Messaging is infiltrating everything right now — from customer support, to banking apps, to analytics. You could be forgiven for thinking we’ve slipped into an alternate dimension, where AOL Instant Messenger never died! What began as a simple, uncluttered 1:1 communication channel for teenagers has inexorably grown toward near-complete ubiquity among netizens.

Upon a bit of reflection, it’s not difficult to see why:

It’s instant, intuitive and easy:  Type, tap, send.
It’s efficient: “I concur with your thoughtful analysis, my good sir” condenses down to “k” in text-speak
It’s forgiving: The hamun biarn is remakrably adbatpale and toarelnt wehn it cemos to intterpreing ttxe, so lnog as the fsrit and lsat leettrs of ecah wrod are coerrct.
It’s cheap: SMS is either very cheap or free for most of its users, meaning…
Everyone’s using it: A Pew Internet study found that 97% of smartphone owners use their SMS app at least once a week.

In other words, text messaging offers users a simple, frictionless interface that’s easily understandable in any culture, and is available to everyone worldwide. So it should come as little surprise that we’re starting to see the familiar text-driven chat interface popping up as a new-but-kinda-old solution to UX design problems in the form of chatbots.

chat·bot
ˈCHatbät/
a computer program designed to simulate conversation with human users, especially over the Internet.

Chatbots are at the nexus of several major trends that are converging fast: 1) consumers becoming accustomed to directly communicating with services as well as each other via text-based interactions, 2) companies leveraging marketing automation to respond to their customers in a customized yet scalable way, and 3) speech recognition technology improving its speed and accuracy by leaps and bounds — enough to enable the real-time synthesis of contextually meaningful responses to users’ queries.

Chatbots are basically the Devastator of user interfaces today — and they’re going to have an appropriately-sized impact on UX design

And boy, have they multiplied as a result! Within just a couple of months of opening its messaging platform to chatbots in April 2016, Facebook Messenger alone now offers over 11,000 of the darn things; so we can safely say that learning how to create a useful, engaging chatbot is a legitimate concern for anyone whose work impacts the customer experience of their company.

In this post, we’re going to discuss some design approaches and critical areas to consider when developing a customer-facing chatbot for your organization. We’ll help you choose the best approach for your needs, as well as cover some of the critical UX aspects to pay attention to, then briefly analyze some of the most popular chatbots around for a little inspiration.

K? 😉

Preparing to Build Your Chatbot

Before you start writing scripts and choosing fonts, it pays to invest time upfront clarifying the best use of chatbots in your organization.

Find a Good Home
Chatbot technology hasn’t yet advanced to the point of passing the Turing Test (but perhaps we’re close?), so they’re not yet flexible enough to handle every potential customer interaction. Simply creating a chatbot without carefully considering where it should live within your overall customer experience will likely end up with a big, loosely-defined investment that duplicates work, confuses your customers, and yields little results. Ew.

Ugh.

Instead, your first step should be to either refer to (or create) an Experience Map to identify wherein the customer experience a chatbot could best help your company engage or retain customers at scale, while lowering or stabilizing your costs to do so.

The fact that chatbots aren’t a silver bullet solution means you need to be on the lookout for key indicators of an appropriate application. It also means as the bots get better, they’ll expand to encompass multiple touchpoints, consolidating into one conversational interface for your customers that potentially spans their entire experience from end-to-end!

Until that magical day, the most appropriate points to leverage chatbots in the customer experience are those where your organization…

…sees similar customer inquiries or needs that occur frequently…
…which benefit from a rapid response, and…
…are typically short, transactional conversations, with…
…a relatively limited range of potential outcomes.

Applying your chatbot to these scenarios will help you best leverage the technology’s current strengths: 24/7 availability, always on-brand, and cheap to scale by orders of magnitude.

Define Its Scope

Once you’ve planned out where your chatbot lives in the overall customer experience, you need to define what role it should play. Should it be handling the first touch of a new customer? Facilitating their purchase? Or, is it better suited to onboarding new users and responding to common support queries?

Another key question to answer upfront is whether your chatbot is peripheral to your product, or if it should serve as the actual interface for it. (More on this in a sec.)

Choose a Platform

Once you’re clear on how your organization can benefit from a chatbot, and what its requirements are, it’s time to go shopping for a platform to power these interactions with your customers. Good news: just like the bots themselves, the tools to build them are proliferating and becoming easier for developers and non-developers alike to use for prototyping & building. Folks like Motion.ai, Rebot.me, and Chatfuel have even made the process as simple as drag & drop — although your results may vary.

The more complex your needs, the more you’ll need to lean on developer-centric chatbot libraries like Pandorabots or even Facebook’s Messenger platform, which incorporate more robust text recognition for your users’ queries, but will also require more regular developer TLC. Hey, it’s just like building a website, right? The simpler your needs, the more off-the-shelf components you can use.

How to Design an Engaging Chatbot

With all the preliminary legwork done, let’s blast off with some insights into crafting a chatbot that’s a pleasure to talk to.

Language = Personality

Compose your chatbot’s responses as carefully as you would the fonts and color schemes of your website.

I’ve touched on this subject in a previous post on the UX of voice interaction, but it’s equally applicable here. As we transition away from visual interfaces, designers’ selection of vocabulary for their chatbots is critical. Words ARE the entire interface in this context, so compose your chatbot’s responses as carefully as you would the fonts and color schemes of your website.

This is where the fun of design thinking really comes in, as your chatbot’s vocab choices enable you to craft a whole persona around it. Should your chatbot be perceived by users as male, female or neither? Young or old? Casual and friendly with lots of emoji, or formal and polite? Within this new old medium, UX designers will be called upon to invent whole characters, seemingly living just behind the typed messages.

Your chatbot’s responses can be crafted to project any kind of persona

With great power comes great responsibility, however, and we’ll need to go beyond simply understanding our customers’ personas and into crafting a compatible chatbot counterpart to engage with them — potentially even developing multiple personas for the same bot!

It’s also important to intentionally decide whether or not your chatbot is opaque to the user. If it’s good enough to fool your customers into thinking they’re interacting with a real person, what’s the appropriate way to deal with the “ick” factor when they discover they’ve been talking to a bot this whole time? Granted, this may only be a generational concern that doesn’t even bother future users, who are young enough that they won’t even remember a time before conversations with computers were the norm.

It may sound far-fetched, but remember that people are neurologically hard-wired for the rapid onset of boredom when repeatedly exposed to the same stimulus — so it’s this element of unpredictability in your chatbot’s linguistic repertoire that’ll keep your customers engaged.

Personality Example: Quartz
The Quartz news app takes a unique approach to delivering you breaking news headlines — not only does it feed you the headlines one-by-one in a chat thread-styled interface, it uses a casual tone, jokes, funny GIFs and emoji to do so, (i.e., it’s totally biting my style.)

This gives the app a living personality, making it feel like you’re having a good conversation about the news with a friend. It’s especially engaging through the use of relevant emoji as the clickthrough CTA for each story, providing those little hits of soul, variability and unpredictability that users crave.

The chat-based interaction could use refinement in a couple areas. Firstly, although heavily-styled to look like one, this isn’t actually a conversational interface; the user is simply clicking through the stories as they’re fed to them, conveyor-belt style. Future iterations may add the ability to search for relevant articles via actual written responses — e.g. “What’s going on in South Africa right now?” or “What’s with the stock market today?” — and have them presented in the same engaging style.

The second problem? Keeping up with longer messages. When the chat interface serves up longer prose, it becomes difficult to read while it’s auto-scrolling:

While not a perfect solution, Quartz is a compelling demonstration of how chatbots can be leveraged to provide a far more engaging interface for traditional content.

Try it yourself: get the Quartz app on iOS or Android.

Keywords are the New Buttons

If the message thread is the interface of a chatbot, then your users’ words are its buttons — so they must be designed to be as easy to discover, intuitively labeled and memorable as possible.

We’re already seeing the need for this kind of focus on user-centered design with chatbots’ more sophisticated sister, Voice Interaction. Although the voice-driven interface of Alexa has received much fanfare this year for launching with thousands of apps available (called “Skills” in Amazon’s parlance), less than 3% of them are retaining users’ attention after 2 weeks of installation. The tech is working all right, but clearly, there’s still something missing.

Keyword Examples: GrowthBot and Digit

GrowthBot by Hubspot is built on the Facebook Messenger platform, and helps marketers by providing them a way to quickly query analytics and dig up information from different services.

Using Growthbot feels like using a search engine for marketing analytics: when you send it keywords and command phrases, it responds with instant answers. While it’s not something I’d want to chat with in my spare time, it dramatically speeds up the typical marketer’s workflow, by making critical — but separate — tools and data sources like Google Analytics and Moz, instantly accessible through a single, unified chatbot interface.

Digit is a mobile finance app that aims to ease the difficulty of regularly saving money. It leverages text message conversations to keep you informed of your current balances, as well as facilitating straightforward transactions triggered by simple, memorable and intuitive keywords.

It was a conscious design decision to build such core functionality to Digit’s product around SMS, in order to keep it lightweight and intuitive enough that very little training was necessary to productively interact with the app.

Typesetting Will Rise Again

That old saying “it’s not what you say, but how you say it” is more prophetic than ever when it comes to designing a chatbot. Fonts, legibility, formatting, kerning, line spacing — all of the ancient traditions of working with type — suddenly take on profound new relevance when dealing with a text-only interface, as they’re all aspects of a written message that convey meaning to the recipient and affect how your chatbot is perceived.

We all have that one relative/Grammy award-winning, controversial, but talented musical artist in the family who insists on emailing you in all caps. We must use our design powers wisely to avoid unleashing a chatbot-powered version of that relative — one that never sleeps and can have thousands of conversations around the world simultaneously. shudder

Funnily enough, as important as I predict type treatment will be in the design of an engaging chatbot, I couldn’t actually find any current bots that leverage it — which makes this a prime opportunity for some innovative design thinking in the space!

Interaction: Designing Your Chatbot’s “Fist”

This is a true story.

During World War II, British spies were able to not only decrypt the coded messages being sent by the Germans, but over time were also able to identify specific radio operators based on what was called their “fist” — their individual style, cadence, speed & frequency of their morse code messages. Tracking the different broadcasting locations of each fist as the radio operators were moved around enabled the Allies to keep tabs on the Germans’ troop movements, thereby anticipating where they were winding up to attack, and where they were vulnerable — in part turning the tide of the war!

Even with a messaging language as simple as the dots and dashes of Morse Code, humans can’t help but convey additional information with their unique “accents”, (like the colorfully-named Banana Boat Swing.)

If your chatbot is to successfully engage your customers, it’s important to consider its fist, or its interaction design. In this case, we’re using ‘interaction design’ to describe all the information you get from a text-based message, outside of the actual content-stuff, such as:

All important questions demanding answers.

Interaction Design Example: Slackbot

Millions of people are exposed to some variation of Slackbot as Slack continues its reign as the fastest-growing enterprise software in history, so it’s not surprising that they’ve paid attention to refining and honing what it’s like to chat with it.

Slackbot enables multiple UX techniques to convey meaning beyond its words. For one, it’s capable of using emoji to communicate reactions, thoughts and emotions to the user:

Receiving instant replies to your answers can be jarring in an extended chat conversation, so Slackbot solves for this by intentionally inserting a delay before responding to a command and notifying the user that it’s typing an answer — a useful lie, since there’s no need to actually give software time to type!


Small, intentional design touches like those can turn an annoying or distracting chatbot into a tool that’s tolerable enough to use frequently throughout the day.

Summary

Hopefully this post has given you an appreciation for some of the details involved in crafting a chatbot that people enjoy conversing with. To summarize:

Many folks see chatbots as an oddity right now, or as retro, stripped-down versions of more traditional visual interfaces — but by embracing the rich potential that chatbots offer in their simplicity and intuitiveness, we’ll open up a wide new frontier of interaction design, ready to be explored through thoughtful innovation and populated with memorable, meaningful interactions.

Comments
  • Some great points here, particularly in the design and conversation realms. My take is that we should focus on maximizing #chatbots in places where their current limitations are least damaging, for example on landing pages. At the same time we should be de-hyping the space and diligently working on creating an #ai driven conversational #chatbot standard that has a chance of going mainstream.

    • Cheers Simon! Agreed, until the technology becomes self-aware and herds all humans underground into its quartz mines, designers should pick the right tool for the right job. Brave is the soul who attempts to run their entire business funnel via chatbot today – but this will soon no longer be the case.

      • Hopefully, a lowly quartz miner’s salary will be enough to afford at least some of the crap that’s hawked to us these days. 🙂

  • I like your article, especially the way you explain the example of interaction design is simply awesome.

    • Thanks Acton! Yeah, these microinteractions can have an outsized effect on user engagement with a chatbot, so it pays to keep them in mind.