From today’s perspective, the chatbots advancement and popularity may not be as revolutionary as expected by the enthusiasts a few years back. Anyway, the number of bots on various platforms has grown to such an extent that we can now speak of chatbots as a new way of user assistance that Information Developers should be aware of.
Idea behind chatbots: simplification of user assistance
Back in 2016, while presenting the Facebook Bots for Messenger platform, Mark Zuckerberg said that “you should be able to message a business the same way you message a friend”. If you want to order a pizza or get a taxi, you do not have to call a company or download and install a specialized app. A much simpler way of getting what you need is just opening your messenger and texting the company’s chatbot. The concept of bots as mini-apps working in messengers was a step towards the simplification of customer-service relations and better user experience. Already in April 2017, on Facebook alone, there were 100,000 bots, and the number of messages sent between consumers and businesses reached 2 billion a month, according to David Marcus, Head of Messaging at Facebook.
How do chatbots work?
Depending on the platform, bots can be website-based or messenger-specific (added via bot stores in Facebook, Telegram, Skype, Slack, and so on). As Chatbots Magazine explains, “usually all the bots connect to a platform like Kik, Slack, Facebook, Messenger, and many others via an API to send and get the messages.“ In their interaction with the users, chatbots employ the conversational user interface (CUI) instead of the traditional GUI with menus and buttons to be clicked or tapped. From ordering pizza to answering “how-to” questions, chatbots fit into a variety of user-assistance cases.
For further reading, I highly recommend the following:
- A great post on the progress of user interfaces and using AI for better user experience.
- An introduction into conversational UI by Google.
Types of chatbots
There are numerous chatbots classifications. For example, Chatbots Magazine provides the classification of chatbots according to the value they bring to the users. Another classification divides them according to their ability to maintain the state of the conversation, or “remember” the context.
To put it simply, most classifications are about just how smart a chatbot is in having a conversation. As Ellis Pratt sums up in his great text “Artificial Intelligence and Chatbots in Technical Communication – A Primer“ (which is indeed a must-read for all Information Developers), we can differentiate between script-based chatbots and AI-based chatbots.
- Script-based chatbots
These “not-so-smart” chatbots work according to the rules or scenarios written for them. They are able to understand you only if your answer strictly coincides with the answers they expect. Usually, during a conversation, you are offered several options that you can select as an answer. Options may be presented to you as buttons.
For example, take a look at Scoop, the news chatbot by Microsoft.
Notice how news summaries are presented in a conversational form.
Additionally, Scoop can search for the “unexpected” requests:
- AI-based chatbots
These “smarter” bots use the Artificial Intelligence technology to analyze your responses, find the most relevant answer, and learn in the process. Take, for example, the Mitsuku bot. In 2016, Mitsuku won the Loebner prize for being the most human-like chatbot, which means it passed the Turing test!
A good example of AI-powered chatbot is Poncho, the weather chatbot that searches for a weather forecast for your location, offers you predefined options to choose from, understands your replies, and occasionally cracks a joke. Poncho creator Sam Mandel is quoted saying that “thanks to a team of writers updating the bot’s library of knowledge every day, Poncho already has opinions on random, non-weather related topics like guacamole recipes and superhero movies”. Writers, see?
You can type something else and see if Poncho understands:
What a sly guy!
Why should we care?
I can think of several reasons why Information Developers should pay close attention to the emerging chatbots trend.
Reason 1: Chatbots require good writing
With bots employing the conversational user interface, the quality script is a major component of a bot’s success. Bot’s authors must possess exceptional audience analysis skill, understand user goals, and be able to write with empathy—the qualities required from InfoDevs. Well, and have a good sense of humor.
Now, take a look at how it is done:
- Here is a great story about experiences and lessons learned of a TechComm team working on a chatbot script for the first time.
- I also invite you to read a great experience-sharing post about dealing with conversational design form the content writer’s perspective.
Reason 2: Chatbots are a channel for our existing deliverables
Use cases for chatbots are not limited to helping a user perform a simple practical task, such as ordering pizza, getting news reports, or booking a hotel. Chatbots can also help users out with an application or website, offering relevant Help topics, FAQ sections, or knowledge base entries.
What would it mean in practice? As Ellis Pratt suggests, we need to make sure that our writing is chatbot-ready.
To be provided via a chatbot medium, our writing should be structured so that chatbots can process its meaning or semantics, for example, reading metadata or keywords from the tag in HTML. Metadata would help a chatbot “understand” which topic is relevant to a user’s question. Metadata can include a component and feature name, task, user role (admin or non-admin user), and more.
Additionally, we get one more reason to make sure that each topic answers one specific question a user might ask.
Chatbots are coming. In fact, they are already here, so let’s not overlook this trend as it gains momentum. Being aware might be your competitive advantage. Even if you are not writing for a chatbot now, you can keep this potential possibility in mind.
What are your experiences with chatbots? What do you think of a chatbot delivering users Help topics? Please share your ideas.