One of my free time projects has been to learn how to create a chatbot. If you have no idea what a chatbot is please read this.
There’s a pretty impressive amount of tech already out there that can get you started in no time… but I believe the key to building a good chatbot is the conversational user experience and that’s more in humanists territory than technologists.
Here’s my journey and my considerations about building a good chatbot.
Wit.ai
Since deep down I am in love with technology I got immediately hooked with Wit.aithat seemed to offer the most configurable, Swiss Army knife solution, for everyone’s need.
Wit.ai (which is now owned by Facebook) has great Natural Language Processing (NLP) powers and can be taught to easily understand almost all you can come up with. The downside is that it’s quite complex and long to train and it takes quite some time to even come up with proper questions (and answers).
I realized that I needed something simpler.
api.ai
But because I wanted to stick with my free-text + natural language processing idea I steered towards api.ai (now owned by Google) which looked a bit more simple.
In fact training the Artificial Intelligence (AI) behind was a breeze. It even allows you to speak to the editor to train the bot… but still you have to come up with the questions and answers and having free-text input form left too much freedom and uncertainty on the kind of answers you could actually do. Resulting in a terrible user experience (UX); way too many times the bot understood something different and gave either the default “Sorry I don’t understand” answer or completely the wrong one.
Give a person a text box and she will come up with 101 ways to surprise your bot
Conversational UX Design
I quickly realized that the tech behind wasn’t really the key to what I was looking for so I decided to start over and focus on the Conversational UX.
“What’s that?!” you might ask? Well… I just made that up.. :)
What I refer to is the design of the conversation that the bot has with the user.
Every conversation has a beginning and an end and each party in the conversation has something in mind, somewhere where they want to lead the conversation. And so bots have to.
If you are able to guide the conversation properly you can easily take the user through a funnel without her even realizing. Maybe increasing your conversion rate without ever changing your e-shop.
I had to first think about the flow and only later focus on the tech supporting that flow.
Welcome Chatfuel
After working a bit on the conversational user interface above I built for my hobby project I headed to chatfuel.com (that I initially discarded for being too simplistic).
Chatfuel can be easily integrated with Facebook (and Telegram) with just one click (really just one). It has a very basic keyword-based AI and its strength is leveraging the card-based chat interactions of Messenger as a default. It’s like the most basic thing you can do with a chatbot today.
But because of this simplicity you are forced to think at three simple questions that the user might have when approaching your bot and how you can easily guide the user towards the desired action(s) you have in mind.
It turns out that even with such a simple tool you can achieve quite remarkable results that feel very professional and friendly for the user.
It is not the all-encompassing experience that you would like to have with free-text input but it fits extremely well many use cases: from e-commerce to events information, to customer support triage and it feels real.
Technology is secondary
As a technologist this hurts but I have to admit that I believe in this case the user experience matters way more than the technology behind it.
We are not quite there yet with the NLP/AI platforms to achieve full comprehension of human language in a way that can provide a smooth user experience and card-based chatbots solution offer such a great and low cost alternative that I wouldn’t see why anyone should do anything different than that.
The key is to properly design the Conversational UX so that the user feels at home when talking to the bot and manage the expectations.
Start playing with Chatbots, now
Chatbots are still at very early stages, with Facebook, a behemoth with 2 billion users, paving the way to mass adoption. We are at the point where we were before mass adoption of websites. When having one meant being almost the only one.
As a business if you don’t have a strategy to offer even a simple chatbot to your users you will have to rush building one in the next few years.
Need help with your digitalization strategy?
Ping me on LinkedIn, I’d be glad to help you or visit my website to know more.
This article also benefits of the insights and the discussions between me, Matsand Mikko, so give them a pat on the back. You can read here another very interesting article by Mats on the topic.