in hopes of the service understanding what you were saying, or worse, you may have been forced to speak out when you would have preferred to keep quiet (pro tip: don’t make your password “Big Boy”).
Then along came chatbots, which are “intelligent,” AI-driven services. With a chatbot, you can ask a question just as you would a human, and if trained properly, it will understand your question and give you a meaningful response. It’s unlikely, however, that your experience with chatbots has been positive. How often have you tried to interact with a chatbot only to become frustrated when it misunderstood you?
When you think of a chatbot, you probably think of what’s called an “open” chatbot. An open chatbot is one that only accepts free-form text — that is, you talk to the chatbot as if you were talking to a human (hence, it is a very “open” interaction with no forced paths).
Digital assistants like Siri and Google Assistant are good examples of “open” chatbots.
On paper, keeping your chatbot open might seem like an attractive prospect — after all, users get more freedom, and what could be better than that? In the end though, the experience with open chatbots often falls flat.
Training a chatbot to understand questions takes a lot of time and resources, both of which businesses often find in short supply. Additionally, if users don’t know what kinds of questions they can ask in advance, they can be frustrated when most of their requests are met with errors or dead-ends in the conversation. Users also need to be trained how to use the chatbot, which means what was originally supposed to be a natural interaction now has become rigid and systematic.
In response to this technical challenge, the idea of “locking down” chatbots has been gaining popularity. A fully closed chatbot can only be interacted with through predefined options, and a semi-closed chatbot is one that combines predefined options with free form-text. A popular implementation of this is to give users some commonly asked questions as options, but also allow them to ask a question. At each step in the conversation, the chat can be augmented with predefined options to help the user progress through the conversation.
We use this concept at Orderly Health to aid in helping users find answers they need to questions they may not know how to ask.
Healthcare is a sprawling topic with thousands of possibilities for the types of questions that can be ask.
Adding to the challenge is the fact that many people may not fully understand what they need in advance, so it’s not unusual for someone to not know exactly what they want, much less how to phrase a question in a way that will generate a meaningful response. Using predefined options lets us encourage users to stay on a “happy path” while also giving them the ability to ask questions if they know what they are looking for.
If your business is currently using chatbots with a poor success rate, or if you’re considering adding a chatbot to help improve your business, I highly recommend investigating ways to close down the chatbot.
Computers don’t know what humans want, which is why our — I am a part of the human group — questions are so often misunderstood. If you can offload the burden to the user by means of some buttons or other rich messaging features, your chatbot will seem a lot smarter even though it’s doing less work.
And in the end, you will have happier users because of it.
