Mark Swaine UX Guy Designing Bot Experiences

PROJECTSChatbots
84 february 2017
chatBots
We love our messaging platforms; they give
us the opportunity to communicate in ways we
normally would not when face to face – such as
emojis, memes and atmospherics (iOS 10). The
voice-call as we know it is fading, and not just
in personal use, but in business use also. Even
voicemails are considered archaic and laborious,
with a preference now to be text-based.
a neW paradIgM
Over the past five decades we have experienced
shifts in messaging, from ELIZA, DoS bots and
gaming to MSN Messenger. The core of human
connection to computing has been through the
messaging paradigm. As we live through this
HOWTODESIGNA
CHATBOTEXPERIENCEMark Swaine runs through the factors you need to consider
when designing the user experience for a conversational bot
abouT THe auTHor
Mark sWaine
w: www.uxguy.com
t: @UX_UI_Guy
job: Head of experience
design, UX Guy
areas of expertise:
UX and UI design
q: if you could change one
thing about the industry,
what would it be?
a: The execution of design
education
ViDeo
This video walks
through how to build
a simple Facebook
Messenger chatbot
prototype with Chatfuel:
netm.ag/chatfuel-289
It’s your morning commute. Look around you;
most people are heads-down, looking at their
phone, whilst at the same time ordering anything
from coffee to a train ticket. They are semi-aware of
their surroundings and hooked within a messaging
communication application window of some type on
their device. Messaging windows are everywhere,
and the rate of use and number of options available
are both intensifying.
Whether we like to admit it or not, messaging
interaction platforms are contributing to our daily
mobile screen time. They are the only applications
we allow into our lives intrusively through
notifications, and the only type of application
we truly want to spend our precious time in.
NET289.tut_chatbot.indd 84 12/12/16 11:54 AM

february 2017 85
Chatbots
current cycle, new messaging platforms with new
usable and useful interfaces, and with much more
personal user context are being created.
Platforms such as Facebook Messenger, Google
Allo, WhatsApp, WeChat, Slack, Skype, kik, Twitter,
Viber, Google Home and Amazon Echo – to name
just a few – are all creating messaging, voice and bot
experiences for their user bases. These experiences
are being created for personal, commercial,
enterprise and automated use.
In China, the number of daily users of WeChat
alone has risen this year to over 762 million. And
with 50 million business users, Facebook Messenger
is exploding its commercial API capabilities weekly,
allowing brands and services to order food, book
a flight, contact local emergency services or send
a payment through simulated bot services.
The White House has just launched the first-
ever government bot (AI) on Facebook Messenger,
allowing users to ask the President questions. Some
users might even think they are actually speaking
to the President.
The death of apps
Many online editorials have detailed the death of
apps. At the start of 2016, Uber’s Chris Messina
predicted that, “you and I will be talking to
The voice-call as we
know it is fading, and
not just in personal use,
but in business use also
As with any new technological cycle, tool options for
developers and designers will emerge quickly. New tools for
creating and designing bot experiences are launching daily. Below
are some common ones you might want to take a look at.
Wit.ai
www.wit.ai
Used to help develop natural language within bot experiences
using text and voice.
Botsify
www.botsify.com
This helps you build self-learning chatbots for Facebook Messenger.
Flow XO
www.flowxo.com
A business bot platform for Facebook Messenger, Slack, SMS
and Telegram.
Converse.AI
www.converse.ai
Create an intelligent chatbot on any channel, including Facebook,
Twillio, SMS, Telegram, Twitter, Smooch and Slack.
api.ai
www.api.ai
Build brand natural language interactions for bots.
Chatfuel
www.chatfuel.com
Allows you to build bots for Facebook Messenger without coding.
Tools
Focus on
White House Facebook Messenger has just welcomed the first ever
government AI, where people can speak to the White House

PROJECTSChatbots
86 february 2017
brands and companies over Facebook Messenger,
WhatsApp, Telegram, Slack, and elsewhere before
year’s end, and will find it normal.”
However, the singular-threaded conversational
utopian OS/interface state, that will inevitably
replace apps is still a long way off. For singular
tasks, it will always be necessary to download a
software application to your device, but perhaps
there will be fewer of them as time goes on.
Bots are maturing fast, as customer experience,
service and marketing teams explore the potential
of conversational UI for solving issues with
self-serve processes (such as booking a hotel or
changing flight details). And bots such as Cently
enable users to purchase from retail stores within
Slack. Bots do not solve all service requests, and
human intervention is still needed to address more
ambiguous tasks.
A UX design party
As the user’s pace of input and absorption of
content within messaging windows increases,
brands and services will need to meet this demand.
A truly seamless, end-to-end ecommerce customer
service bot experience is still a way off, but
platforms such as Agent.ai are progressing fast
in this area. The case for exploring purchasing
simplification through conversational UI has
never been so important.
With the growth of online communities, shared
resources, tools and APIs, a conversational UX
design party has truly started. Now is an incredibly
exciting time to be a UX/UI practitioner exploring
bot and AI automation and machine learning in
natural language messaging interfaces.
As a designer, emphasis on the ‘design of
words’ (natural language) and user intent within a
messaging window will be at the fore of your work.
To convey the pace and depth of practice emerging
in this space, in this article I have covered only a
Slack bot The Cently Slack bot enables users to purchase from retail
stores within Slack
Learning resources
in-depth
If you want to explore chatbot UI design further, there are
plenty of resources available. Here are some of the best:
The Ultimate Guide to Chatbots
netm.ag/toscano-289
This article by Joe Toscano on Medium explores why chatbots
are disrupting UX, and takes a look at some of the best practices
for building them.
Slack bot directory
netm.ag/slack-289
Slack’s app store is a great place to browse bot services that can
live within a Slack thread.
Bot List
botlist.co
A great bot aggregator website, listing new bots added daily across
all messaging platforms.
Chatbots Magazine
chatbotsmagazine.com
Billed as ‘the #1 place to learn about chatbots’, this is an amazing
resource for new best practices and how-to guides.
UX Design CC
uxdesign.cc
A great source on Medium for bot UX and best practices.

february 2017 87
Chatbots
few key areas for consideration when designing
a successful bot user experience.
A good place to start
If you are considering designing a bot for your brand
or service, there are a lot of factors involved. First
you need to decide if a chatbot is the right option
for your product. Consider the following:
● Can your core products and services translate
into text in an organic and natural language
paradigm?
● Will a bot speed up purchasing, ordering and
customer service issues, and make things more
convenient for your users compared to using your
current website or app?
● Will your new bot align with current company
business requirements and objectives?
● Can core services and product relationships be
decoupled and broken down to function as micro-
services within a messaging UI?
● What problem would a bot’s information
architecture attempt to solve?
Creating a successful automated (end-to-end)
chatbot is hard. Because it is a relatively new
medium for organisations, best practice for
implementation is learned, tested and evolving
daily and globally.
Be sure to spend time researching, testing and
breaking bot flows across multiple platforms, like
Messenger and Slack. Learn how different companies
handle intent (mapping a phrase to an action),
on-boarding, responses, suggestions, dead ends
and natural language.
You also need to make sure you have the ability
to put this into practice. Skills such as natural
language programming, conversational UX, and
in-house technical competencies are vital, but you
also must factor in the constant effort required to
maintain a successful bot experience.
Designing a system
We are still in the early days of building and
interacting with automated commercial services.
The assumption for users when interacting with
an automated service is that their intentions will
The dialogue flows
need to be short and
simple, with reduced
interaction points
be misunderstood, and attempts to interact with
the service will fail. As such, it’s important to
thoroughly consider your system dialogue flow.
First, prototype your flow as information
architecture. You are designing a systematic set
of conversational flows that take a service from
beginning to end. The flows need to be short and
simple, with reduced interaction points – the aim is
to enable users reach their intended result as easily
as possible.
If user intent is misunderstood or not achievable
by the system, be honest and let them know they
need to try a different approach. The chances of
this happening can be reduced by including
introductory contextual on-boarding, which
includes suggestions or prompts based on the
capabilities of the system.
It’s vital to accommodate for any potential dead-
end scenarios. Here are some tips:
● When designing your branching dialogue,
document a matrix of possible dead-end
situations, and establish alternate routes
for each of them
● Users need an exit option when navigating the
wrong branch of a conversational flow, otherwise
they may drop off. Always pose an escape option
to allow users to correct the situation
● To guide users in the right direction, your
system needs to provide simple expected dialogue
direction for their next steps, such as ‘Browse
latest sneakers’ when attempting to view desired
products, or ‘Order Taxi’ when booking a cab
Use of natural human language within menus and
micro-button options need to be contextually exact
in meaning to allow swift progression and prevent
Joining in Messaging
platforms from Whatsapp
to Telegram are all creating
voice and bot experiences
for their users

PROJECTSChatbots
88 february 2017
‘guess selection and input’. This also prevents users
starting the conversation a second time.
System responses and prompts
Bots are primarily conceived to provide expertise,
increase convenience and speed of interaction with
a brand, service or general information request.
There are some fundamental approaches to bot
responses that need to be considered.
First off, responses need to be short and concise,
to avoid misunderstandings resulting in back-and-
forth dialogue with the user. Always provide direct
button option prompts, like ‘Yes’ and ‘No’ or ‘Read
this story’ within a thread to help the dialogue
progress. For a guide, take a look at the CNN bot
service on Facebook Messenger.
If a customer must input more than two or three
times to correct intent or flow, the experience
has failed. System responses such as ‘Sorry I
didn’t get that’, ‘Do you need help?’ and ‘Would
you like to see our deals?’ can be frustrating and
misleading, depending on the user’s task. If a user
must seek help to establish the correct response
that will rescue their position within a flow, they
will become frustrated, restart or drop off. Finally,
always confirm a user’s input, be it right or wrong,
before progressing the dialogue.
Facebook Messenger
Case study
Facebook currently has over one billion users globally,
and 900 million using Messenger. As a result, brands and
organisations are creating and testing bot experiences on this
platform first. With over 18,000 bots deployed and no searchable
‘bot store’ on the platform, the Messenger ecosystem is still in its
early days, with capabilities being updated continuously.
Messenger has made it easy for brands to customise, prototype
and test bot experiences on its platform, using built-in messaging
templates, navigation patterns, rich media and calls to action. Once
your bot is deployed on Messenger, your new business experience
is accessible to millions across mobile, tablet and desktop on iOS,
Android and Windows.
An exploding third party ecosystem of bot-building platforms,
such as ChatFuel and Flow XO allow brands to deploy simple
experimental brand and business bot experiences within hours.
With the purchase of the wit.ai – a natural language bot engine –
Messenger can allow to you create more advanced bot experiences
using ‘Speech to text’ input capabilities.
For a list of Messenger’s template capabilities for your bot
experience, take a look at netm.ag/messenger-289. Here you’ll also
find further guidelines on how to treat greeting text and persistent
menus, and how to test your bot. Whatever media you
use in your dialogue
flow, make sure it is
used sparingly
CNN bot CNN’s Facebook Messenger bot displays the latest articles,
suggestions to progress, images and excerpts

february 2017 89
Chatbots
Design your bot
As a designer, your job is now to design words, while
considering dialogue flow, voice and thread position.
When designing bots on Facebook Messenger, you
will find that customisation capabilities are limited.
This is a good thing – if all brands implemented
custom design for their bot experiences, the UX
would become chaotic, distracting and inconsistent
across different platforms.
The design needs to be controlled and adaptable
within the guidelines of the particular platform, as
the user would expect it to be. As a result, you need
to accept that your flows may look a little different
across platforms.
Here are some tips for designing your bot:
● Begin with sketching the branching flow dialogue
of your bot, focusing on the key actions it will
address for users
● Tools such as motion.ai, Twine and Mind Node can
be useful for prototyping dialogue branching
● In your initial prototype, detail all dialogue
relationship flows, sentences (verbs), subjects,
user tasks, navigation and potential dead points
● The goal is to always steer the conversation
forward using contextual suggestions, options
and prompts based on understood user intent
Although you will have little GUI design control,
some messaging platforms allow you to add product
photos, icons and menus, edit the colour palette,
use animation, sound and video, and insert stickers
and conversation landmarks. Whatever media you
plan to use as part of your dialogue flow, make sure
it is used sparingly, is on-brand and contextually
relevant. Design elements should not interrupt the
dialogue flow. Similarly, the design of selectable
button options and prompts should act as a support
to the dialogue, not overtake the UI.
However, be sure to use interface elements such
as animation and live typing indicators. Users are
used to these interaction paradigms as part of
conversational UI messaging windows.
Conclusion
As the area of automated conversational UI heats
up, we are going to see a plethora of messaging
platforms (commercial, personal and business),
natural language integrations (Watson), machine
learning, APIs, tools and UX and UI best practices
explode. Platforms such as Facebook Messenger
and Slack will attempt to spearhead the trend.
Brands now need to ramp up and self-educate
internally at pace. They will have to investigate
whether their products and services can translate
into conversational natural language experiences.
Figuring out whether they can live within this
window as a micro-service, and support users’
expectations while reimagining internal services
and products will be hard. Finding the talent to
lead these endeavours may also be hard.
For designers, we need to accept that the
expectations of our role are broadening quickly.
Automated and bot conversational UI is just one
of many current trends in technology that we need
to invest in, self-educating daily. The role of a user
experience designer is increasingly an ‘always on’
practice, evolving daily.
Next issue: Luis Rodriguez explains how to build a chatbot
Wall Street Journal
An example of solid use
of main menu navigation
within the conversational
UI flow
Resource
Motion.ai is a handy
resource that helps
you design your
bot’s dialogue flow:
www.motion.ai
Starbucks bot Customisation capabilities are limited to ensure consistency
within a particular platform

Mark Swaine UX Guy Designing Bot Experiences

More Related Content

What's hot (20)

Similar to Mark Swaine UX Guy Designing Bot Experiences (20)

Mark Swaine UX Guy Designing Bot Experiences