Writing Chatbot Conversations: What You Need to Know — Part I

By Kristi Colleran - January 25, 2017


While chatbots are not new, having got their start in 1966, they have become increasingly relevant. This is being driven by advances in Artificial Intelligence (AI), more specifically in the areas of Natural Language Processing (NLP) and Natural Language Understanding (NLU). These advances have given us many new capabilities including smarter and more sophisticated chatbots.

Then why are we seeing the tremendous excitement in chatbots be tempered with such considerable disappointment?

While the bot and messaging platforms have made it relatively easy to stand up a chatbot, we are quickly realizing that making a good one isn’t trivial. As with any new technology and the capabilities it affords, there’s a lot to figure out and new areas of expertise to be developed.

One of those areas of expertise is writing chatbot conversation. I wrote my first chatbot conversation in April 2016 and can attest to the dearth of resources and advice available on this topic! I would like to change that by sharing some of my real world experience designing and writing conversations for chatbots used in the event industry, the enterprise and services businesses. If you are: faced with writing or being responsible for chatbots in your organization, a bot builder for whom writing is not your core competency, a writer considering a new medium for your talents or just genuinely curious about this topic, then it is my hope that you will find value in this series on writing chatbot conversations.

Let’s get started!

Know Your Audience

One of the basic tenants of writing is to know your audience. Writing chatbot conversation is no different. Here are three key things you should know about your audience along with the implications they will have on your chatbot’s conversation.

#1 — Solve Their Problems Simply

Ensure your chatbot has a clear purpose and adds value for your audience. If it doesn’t, it’s not a good use of your time or that of your audience. Seek out use cases for chatbots that leverage their unique attributes (available 24x7, no app to download, no waiting on hold or for live chat to respond, etc.) over other communication channels to solve problems and address needs more quickly and simply than existing solutions. To get it right means you need to do your homework to understand what your audience is trying to get done, why they are trying to do it, how they are currently approaching it and what is a satisfying outcome for them. Two methods that can help you get this right are: Jobs-To-Be-Done and User Interviews.

#2 — Be Where They Are

Once you have clearly defined your chatbot’s purpose, you will need to identify what messaging platforms your target audience uses and understand the capabilities/limitations of each. Popular messaging platforms that support chatbots include: Facebook Messenger, SMS, Web, Kik, Skype, Telegram, Viber and Slack. Below are 4 dimensions the platforms vary on, along with examples to illustrate the differences that will factor into your selection:

  • Geography — What’s App is the dominant world leader but currently doesn’t support chatbots. Of the platforms that do, Facebook Messenger is popular in the US, Canada and Australia, SMS in North America but less so in other parts of the world where it’s pricey and Viber in Eastern Europe and the Middle East.
  • Demographics— Kik focuses on teens and early 20-somethings, Facebook Messenger is more popular with women than men and SMS with baby boomers.
  • Setting — Slack is primarily used in the workplace while the other platforms tend toward more B2C or personal use.
  • Capabilities/Limitations— SMS, the oldest platform having been in existence for over 20 years, is limited to texting. The newer messaging platforms, often referred to as OTT (Over the Top), offer more efficient and richer experiences via visual affordances such as buttons, quick replies, carousels and cards.

The capabilities of the platform(s) you select will have a significant impact on the design and implementation of your chatbot’s conversation. Let’s take a look at two examples.

Buttons — users can simply tap a button to trigger an action. For example, tapping a button labeled Parking presents the user with a text response, “here are the available parking options”, with relevant information like an image of a map showing the available garages and links to driving directions. On platforms that don’t support buttons like SMS, your users will have to type the button label vs. tapping. This means you’ll want to keep labels succinct. On platforms that support buttons, you can include more descriptive text and images on button labels to more effectively guide users and help them discover what your chatbot knows.

Presentation of Buttons on SMS vs. Messenger

Carousel — is a stream of horizontal scrolling containers that can include images, text and buttons. Carousels work well for presenting catalogs, agendas and menus of items. On platforms that don’t support or fully support carousels, the carousel will be presented vertically. It will be functional with text and actions and may or may not support images. Below you can see that Messenger fully supports carousel (swiping left or right scrolls the Agenda horizontally, SMS does not support carousel and Telegram partially supports carousel.

Presentation of Carousel on Messenger vs. SMS vs. Telegram

As you can see there are a lot of implications for your chatbot when it comes to platform choices!

#3 — Speak Their Language

In this case, I’m not referring to English, Spanish or German, rather these are factors that can influence how your chatbot converses with your audience. Three key factors include:

  • Demographics — The good news is that texting has become ubiquitous and messaging apps are now the most widely and frequently used apps on a smartphone. With 92–100% of US adults texting daily, the only thing new about a chatbot experience will be who is on the other end! Of course, you may need to adjust the text-speak your chatbot uses for your audience’s age and gender.
  • Literacy Levels —Consider how your chatbot will handle queries from users who may struggle with spelling or literacy. Or what about users who don’t speak English as their first language? Literacy rates are lower in the US than most people realize so these are not unfounded concerns. Fortunately, AI tools are available that can help your chatbot address these challenges. Microsoft Cognitive Services is a good place to start learning about these.
  • Context and Situation — Think about what your audience will be doing when they interact with your chatbot and where they will be. How your chatbot converses with a busy professional that needs to know where to charge their iPhone at a conference vs. a user relaxing at home with basic questions about a product likely will be different. Regardless, just ensure your chatbot solves their problem or addresses their information need quickly and simply. Then, everybody wins!

In Summary

While it’s relatively easy to stand up a chatbot, making a good one isn’t trivial. There’s a lot to figure out and new areas of expertise to be developed including designing and writing chatbot conversation. The more expertise we develop, the better chatbots will. So what are you waiting for, roll up your sleeves and get started today!

Check out Part II — Bringing Your Chatbot to Life where I cover the important aspects of defining your chatbot’s identity, making it human-like or not, the four dimensions of voice and to assign gender or not.



Case Study

Drive Event

Deploying Sciensio's Eventbots solution to improve and enhance a major industry trade show.

Get Case Study

We promise that we won't SPAM you.