THE

ENTERPRISE CHATBOT

GUIDEBOOK
 Contents

CHAPTER
1 WHAT IS A CHATBOT
ANYWAY?
CHAPTER
5 CHATBOT FEATURES
AND BENEFITS

CHAPTER
2 A BRIEF HISTORY
OF CHATBOTS
CHAPTER
6 HOW AI CHATBOTS
DRIVE BUSINESS VALUE

CHAPTER
3 CHATBOTS VS. CONVERSATIONAL
AI - WHAT’S THE DIFFERENCE?
CHAPTER
7 USE CASES AND
CASE STUDIES

CHAPTER
4 CONVERSATIONAL AI
- HOW IT WORKS
CHAPTER
8 CHATBOT BEST PRACTICES
FOR ENTERPRISE

2 3
 0 1 2 3 4 5 6 7 8

INTRODUCTION

Welcome to the enterprise chatbot guidebook! In the

pages of this guide, we will take a comprehensive look at
chatbots, conversational AI and their related technologies.
We will outline how and why conversational AI has
become crucial to the successful digitization of customer
service for large-scale enterprises, and go over use
cases and best practices for how to best get started
with a chatbot project.

Whether you’re just curious about automation or

researching in preparation to deploy your 8th chatbot
(go, you!), we’re confident that you’ll find everything
you need here to walk away with the knowledge
necessary to help your business thrive in today’s
world of digital-first customer service, sales and
support.

4 5
 0 1 2 3 4 5 6 7 8

WHAT IS A CHATBOT ANYWAY?

Let’s start with the basics. In its purest form, a chatbot is a computer just customer support and service,
program designed to allow interaction between humans and technology. but an organization’s operational
As the name implies, this was originally confined to just text-based efficiency, too.
communication, however, over the years, this interaction paradigm has What really distinguishes a good
evolved to include other input methods such as voice and gestures. chatbot from a bad one is the
underlying technology powering
Chatbots have also, in the recent past, chatbots has improved, so have the it. Artificial intelligence has
almost exclusively been deployed as number of use cases that they cover. revolutionized what’s possible to
customer service tools. They allow The 21st-century chatbot has evolved accomplish with a chatbot. Even
consumers 24/7 access to businesses from a simple question-and-answer the most basic bot with off-the-
and empower self-service, but that too bot into a swiss-army knife of shelf tech can automate simple
is changing. As the technology powering automation that can enhance not tasks. Larger enterprises, however,
have equally larger needs and
require sophisticated chatbots with
more powerful technology in order
to automate customer interactions
at scale.
What’s in a name?
You’ve no doubt noticed that there are myriad different
names out there for chatbots. Some popular ones That’s where conversational AI
include virtual assistant, AI chatbot, intelligent assistant, comes in and when things start
conversational agent... and the list goes on! to get really interesting, but…
hold that thought, we’ll get back
These names are basically all interchangeable and will vary
to this topic shortly.
based on the vendor or solution. At boost.ai, we prefer the
name ‘virtual agent’ because an agent, by definition, is First, let’s rewind a bit and look at
“a person who acts on behalf of another person or group”. where chatbots came from.
It makes sense then, that chatbots powered by conversational AI
should go by another name when, as you’ll learn from this guide,
that they can do so much more than just chat!

6 7
 0 1 2 3 4 5 6 7 8

A BRIEF HISTORY OF CHATBOTS

The first chatbot - or proto-chatbot, if you will It used a basic pattern-matching algorithm and substitution
- was developed in 1964 at MIT by computer scientist methodology to give users the illusion that the chatbot
Joseph Weizenbaum. Named ELIZA, it is considered understood them, but in reality, had no ability to contextualize
an early example of a natural language processing events. Even though ELIZA was pretty basic, it laid the
computer program and was designed to simulate groundwork for the future of chatbots. Below are some other
communication between humans and machines. significant milestones in chatbot history that brought the
technology to where it is today.

The Turing Test PARRY A.L.I.C.E Siri

Described as “ELIZA with an attitude”, this Inspired by ELIZA, the Artificial Linguistic Launched initially as a standalone iPhone app,
Even before ELIZA was a glimmer in its creators’
Standford-developed chatbot attempted Internet Computer Entity was the natural Siri was integrated into iOS with the launch of
eye, Alan Turing posed the question of whether
to simulate a person with paranoid language processing chatbot that itself the iPhone 4S in 2011. This ushered in the era
a machine could think in his seminal paper
schizophrenia and successfully fooled many served as the inspiration for the 2013 of voice-enabled virtual assistants that included
‘Computing Machinery and Intelligence’.
experienced psychiatrists. film Her. Google Assistant and Amazon’s Alexa.

1950 1995
1972 2010

1966 1988 2016

2001
ELIZA
Jabberywacky SmarterChild
The world’s first chatbot was proof positive Facebook Messenger Bots
that humans were eager to communicate with Jabberwacky was an early attempt at creating Available on AOL Instant Messenger and MSN
an artificial intelligence through human The wide-spread adoption of chatbots
machines. ELIZA could carry on (relatively) Messenger, SmarterChild was the first chatbot
interaction and was designed to simulate exploded when Facebook announced that it
convincing conversations by mimicking to achieve mainstream adoption by millions of
natural chat in a humorous way. It was would begin allowing bots onto its popular
response users in the early 2000s.
eventually released online in 1997. messaging platform. By 2018, there were more
than 300,000 active chatbots on Facebook
Messenger.

8 9
 0 1 2 3 4 5 6 7 8

CHATBOTS VS. CONVERSATIONAL AI - WHAT’S THE DIFFERENCE?

Basic chatbots only have the capacity to complete a limited number of tasks. Typically,
this means answering simple FAQs and not much else. In order to meet the demands
of larger enterprises, chatbots need conversational AI to enhance their ability to
understand human language and to provide transactional functionality in addition to
their informational capabilities.

Conversational AI is the synthetic brainpower that

makes chatbots capable of understanding, processing and
responding to human language.

Conversational AI can be used to power chatbots to become

smarter and more capable. But it’s important to understand
that not all chatbots are powered by conversational AI.

Scalability is king!
Large organizations, like banks or government
agencies, often have thousands of customer
service requests every day that can vary wildly.
This can be impossible for basic chatbots to keep
up with. Conversational AI can enable a chatbot
to scale its language understanding and capacity
without any reduction in accuracy.

10 11
 0 1 2 3 4 5 6 7 8

CHATBOTS VS. CONVERSATIONAL AI - WHAT’S THE DIFFERENCE?

Using sophisticated deep learning and natural language

understanding (NLU) algorithms, conversational AI makes Chatbots with
it possible for a chatbot to go beyond translating website Basic chatbots conversational AI
content into simple chat responses.
Online 24/7
It can unlock the potential for enterprises to empower customer
Natural language understanding Keyword-based tech
self-service by automating complex interactions such as blocking
credit cards, filing insurance claims, upgrading data packages,
Dynamic, context-based navigation Button-focused navigation
generating invoices and much, much more.
Multi-level intent hierarchy If/Then statements

Unlimited scalability Limited improvement capacity

Broad scope Narrow scope

3rd-party integration support Limited understanding model

Self-improving over time

Consistently high-resolution rates

Omni-channel

Entity extraction

User authentication

Voice and conversational IVR

Multi-lingual

Privacy & security compliant

12 13
 0 1 2 3 4 5 6 7 8

CHATBOTS VS. CONVERSATIONAL AI - WHAT’S THE DIFFERENCE?

hatbot technology Computational Linguistics

In the 2019 report, _‘Competitive Landscape: Virtual A flexible approach for dealing with multiple languages
-Assistant Platforms, Worldwide’ , Gartner outlined by tackling customer requests at various linguistic levels.
three types of technology that most of the chatbots
you will come across on the internet are built on.
Easily adapted
Understanding the difference between these three to new languages

key chatbot technologies will help you make the right

High level of variation
decision on which one is right for your business:
requires regular maintenance

Rule-based programming
Common architecture of most basic chatbots. Machine Learning (conversational AI)
Uses keywords and other language markers to trigger
predefined responses. Most advanced implementation. Uses a large set of
training data that enables deep learning algorithms to
classify intents and better understand human language.

Easy to implement High response accuracy.

Scalable

Deployment can be
Does not scale
resource-intensive

14 15
 0 1 2 3 4 5 6 7 8
In use

CONVERSATIONAL AI - HOW IT WORKS

We are open
What are your
Please help Does my travel
opening hours? NLP NLU 9am to 4pm, It sure does! I’m sending a new
insurance cover a broken Mon-Fri one your way right now!
fishing pole? Cleaning up Understanding Formulating a
the request what is being said human-like answer

INTENT
PREDICTION
ASU TM
ENTITY
EXTRACTION

Identifying the Boosting Extracting important

correct topic understanding and information (entities)
reducing errors from the text

An ‘intent’ refers to the goal or topic a customer

How does conversational AI translate human language into something has in mind when typing in a request.
Once the request has
it can respond to and action on? On the surface, it’s deceptively simple
been preprocessed, a variety of deep important information that can be used to
- a customer interacts with a chatbot and is given an appropriate learning and machine learning models trigger additional actions such as context,
response. But there are actually a number of different technologies take over. account preferences, entity extraction,
working behind the scenes to ensure this interaction goes smoothly. sentiment classification etc.
These are collectively known as Natural
Language Understanding (NLU), and are Tldr; NLP is anything that has to do with
The first step involves Natural Language interpret grammar and break down a a subfield of NLP. NLU is responsible the processing of natural text, while NLU
Processing (NLP). It’s the job of NLP to request into words and sentences that for identifying the correct intent (i.e. is more specifically about making sense of
correct spelling, identify synonyms, make it easier for a chatbot to understand. topic) of a request and extracting other a chunk of text.

16 17
 0 1 2 3 4 5 6 7 8

CONVERSATIONAL AI - HOW IT WORKS

Once the request is understood, it’s

time for the chatbot to formulate a
response. Conversational AI outshines
basic chatbot solutions in its ability
to communicate in a personalized
manner. By combining the information
gathered by the NLU (customer
intent, contextual information, etc.)
with a structured hierarchy of
conversational flows, a chatbot
is able to respond appropriately
and in a conversational manner,
whether it’s answering a simple
question or carrying out a complex
transaction on a customers’ behalf.

As the chatbot has more and more interactions,

conversational AI can grow smarter, improving how
Proprietary power-ups! it responds. This can be aided by technology like
self-learning AI that improves based on conversation
At boost.ai, our NLU stack includes a
data, and by human AI Trainers that tweak the model
proprietary algorithm called Automatic Semantic
to make it more effective and efficient over time.
Understanding (ASU). ASU helps to enhance the
overall language understanding capabilities of
our virtual agents, increasing their ability to parse
complex sentences and reduce the occurrence of false
positives by up to 90%.

18 19
 0 1 2 3 4 5 6 7 8

CONVERSATIONAL AI - HOW IT WORKS

Do more with conversational AI AI-assisted human chat

Thanks to advanced language understanding capabilities Connect a chatbot to existing knowledge
conversational AI can enable chatbots to do so much more. base systems and supercharge its ability to
assist front-line customer service staff.
Conversational AI can provide helpful
Self-learning AI suggestions and increase employee
Build smarter and deploy faster with artificial productivity without needing to put
intelligence that improves with the more data you customers on hold.
feed it. This can mean scanning a website (or another
chatbot) to build a viable model in just a few hours Integrations
or analyzing conversation data to help AI Trainers Conversational AI can make
optimize in real-time. it possible for a chatbot to
integrate with applications
Voice and you already know and
conversational IVR love. That means front-end
With powerful language understanding, a chatbot integrations like Zendesk,
can go beyond text-based interactions and become Facebook Messenger and
the foundation for voice-driven customer service using Slack, and powerful back-end
text-to-speech and speech-to-text integrations. integrations like RPA and OCR.

Virtual Agent Network

Large enterprises can break down departmental
silos by connecting multiple chatbots together in
a network. Advanced NLU detects and transfers
users to the correct bot seamlessly and within
the same chat window to maximize the customer
experience.

20 21
 0 1 2 3 4 5 6 7 8

CHATBOT FEATURES AND BENEFITS

Now that we’ve established what chatbots are and how they work, Scalable intent hierarchy
it’s time to dig in and find out how they can help.
Many basic chatbot solutions are only able to automate actions and answer
When deciding on the right conversational AI platform to deploy for your questions on 100-200 topics at most. This is fine in the case of a florist or food
business it’s important to have a clear understanding of the key features and delivery service but soon becomes untenable for larger enterprises such as banks,
benefits that an AI-powered chatbot can deliver. insurance companies, telcos, e-retailers or government organizations.

Here’s a list of enterprise-grade features you should look for and expect: Larger organizations often have complex product and service offerings that
require a scalable chatbot solution. A robust conversational AI platform will allow
High accuracy & resolution rates a chatbot to handle a number of intents orders of magnitude higher than rule-
based chatbots, averaging in the thousands, rather than just a few hundred.
A chatbot can’t help if it doesn’t understand. That’s why conversational AI
needs to be built on robust language understanding algorithms in order Conversational AI places intents into a hierarchical structure making it easier to
for it to be effective at automating customer requests. A chatbot must scale. Having all intents at the same level (which is typical of many solutions)
be able to: makes them difficult for the AI to keep track of. Instead, categorizing intents per
σ Interact with customers in a conversational manner subject matter (i.e. insurance types, banking products, etc.) negates the need to
scroll through a long list in order to find the correct intent.
σ Understand and act on customer intent, regardless of how
complex the request is
Broad scope vs. narrow scope
σ Identify multiple intents in the same request and provide
customers with actionable responses for each Trying to solve a narrow set of problems with a chatbot may seem like a quick fix
σ Understand context to keep interactions from veering off-track but, ultimately, will not lead to sustainable returns on investment. Conversational AI
σ Ask follow-up questions to clarify information and gather actionable data
makes it possible to cast a wider net and deploy chatbots that have a broad scope.

These crucial features make it possible for conversational AI chatbots This approach is particularly applicable to enterprises that have large volumes of
to achieve high automation rates of up to 90%. Without these in place, customer service traffic every day, and goes hand in hand with a chatbot’s ability to
you run the risk of frustrating customers and reducing their confidence scale and automate at consistently high resolution rates. A conversational AI chatbot
in your chatbot. with a broad scope can help an organization achieve long-term strategic goals
instead of just short-term wins.

22 23
 0 1 2 3 4 5 6 7 8

CHATBOT FEATURES AND BENEFITS

Total cost of ownership By giving an organization complete ready-made content for your specific Conversation analytics
ownership of the solution - from vertical. Instead of building banking
As conversational AI becomes smarter building and implementation to intents from the ground up, you can Analyzing and understanding what your
and more capable, it’s natural for optimization - you achieve greater implement a banking-ready chatbot customers are asking about is crucial to
enterprises to want to be more in organizational buy-in which results in that has knowledge about credit cards, chatbot optimization. Look for analytics
control of what their chatbots can better customer experiences overall. mortgages and more, and then tailor tools that allow an organization to
achieve. This means moving away from its responses to match your brand mine conversation and interaction data,
the need to hire data scientists and Pre-built, vertical-specific guidelines. with features such as advanced search,
developers towards solutions that offer content tagging and quality rating, clean-up
low-code or no-code software and puts Combined with self-learning AI, this
reports and API support.
the power of automation into the hands Launching a brand new chatbot from can dramatically reduce implementation
of customer service reps, leveraging scratch can mean a lot of initial work. timelines from weeks and months to
This can be mitigated by selecting a only a few days.
their expertise and experience to build Seamless human hand-off
dynamic customer interactions. conversational AI platform that offers
The advantage of robust natural
language understanding is that it
allows a chatbot to correctly identify
when it doesn’t recognize a customer
request. Conversational AI can use
this information (complete with a
chat transcript and account info) to
seamlessly loop in a human agent to
assist without the customer experience
breaking down.

By recognizing when something is

outside of a chatbot’s scope and
leveraging human support to help
Value per interaction Number of interactions Value created instead, you stand to increase customer
satisfaction and loyalty.

24 25
 0 1 2 3 4 5 66 7 8

HOW AI CHATBOTS DRIVE BUSINESS VALUE

Customers prefer to use chatbots over legacy channels like phone and Open up new channels
email because they are convenient, easy to use and empower self-service.
_Chatbots saw a 92% use increase since 2019,
Chatbots also have a number of benefits for enterprises including helping _making it the brand communication channel with
to reduce operating costs and increasing employee productivity. _the largest growth._

Chatbots allow enterprises to deploy and test new,

fully-automated omnichannel experiences for sales,
Here is a list of the top areas where may be interested in, helping to increase
service and support without tying up resources to
conversational AI-powered chatbots sales and drive revenue.
onboard and manage new teams.
can drive value to both businesses and
Keep costs down
consumers: Bolster brand loyalty
_Chatbots can cut operational costs by
Instant response times _up to 30%.
_67% of US Millennials say they would purchase
_a product/service from brands using a chatbot.
_69% of consumers prefer chatbots
A scalable conversational AI-powered
_because of their ability to provide fast Deliver your customers a dynamic and memorable
chatbot can do the equivalent work of
_answers to simple questions. self-service experience and they’ll keep coming back
_hundreds of human agents , without
(and spending more!).
Consumers value speed and an organization needing to onboard any
convenience when seeking help online. additional staff. Available 24/7, 365
AI chatbots eliminate the barrier
between a brand and its customers by Increase employee _Over 50% of customers expect
_a business to be open 24/7.
providing immediate responses. efficiency
In today’s connected world, it’s not unexpected that
Increase revenue _70% of white-collar workers will
_interact with conversational platforms customers should want to reach you outside of office
_Chatbots expected to drive $112 _on a daily basis by 2022. hours. Chatbots present a fast and scalable way to
_billion in retail revenue by 2023. answer customer queries about your business when
Free up employees to focus on high-
it suits them best.
Chatbots powered by conversational AI value customer interactions by using a
can interface with a customer’s account chatbot to empower self-service and
and suggest products and services they automate repetitive inquiries.

26 27
 0 1 2 3 4 5 6 7 8

USE CASES AND CASE STUDIES

With the right conversational AI platform powering it, a chatbot can be

a powerful tool for both customer-facing support/sales and optimizing
organizational efficiency.

In this chapter, we will explore a variety of chatbot use cases that can be
applied across industries, and share examples of client success stories.

ID-126
Banking, credit unions & financial services Add test da
ta to the la
Car Insura test intent
nce , Electric
Automate core banking services Test data
High
Whether it’s opening a new account, reporting a lost card, checking account
balance, processing mortgage payments or any other core banking services,
conversational AI makes it possible for customers to have agency over the process
via a friendly, conversational interface without needing to involve a human operator
unless they want to.
Case study: DNB
Instant & personalized service
Challenge Norway’s largest bank needed to scale customer
Break down customer service and support barriers by delivering instant, accurate support to handle high volumes of incoming chat traffic
and consistent answers to questions related to your bank’s products and services.
User authentication can allow for personalized and proactive responses, tailored to Solution Routed all customer service chat traffic through
individual customers’ needs, that improve over time. conversational AI chatbot

Onboarding and internal support Result Chatbot automated over 50% of all incoming
chat traffic in 6 months.
Combine conversational AI with RPA to automate time-intensive back-office
processes, such as document management and contract review, ultimately Now accounts for 20% of all
increasing employee productivity and streamlining operations. Onboard new staff customer service automation,
with training and knowledge base content built directly into the AI chatbot. including channels like phone
and email

28 29
 0 1 2 3 4 5 6 7 8

USE CASES AND CASE STUDIES

Insurance
Automate claims processing Case study: Tryg
For insurance companies, ensuring a fast and frictionless claims process is key to
Challenge Denmark’s largest insurer needed to reduce
maintaining strong customer relationships. A chatbot can answer questions about
complexity for human support staff in order
claims instantly, while interfacing with back-end systems and 3rd-party integrations
to increase operational efficiency
(i.e. Optical Character Recognition), to fully automate the claims process.

Boost employee productivity Solution Deployed internal AI-powered chatbot to

assist customer support staff by answering
Complex policy structures and documentation can be accessed by support and questions about policy and products without
service staff via an internal-facing chatbot, instead of needing to call back-office needing to put customers on hold
support lines. This can increase productivity and help to maintain consistency of the
information provided to customers, while simultaneously reducing overall call times. Results Chatbot answers questions on over

1,200 topics with a 95% success rate.

Assists over 750 employees daily

Fight fraudulent activity

Chatbots powered by conversational AI can help insurance firms to fight fraud and
protect customer privacy. Natural language technologies can be trained to identify
early warning signs of fraud and allow for automation processes put in place to
mitigate potential challenges.

30 31
 0 1 2 3 4 5 6 7 8

USE CASES AND CASE STUDIES

Case study: Telenor
Telecommunications
Challenge The largest telecommunications company
Powerful upselling potential
in the Nordics needed a channel to deliver
Combine existing customer data with conversational AI to create intelligent cross-
consistent, on-brand customer self-service
selling opportunities. Purchase history and account info can be leveraged to offer
personalized plans and services that customers didn’t even realize they needed.
Solution Deployed a customer-facing AI chatbot with over
Increase acquisition rates 20 unique integrations to automate
Chatbots present a unique and dynamic channel for customer acquisition. Potential key processes including requesting a PUK code,
customers can get to know which of your products and services are the best fit for
upgrading mobile data plan, viewing an invoice,
them through targeted questions and AI-powered recommendations based on past
user behavior. and more

Product and technical support Results

Telenor’s chatbot can answer questions on more

If a customer has a question about one of your products or is experiencing technical than 2,000 topics and successfully handled over
difficulties, a chatbot saves them from having to wade through FAQs to look for 630,000 inquiries in 2020. Chatbot integrations
an answer. Conversational AI can help diagnose the issue and either provide
automated support or transfer the customer to the correct human operator for helped Telenor reach its ROI goals in less than
further assistance. 12 months

32 33
 0 1 2 3 4 5 6 7 8

USE CASES AND CASE STUDIES

Chatbots don’t have this problem. They are available 24/7 (yes, even on public
holidays!) and can help citizens with queries outside of standard office hours.
Public sector
Collecting feedback
Easy access to public data and information
Implementing a new policy and want feedback? Chatbots provide a unique
Chatbots are a great way to give citizens access
way for citizens to engage with local and national government through dynamic
to public information that may otherwise
conversations that can act as surveys or simply as a place to let people’s voices
be difficult to track down on government
be heard.
websites or by phone. This can include
anything from information on new
legislation to what day of the week the
rubbish is collected in your neighborhood.

Extending office opening hours to 24/7 Case study: Norwegian Labour and Welfare Administration

Government agencies are famously known Challenge Help businesses and citizens maintain access to key social benefit
for having limited opening hours, making programs during coronavirus pandemic
it difficult to get in touch with them.
AI chatbot was deployed to help assist with questions related to
Solution
pension, child support, unemployment benefits, employee sick

leave and more

Result Conversational AI allowed chatbot to scale to handle over

270,000 inquiries at peak of the pandemic, doing the

work of 220 FTE with an 80% success rate including

channels like phone and email

34 35
 0 1 2 3 4 5 6 7 8

USE CASES AND CASE STUDIES

E-commerce
Promote sales and marketing campaigns

A chatbot can intelligently make recommendations to customers based on

active sales and marketing campaigns by asking questions or interfacing
with a CRM system for personalized recommendations.

Re-engage customers

Avoid abandoned carts by using conversational AI to remind customers

that they have an unfinished order. Chatbots can help complete Case study: Posten
lingering purchases by providing additional product information
or helping to resolve any issues a customer may have. Challenge Norwegian postal service needed to scale customer

24/7 order support service and support during busy holiday period

Extend your support hours to 24/7 with the help of automation.

Solution Chatbot launched with advanced functionality
Offer automated return processes and assist customers with
inquiries outside of contact center office hours so that they including parcel tracking and address change
can keep buying your products and services when it suits
them, not you. Results More than 370,000 customer service inquiries

successfully automated in December 2019.

In December 2020, Posten’s chatbot surpassed

over 1 million customer interactions

36 37
 0 1 2 3 4 5 6 7 8

USE CASES AND CASE STUDIES

Healthcare
Provide critical information

A chatbot can be a great resource for a hospital or medical Case study: Västra Götaland
organization to provide accurate and updated information
on procedures, illness symptoms, mental health, health Challenge Sweden’s second-largest county needed a reliable
insurance and more.
and always-available channel for its 1.7 million
Support doctors residents to access information about the latest

Conversational AI can interface with medical databases coronavirus guidelines

and other systems to help doctors retrieve information
that can be helpful during patient diagnosis. Answers to Solution AI chatbot developed and launched in just
queries on symptoms, medication, dosages and more
are just a chat away. 6 working days in collaboration with local
healthcare professionals
Assist patients

More than just answer-bots, AI chatbots can also help patients Results More than 800 conversations automated
in a proactive manner. They can remind you when it’s time to
each day. Chatbot proved so effective that it
take medication, provide instructions for how to apply simple
treatments and, with the help of 3rd-party integrations, was adopted by the wider Swedish population
help monitor patient health.

38 39
 0 1 2 3 4 5 6 7 8

CHATBOT BEST PRACTICES FOR ENTERPRISE

Building and implementing a chatbot powered by conversational AI can

be a daunting task - but it doesn’t have to be! In this final chapter, we
2
Scope
want to share a list of best practices that the boost.ai delivery team has
developed from hundreds of successful virtual agent implementations
across Europe and North America.

Following these recommendations can help you, as a project manager, successfully

Delivering
deliver your next conversational AI project quickly without compromising on quality. an AI-powered
Decide on the correct use-case(s) for your chatbot virtual agent
Identify which parts of your organization would benefit most from automation and 1
Identify use-case
then decide on relevant use-cases. Use-cases can include:

3
Build
Service and support Internal knowledge Sales optimization
(external chatbot) (internal chatbot) (external sales chatbot)
σ Automated first responder σ Centralized knowledge base σ Automated sales
σ Automate a portion of σ Convenient and anonymous σ Self-service and web-form
customer service and support guidance
σ HR, IT and payroll support
σ Increase self-service rates σ Increase up-sell and cross-sell
σ Personalized
σ Increase support capacity σ Revenue goals with
σ Cost-effective 24/7 support
tracking and analytics For more information on­
Cost-effective 24/7 support
σ
σ Handle unexpected peaks industry-specific use-cases
σ Design and optimize
σ Handle unexpected peaks
σ AI assist with Smart Replies customer journeys refer back to chapter 7.

40 41
 0 1 2 3 4 5 6 7 8

CHATBOT BEST PRACTICES FOR ENTERPRISE

Set and agree on a clear scope Identify any limitations that could slow down or impede the project from
moving forward
Establish the scope and KPIs for your chatbot and ensure all stakeholders are
aligned. This is where you can set goals for your project and assess what resources Assemble the right (dream) team
may be needed. When setting scope and expectations, remember to:
Buy-in across all levels of your organization can be crucial to your chatbot’s
Outline goals and clarify expectations, so that all stakeholders are on the same page success. We recommend the following structure for your internal (your company)
about what they want to achieve and external (the chatbot vendor) teams. This can, of course, vary from
organization to organization but is a good starting point for how to set up
Define the requirements for the product, process and project. This a winning project team.
means focusing on what questions need to be covered by the chatbot

Internal team External resources

σ Executive sponsor (critical) It is important to include an ‘Executive σ Project manager (optional)
sponsor’ on your internal team. This should
σ Project manager ideally be someone in a senior position who σ AI Supervisor/Platform expert
can act as a champion for the project and
σ Content Designers/AI Trainers help to remove any roadblocks. σ AI Trainer

σ User testers σ Technical specialists

σ Frontend developer

42 43
 0 1 2 3 4 5 6 7 8

CHATBOT BEST PRACTICES FOR ENTERPRISE

Integrate your chatbot dynamic interactions and conversation Also, it can be important
flows that make for the best to monitor what types of
with your support team
automated customer experiences. inquiries your human
Some AI chatbot solutions are black support staff see an
boxes that need a team of developers Choosing the right KPIs increase or decrease in
or data scientists to operate. This is once the chatbot has
When setting KPIs, you need to be
not the right approach for enterprises been implemented.
mindful of the use-case that you have
that may require greater control over This will give you a
selected for your chatbot, and the
their customer service automation. good indication of
scope you have landed on.
where the bot may
It’s important to opt for a be helping most.
Here are some common performance
conversational AI platform that has
metrics that you should look to
a user-friendly, no-code or low-
measure:
code, interface. This will allow for an
organization’s best customer service σ Resolution rate
resource - its existing customer service
σ Deflection rate
staff - to take ownership of training
and maintaining the chatbot. These σ Avg. handling time
subject matter experts, due to their
σ Reduction in queue times
experience and expertise, are best
positioned to develop the kinds of σ Number of sales automated

44 45
 0 1 2 3 4 5 6 7 8

CHATBOT BEST PRACTICES FOR ENTERPRISE

Chatbot visibility matters If you make your virtual agent Anticipate and mitigate risks
difficult to find, customers will
Visibility is one of the core principles of typically interpret this as you not All projects inherently have some level of
(interaction) design. The more visible wanting them to get in contact risk attached to them. If you can plan for
an element is, the more likely users will - a cardinal sin in customer service. these risks in advance, then it can mean
know about it and, in turn, interact with Here are some best practices to the difference between being surprised
it. This applies to chatbots just as much follow for virtual agent visibility: by an unexpected roadblock, and deftly
as it does to websites, buttons, navigating around it.
or content.
These are some recommendations for
how to think about risk mitigation for
σ Ensure your chatbot is easy to find
a conversational AI chatbot project:
on your webpage

Identify and describe potential risks

Consider going σ Design a visually appealing avatar σ
that is instantly recognizable
‘chat-first’ σ Determine the probability of how likely
σ Use text to encourage engagement it is for a risk to become an issue
For the best ROI, we recommend with the chatbot, i.e. “I can help!”
that our clients adopt a ‘chat- σ Categorize the level of impact the
σ Use dynamic designs such as avatar
first’ approach. This is where all issue might have on the project
animation or text pop-ups when hovering
incoming chat traffic is directed over the chatbot σ Have a plan in place for what
through the chatbot first before mitigation measures to adopt
σ Proactively launch the chat window
other channels like phone or live to grab the customer’s attention to deal with the different
chat. This allows an organization to categories of issues that may arise
automate the highest possible amount Chatbot visibility should also align with
of interactions, and our NLU will your KPIs. There is no sense in hiding your
intelligently transfer customers to the chatbot away on the ‘Contact’ page of your
right human agent if a request is outside website if you set a KPI goal to automate
of the chatbot’s scope. 50% of all chat inquiries. It needs to be
visible on the front page and promoted
We have many clients that successfully
so that it can reach its full potential.
use this approach to automate upwards of
40% of their total customer service traffic,
including phone and email.
46 47
 0 1 2 3 4 5 6 7 8

CHATBOT BEST PRACTICES FOR ENTERPRISE

Chatbot project pitfalls to avoid

Without proper planning and foresight a large-scale chatbot project can quickly
derail. Here are five of the most common pitfalls to avoid to ensure your virtual agent
launches on-time and on-scope.

σ Insufficient planning - Failing to plan is a big no-no. Establishing a detailed

roadmap for where you want to see your chatbot in 12, 18 or 24 months
crucial to reducing overall risk and failure rates.

σ Scope creep - Keeping your eye on the prize, and scope under control,
is critical for efficient delivery. There will always be time later to add in
additional functionality to your new chatbot.

σ Not anticipating hiccups - Even the most carefully planned project can run
into trouble. Sometimes hiccups are unavoidable, but it helps to
acknowledge that they can happen, and to try and have an contingency
plan in place in case they do.

σ Undefined or misaligned expectations - Your c-suite may define success in

a completely different way than your front-line support team, so it’s
important to make sure your Critical Success Factors are easily measurable.

σ Not thinking about the ‘ever after’ - You need to be thinking ahead to after
a project is completed; how will the solution be operated, built-upon and
continue to grow? Taking time to consider these factors will make you
ready to tackle them later.

48 49
 0 1 2 3 4 5 6 7 8

CHECK OUT OUR OTHER GUIDES

_Conversational AI trends in customer service: 2021 and beyond

_5 essential steps to enhancing customer interactions with conversational AI

_How conversational AI can significantly improve customer experience in the

t

_financial services industry

_Conversational design vol. 1 - Interaction: Using data-driven insights to build

_a better user experience for your virtual agent

_Conversational design vol. 2 - Visibility: How to maximize user experience for

_your virtual agent

_Conversational design vol. 3 - Avatars: Considerations, approaches

_and recommendations for virtual agent avatar design

50 51
To learn more about how
you can use conversational AI
to help your business put
customer experience first,
contact us today.

www.boost.ai | [email protected]