White Paper

AI in the Contact
Center:
Strategies to
Optimize the Mix
of Automation
and Assisted
Service

Prepared for:

February 2018
Table of Contents
Executive Summary……………………………………………………………………. 1
AI Overview……………………………………………………………………………….. 2
The Basics……………………………………………………………………… 2
Current State of AI………………………………………………………… 3
The Case for AI in the Contact Center………………………………………… 5
Key Drivers for Chatbots……………………………………………….. 5
How Chatbots Add Value………………………………………………. 6
The Chatbot Ecosystem…………………………………………………. 7
Developing a Successful Chatbot Strategy…………………………………. 9
Setting Objectives…………………………………………………………. 9
Managing the Process…………………………………………………… 11
The Bigger Picture…………………………………………………………. 14
Conclusion – Moving Forward……………………………………………………. 15

List of Tables
Table 1 – Key Uses Cases Where Chatbots Add Value………………… 7
Table 2 – Chatbot Ecosystem – Type of Platform……………………….. 8
Table 3 – Chatbot Ecosystem – Type of Developer…………………….. 8

Executive Summary
Artificial Intelligence seems to be everywhere now, and as enterprises start to see business value, the
contact center quickly becomes part of the conversation. The associated technologies are evolving
rapidly, and while that creates a sense of urgency to leverage AI, it remains poorly understood. Not only
do decision-makers lack basic knowledge about AI, but there are often unrealistic expectations around
its capabilities as well as the outcomes once deployed.

Most contact centers now struggle with a fundamental dilemma to provide good customer experiences
both with live agents and self-service options. No contact center can afford to provide assisted service
with their agents for every inquiry, and existing self-service capabilities are limited and often ineffective.
There is a clear need for better forms of automation, and this is where the promise of AI really
resonates.

This white paper has been written to help contact center decision-makers better understand AI’s
potential value, and more specifically, how chatbots can help address this fundamental dilemma. The
key lies in developing a holistic strategy for using chatbots to strike a balance between improving the

AI in the Contact Center J Arnold & Associates Page 1

self-service experience and optimizing the use of agents to provide assisted service. Along with outlining
such a strategy, this white paper reviews industry-based best practices for developing a chatbot project,
along with what to look for in choosing a chatbot provider partner.

AI Overview
The Basics
Artificial Intelligence is one of the most important technology trends in 2018, and it holds tremendous
promise for the contact center. While this field of computer science has been active for decades, only
recently has it evolved to provide value in the enterprise. Our view is that AI can be particularly effective
for improving the customer experience – CX – especially for helping contact centers optimize the mix
between automated self-service and assisted service from live agents.

With so much changing so quickly, our research shows both a sense of urgency and anxiety among
contact center decision-makers to deploy AI. Expectations are high, but so is the risk for jumping into
projects with emerging technology that is not well understood. To mitigate that, the following is a basic
outline of AI’s building blocks, and a summary of the current environment surrounding AI.

Artificial Intelligence

AI is a computer science discipline focused on emulating human intelligence and behavior with
machines, namely computers. Contact centers can use these capabilities to automate various forms of
customer service, as well as to make agents more productive, and ultimately provide a better CX. The
term “AI” may sound like a solution, but it’s really a framework supported by several technologies, along
with applications that build on these technologies. Aside from there being many supporting pieces, they
are often inter-related, making it difficult to clearly define AI. In terms of this white paper, however, the
following three are the most relevant.

1. Natural Language Processing

This branch of AI focuses on using human language to interact with computers – both with voice and
text. The key challenge is to address the ambiguity of language so computers can understand the
context of the communication and the intent of the speaker to determine the right meaning. NLP is
based on software algorithms that continuously analyze human speech samples until patterns are
recognized to minimize ambiguity and produce highly accurate results that can be used to develop other
AI applications such as chatbots.

AI in the Contact Center J Arnold & Associates Page 2

 2. Machine Learning

As a different but related branch of AI, this is a form of learning theory based on using the
computational power of computers to process large volumes of data, recognize patterns, and from that
make statistically-driven predictions. ML represents a highly scientific approach to learning where
computers can continue refining these statistical models without external programming. In theory, the
more data there is to work with, the more accurate these models and their predictions become. The
recent advent of cloud-based networks has been a boon to ML, and by analyzing large volumes of
customer data, contact centers can benefit by enabling agents to make better decisions that match
customer needs.

3. Chatbots

As the term implies, this application is about enabling interactions between people and computers to
automate tasks and streamline business processes. Similar to NLP, there are both voice-based and text-
based chatbots, but despite being software-based, they are not always fully automated. Current
capabilities are still rather limited, and chatbot interactions often require some degree of human
involvement that customers may or may not know about.

The value of chatbots is largely determined by AI, but it should be noted that chatbots are not always AI-
based. In terms of the contact center, however, the above AI elements do play a key role. First, NLP
enables more conversational interactions so customers can get better self-service outcomes, and
building on that, ML enables chatbots to make better choices when handing self-service inquiries off to
live agents, as well as anticipating customer needs with greater accuracy. While AI provides the overall
framework for these capabilities, chatbots are the applications that contact centers will deploy, and
hence serves as the focus for this white paper.

Current State of AI
Before considering the business case for chatbots, along with developing a strategy to deploy them in
the contact center, some context is needed around the current state of AI. Interest has never been
higher, and realistic expectations are needed to ensure the broad hype around AI doesn’t lead to hasty
decisions or poorly planned implementations for chatbots. To provide that context, here are six reality
checks to consider.

1. While some of the hype around AI is warranted, it should not be viewed

as a game-changing silver bullet to solve all your contact center
challenges.

AI in the Contact Center J Arnold & Associates Page 3

 AI can – and will – be used to reduce costs and drive smart automation, but be wary of messaging
focused on reducing or eliminating the need for agents.

2. Another aspect around AI hype is the growing adoption of digital

assistants, both on smartphones and in the home.
Industry giants such as Amazon, Google and Apple are quickly scaling consumer demand for
intelligent chatbots, and those expectations are now carrying over into the contact center. Rapid
innovation here is transforming the customer experience, and contact centers need to respond.

3. Contact centers do need automation in today’s digital world, but AI has a

long way to go to completely automate operations and/or replace agents.
The underlying technologies are rapidly improving, but you need to understand what chatbots are
currently capable of and where they can truly add value. When considering chatbots for the first
time, small successes are more important than trying to get to the finish line right away.

4. Reducing cost is important – and AI can deliver that - but only in a tactical
manner by replacing agents with automated forms of service.
The strategic value of AI, however, pertains to improving the customer experience, from which
longer-term, sustainable benefits are realized. This outcome requires more effort and better
planning, and serves as a key driver for this white paper.

5. Be mindful when associating automation with labor savings.

IVR has provided self-service options for decades, but agent headcounts have not gone down as a
result. The outcome with AI will be no different if you take a strategic approach. If the rationale for
chatbots is to improve CX and drive engagement, automation will create more customer touch
points, and the need for agents may even increase. Customer-centric businesses actually embrace
this, and for them, automation is about having more engagement rather than less operational
expense.

6. Contact center decision-makers need to understand AI is a broad

discipline, especially if management has made it a business priority.
To whatever extent they understand AI, the contact center needs an informed perspective to
develop and justify their business case. This entails understanding how AI can improve existing
processes, and not just replace them. Replacing agents with chatbots to save money is not a realistic
plan for AI right now, but they can be used today to bring some new automation capabilities that
will also improve CX. By getting executive management to see this bigger picture, the contact center
will have a much longer runway to leverage chatbots in the areas that really matter.

AI in the Contact Center J Arnold & Associates Page 4

The Case for AI in the Contact Center
A basic understanding of what AI is - and is not - provides the foundation for moving forward with
chatbots, and the next step in your journey is building the business case. This section breaks that down
into three core elements from which you can develop a successful strategy. The first element is
understanding the drivers for chatbots, and two sets are presented below – drivers for the contact
center and drivers for your customers. Building on this will be a summary of use cases for how chatbots
can add value, followed by an overview of the chatbot ecosystem to help identify the right type of
partners for your needs.

Key Drivers for Chatbots

There are many reasons why chatbots are coming of age now for the contact center, and while this
analysis is not intended to be comprehensive, it provides a strong set of factors than any contact center
could build a business case around. Whatever mix of drivers aligns best with your situation, the case will
be stronger when the challenges facing both the contact center and customers are considered.

AI in the Contact Center J Arnold & Associates Page 5

Chatbot Drivers in the Contact Center

• AI has matured and there’s now a rich ecosystem for the contact center, with plenty of innovation
yet to come
• There is an ongoing need to reduce operational costs, and with this maturity, AI’s automation
capabilities are drawing strong interest for the contact center
• Countering the need to reduce costs – especially agent staffing - contact centers face growing
demands from customers, not just for the rising volume of inquiries on a 24/7 basis, but also
supporting them over a wide range of channels
• While web chat may be the preferred channel for automating customer service, traffic volumes are
rising and agents must multi-task to keep pace, with the result being poorer performance overall –
to improve self-service CX, contact centers need a better form of automation
• Aside from needing to streamline operations, contact centers need agents to be productive, high-
performing and satisfied in their work – without the right tools, all of these suffer, leading to agent
turnover, higher training costs, and poor CX

Chatbot Drivers for the Customer

• For many companies, the current CX is poor, with key pain points being too many points of contact,
too much repetition of basic information and disjointed handoffs during the session – all of which
could be mitigated with intelligent automation such as using chatbots
• With today’s global, always-on economy, many consumers like the convenience of shopping 24/7,
either online or in-store, and increasingly expect the same for customer service – even if that means
using self-service options to get a resolution in the moment rather than waiting the next day to call a
live agent
• Self-service is a cornerstone of our digital world, and Millennials in particular have a growing
preference for this given today’s technologies – and this extends to their consumer lives, both for
how purchases are made as well as getting customer support
• Related to self-service, Millennials also show a growing receptiveness to engage with chatbots – and
a comfort level with AI in general – so long as their needs can be met
• Another important trend among this younger customer demographic is the emergence of messaging
as a preferred communication channel – this channel is well-suited for chatbots, and for contact
centers to be effective today, support needs to be provided in their native environment

How Chatbots Add Value

The drivers outlined above help explain why there’s a case to be made for chatbots now, and this
section applies that specifically to the contact center. Table 1 below summarizes key use cases, along
with how each provides value both for the contact center and for improving CX.

AI in the Contact Center J Arnold & Associates Page 6

Table 1 – Key Uses Cases Where Chatbots Add Value

Contact center challenge Impact on contact center How chatbots add value
Customers still relying on legacy Channels not well integrated, with By delivering better self-service
channels to reach contact center poor forms of self-service that slow options, chatbots can help migrate
down the process and put more customers to digital channels to
demands on live agents both improve CX and the use of
contact center resources
Consumers are now multi-channel, Limited ability to support all these If developed to support multi-
and can engage across many channels with existing technology, channel, chatbots can provide
endpoints – smart phone, landline, as well as providing a consistent CX better self-service, regardless of
PC, Echo, etc. across them channel used by customer
When customers shift from self- Legacy systems provide no context AI capabilities for chatbots can
service to live agents, CX is usually when handing off to agents, so provide continuous context, so
poor and feels disjointed time is wasted repeating agents can seamlessly continue the
everything, driving costs up and session as it shifts over from self-
customers away service
Providing a good CX around the Poor self-service options drive up Chatbots can provide better self-
clock, not just from 9-5 cost to provide more agents, and service options, which helps
when overworked, their optimize the use of agents, plus,
performance declines along with chatbots work 24/7, and don’t
job satisfaction behave badly under stress
Improving overall CX and meeting Legacy systems constrain ability of Chatbots can help provide faster
expectations of today’s customers agents to meet these expectations answers and better answers to
and get beyond IVR for self-service improve CX, and they are always
learning, so can keep pace with
changing needs

The Chatbot Ecosystem

To some extent, this white paper outlines the buyer’s journey for chatbots, initially for understanding AI,
and then for the role chatbots can play in the contact center. Once those needs have been addressed,
and the business case is well-defined, the next step is to understand the chatbot landscape, especially if
this is your first foray into the world of AI.

There is no single model that defines the chatbot ecosystem, and this white paper can only break down
the basic segments. Aside from being very new, this landscape is quite fluid, with a constant stream of
start-up entries, along with larger players entering the market with their own chatbot applications.

When considering a chatbot for the first time, choosing the right type of partner is essential, and the
starting point is understanding the basic varieties. Tables 2 and 3 below summarize two ways of
segmenting the market – by type of platform and by type of developer - but this taxonomy is by no
means definitive. However, it provides a basic compass for direction in terms of aligning with the type of
chatbot application needed for your contact center.

AI in the Contact Center J Arnold & Associates Page 7

Table 2 – Chatbot Ecosystem – Type of Platform

Platform type: Query-based Learning-based

Basic description: Like IVR, these chatbots are rules-based, Being more AI-driven, these chatbots
and respond to commands from understand both commands and language.
customer inquiries. There is no By using NLP, they can engage with
conversational capability or context to customers in an conversational manner.
clarify ambiguities. This can be an They also use ML to get better over time to
efficient form of self-service, but it won’t handle more complex inquiries.
feel very personal.
Main benefit: Can be developed fairly quickly and Able to manage both closed and open-
inexpensively, and if well-planned, will ended queries, and build up deep domain or
perform fairly predictably. This only vertical market expertise. These chatbots
applies, however, within the parameters take longer to build, and best if developer
it “knows”. knows your market.
Prime use case: FAQ forms of self-service where Best for complex forms of self-service, but
customers are looking for basic can also take automated service further by
information. That may be a limited helping customers define their issue, and
range of queries, but can be a great way “knows” the best time to hand off to an
to start using chatbots. agent.

Table 3 – Chatbot Ecosystem – Type of Developer

Developer type: Self-serve Full-serve

Basic description: Built from off-the-shelf software and These developers handle end-to-end
existing APIs, these chatbots can largely projects, including integrations with your
be created by your IT team, with network and contact center platform. One
minimal involvement from the type would be third party developers who
developer. To be effective, though, IT can customize for your platform and/or
needs to prioritize resources needed to needs of your customers. Another type
build it, test it, create a good UX and would be developers who build chatbots
manage the integrations with your that are native to a particular AI platform
contact center platform. such as Alexa or Siri.
Main benefit: Contact center gets chatbots built Contact center can leverage state-of-the-art
quickly and cheaply, but only for basic AI to greatly enhance service automation
applications. Self-service capabilities will and optimize the use and performance of
be limited, and note that these agents. That said, these projects are
developers won’t have much contact complex and costly, with no guarantees of
center expertise. success.
Prime use case: Ideal for SMBs with limited budget and Conversational chatbots to support both
resources, but wants or needs chatbots text and voice-based self-service, and when
now, and feels the project can be done needed, provide a seamless handoff to
largely in-house. agents for a better CX.

AI in the Contact Center J Arnold & Associates Page 8

Developing a Successful Chatbot Strategy
The analysis to this point lays a foundation of understanding, and is especially important for emerging
technologies like chatbots that have great potential but a short track record. Not only do you need to
select the right type of partner for a chatbot deployment, but you need to be doing it for the right
reasons.

These two factors are closely related, and for them to mesh, you need a chatbot strategy. Based on our
industry research on AI in general, and chatbots in particular, we have distilled the process for
developing this strategy into three basic building blocks:

1. Setting objectives
2. Managing the process
3. Seeing the bigger picture

Building Block #1 - Setting Objectives

This drives the rationale for using chatbots, but given the silver bullet hype surrounding AI, your
objectives need to be carefully considered. Here are five questions to help clarify your thinking, along
with our perspective based on current industry trends.

1. What is your overall, driving objective for using chatbots in the contact
center?
The common, almost knee-jerk response to this question is for saving money. Since AI and chatbots are
largely about automation, this is a natural association to make. More than ever, contact centers are
under pressure to reduce costs, and AI’s appeal is undeniable. That said, this thinking reflects a lack of
understanding of what chatbots can do, as well as just how customer-centric your operation is - or is
not.

Our view is that your overall objective should be about improving the customer experience. Cost savings
are important, but if that’s the driver for a chatbot project – especially your first one – the results will
likely fall short. Even if you achieve that, the benefit will be one-dimensional, and if corners are cut
along the way, CX may suffer. Building your chatbot project around CX, however, reflects a more holistic
approach to what’s most important for the contact center, and by extension, the business overall.

AI in the Contact Center J Arnold & Associates Page 9

 2. What part of the contact center or customer care process are you trying
to improve?
Part of understanding chatbots includes recognizing there is no singular bot that is universal for all
contact center use cases. Across the full spectrum of CX, there is a multitude of touchpoints that
represent moments of truth, and each one can benefit from having some degree of automation. The
more purpose-built the chatbot, the more effective it can be, so identifying a specific need or
application should be a key part of your chatbot strategy.

If your overall objective with chatbots is to improve CX, you’ll need to consider how best to strike a
balance between self-service and live agents. AI may never be able to fully automate customer service,
but agents can’t handle every inquiry either. In that context, a strategic use case would be developing a
chatbot that can comfortably handle a specific set of inquiries, but can then seamlessly extend the
session to a live agent when human interaction is needed.

3. Do you want chatbots only to automate customer service or help agents

perform better?
This builds on the above question, and speaks to expectations around the role automation should play in
your contact center. With these AI applications being so new, every contact center is struggling to figure
this out, but you should not view this as an either/or proposition. To get strategic, business-level value
from chatbots, both need to be addressed.

Building a chatbot strictly to automate one link in the CX chain may work in isolation, but will likely have
unforeseen consequences that impact the performance of agents. All contact centers want customers to
make more use of self-service, and chatbots make it better by being conversational and context-driven.
However, that same capability for service automation will also optimize the use of live agents, allowing
them to resolve problems faster, waste less time on handling routine inquiries, and provide a better
overall CX. Without this holistic view, developing a chatbot just to improve one aspect of self-service is
just a tactical application of new technology.

4. What customer behaviors are you trying to change?

Understanding the full spectrum of CX is challenging for any contact center, and putting yourself in the
customer’s shoes should be part of how you set objectives for chatbots. Doing so allows you to see their
pain points, and that will help define the need for automation. Poor CX is often the result of customers
relying too much on legacy applications such as email, IVR and landlines for service.

Chatbots – and AI in general – provide a fresh start, and a better set of options for self-service.
Customers should only have to deal with live agents as a last resort, but that rarely happens with legacy

AI in the Contact Center J Arnold & Associates Page 10

contact center platforms, so changing their behaviors requires a better experience, especially for self-
service. When thinking this way, chatbots can provide great value, not just for improving CX, but in
moving customers away from legacy channels that create so many problems before they even engage
with your agents.

5. Can chatbots add value to your existing contact center environment?

This is more of an operational question, but in terms of setting objectives, you need to work within
existing constraints. Most contact centers don’t have the luxury of upgrading to an all-IP platform, or
even migrating to the cloud. The current state of your contact center infrastructure may limit how far
you can go with automation, and your chatbot strategy needs to reflect that.

Chatbot applications can support a wide range of scenarios, and you need to choose a partner that can
integrate with what you have. Most contact center platforms represent major sunk costs, so another
way chatbots can add value is by extending the life of your infrastructure, especially if it’s going to be
with you for many years to come.

Building Block #2 - Managing the Process

As with any IT-related project, success with chatbots requires a roadmap. Once you have a clear set of
objectives, a process is needed to guide you through the steps along the way. That’s the second building
block in this analysis, and based on best practices gleaned from our research about chatbots and contact
centers, here are the key elements your process should follow.

Before the project begins

As per the first building block, the more clearly your objectives are defined, the easier it will be to
choose the right partner. The chatbot ecosystem is challenging to navigate, and the landscape analysis
herein is really just a starting point for narrowing down the type of partner that will best align with your
chatbot objectives.

Regardless of the type of partner – such as self-full or full-serve – the key criterion at this stage is
ensuring they understand your space and know how to support it. Many developers are project-driven
and will only be inclined to build a chatbot that fits their framework rather than yours.

If your choice of partner is cost-driven, it will be easy to find generic developers who can quickly turn
around a project and then hand it off to you. That model works well for them, but if you’re only getting a
standalone application, the broader CX benefits will not materialize. In this regard, their ability to
integrate chatbots with contact center platforms – especially yours – should be table stakes.

AI in the Contact Center J Arnold & Associates Page 11

During the project
Developing chatbots is an iterative process, and to effectively leverage AI – and elevate self-service
beyond IVR – your chatbot partner needs to understand that you’re trying to create a conversational
experience with customers. While self-service is driven by automation, the experience must feel natural
and resonate on a human scale. Your chatbot needs to have some personality, and ideally reflect
elements of your brand, such as a distinct style, or look, or attitude, etc.

This is where NLP really comes into play, as the chatbot needs to use language that customers can relate
to. Furthermore, to project human-like qualities, you may want the chatbot to have a playful sense of
humor, either with language or emojis.

Creating this kind of experience also means that the chatbot must be context-driven, and this is where
ML capabilities add value. The more data the chatbot can draw from about the customer’s history and
buying patterns, the more accurately it can address self-service queries, anticipate future problems, and
hand off to the right agent at the right time. In terms of best practices, most chatbot developers won’t
be articulating these needs to you, so to keep the project on track, this is the kind of guidance you need
to provide.

Presuming the project has been managed to plan by this point, there’s an important step before live
deployment. Customers will have little tolerance for self-service applications that don’t work right or
provide an impersonal experience. First-time chatbot projects will rarely be flawless from the start, but
to mitigate these problems, you’ll need to do adequate testing to determine how well the chatbot
responds to a wide range of use cases.

If there’s any doubt about what could go wrong, a quick review of Microsoft’s experience with Tay in
2016 will keep this front and center. Furthermore, you should also use testing metrics to gauge how well
the chatbot is working on your network, as well as integrating with your contact center platform. Testing
vendors are an important subset of the chatbot ecosystem, and aside from vendors that your chatbot
developer may recommend, there are several resources in the AI community where you can easily do
your own search.

After your chatbot has been deployed

For contact centers doing their first chatbot project, customer response is impossible to predict, and
managing the overall process needs to extend to this stage. The requisite testing outlined above only
measures internal success, but the true test is in the field, and here are some best practices to follow.

AI in the Contact Center J Arnold & Associates Page 12

First and foremost, you need to be transparent with customers about their engagement with the
chatbot. Some customers may think this is just another web chat session, while others may not have
encountered this mode of self-service before. Regardless, they need to know they’re engaging with a
chatbot, especially if it’s totally automated and AI-driven. Depending on the type of service being
offered, many other chatbot applications will have some degree of live agent involvement.

Whatever form your chatbot takes, it’s critical to not misrepresent the situation. If your chatbot makes
advanced use of NLP, it could engage in a highly personal manner, leading the customer to think they’re
interacting with a live agent. However, it just takes one tiny miss for the customer to realize they’re
engaging instead with a machine, and if they feel duped, you’ll never get them back for self-service, and
the fallout could be much worse.

Another important post-deployment best practice is the use of performance metrics. This may be the
most valuable criterion for success since chatbots are not well understood, and like KPIs, numbers make
the impact real for all the stakeholders involved. Keep in mind that you should be tracking performance
in two ways – first, how well the chatbot is supporting self-service queries, and secondly, how it is
impacting agent performance. To illustrate, here are key examples for each.

Performance metrics for self-service

• Usage frequency for each self-service category – in NLP parlance, these would be “intents”
• Efficiency of query resolution – how many steps needed, how quickly done, what percentage of
queries could be handled end-to-end by the chatbot, etc.
• Accuracy – how well queries are understood, how many times questions needed repeating, etc.
• Ability to make a seamless handoff to agents – percentage of sessions that remained continuous
• Efficacy for routing calls to the right agent – percentage of queries that only need one agent
• Linking self-service outcomes to customer satisfaction ratings
• Usage reduction for other self-service options such as IVR, email or web chat

Performance metrics for live agents

• Speed and quality of query resolution after getting handoff from chatbot
• Profile of routine versus complex queries handled by agents before chatbot and then after
• Incidence of queries that come direct to agents without first trying self-service options
• Basic productivity – percentage more calls agent is able to handle with chatbot being used
• Number of agents required to handle live queries
• Impact on overall CX – ability to shorten sales cycle, CSAT, NPS, etc.

AI in the Contact Center J Arnold & Associates Page 13

Building Block #3 – Seeing the Bigger Picture

This building block is about stepping back and viewing the overall process in the broader context of
digital transformation, and how businesses must adapt to changing technology. AI is one of those
transformative technologies, and deploying chatbots in the contact center is more than just a routine IT
project. When contact center decision-makers reach the point of wanting chatbots, they are usually
anxious for a few reasons.

First is the recognition that CX is getting worse, and they don’t have the luxury of hiring more agents or
totally modernizing their infrastructure. Not only is their job on the line if things don’t improve, but so is
the stability of the contact center for remaining an in-house operation. A second source of anxiety is the
hype around AI, not just for understanding its nature, but also the false hope it offers as a panacea for
everything technology-related.

To whatever extent this reflects your current situation, our research shows that contact center decision-
makers are coming to chatbots with many questions and great uncertainty about how to move forward.
That is a key challenge this white paper seeks to address, with the intention being for decision-makers to
take a proactive stance rather than being reactive. To make this process and your over chatbot strategy
successful, here are two bigger picture elements to keep in mind.

Big picture element #1 – focus on a manageable problem set

We’re a long way from AI automating everything – if ever – and for a first-time deployment, it’s more
important for the chatbot to perform really well than to take on the most complex customer challenges.
The goal is not to be so good that agents are no longer needed; rather to effectively handle less complex
queries via self-service so agents can focus on problems that only they can address.

Research shows that Millennials are generally quite receptive to engaging with chatbots, and if your
current contact center platform is serving customers reasonably well, they should also be receptive to
your chatbots. There is some implied goodwill here that will be validated if their initial CX is good, so
getting it right the first time is crucial. This is especially true since less tech-savvy customers may not be
ready to trust chatbots, so not only must the CX be effective, but it must be easy to use. Anything less
will mean fewer second chances, so there’s no need to be overly ambitious for the first project.

Big picture element #2 – AI is more than just self-service

Our view is that self-service is the right starting point with chatbots, not just for improving the
automated side of CX, but also in optimizing the utilization of live agents for a more complete CX. Even if
your initial chatbot project is small-scale, there is a much bigger scope to consider. In time, AI’s impact

AI in the Contact Center J Arnold & Associates Page 14

will be felt across our business and personal lives, and for the contact center, the same will apply across
the entire customer journey.

As such, if CX is an important driver of your enterprise business strategy, then this first chatbot project
should be viewed as the starting point for using AI for the rest of the customer journey. Beyond self-
service, chatbots could be developed to make call routing more accurate, to enable agents to respond
faster, to enable supervisors to proactively coach agents, to automate the completion applications or
trouble tickets, just to name a few.

Along these lines, it’s also important to keep in mind that both NLP and ML are heuristic – they are built
to constantly learn, and improve as the data sets get larger and as the patterns become more
established. This is a departure from the static, standalone technologies contact centers have been using
for so long. Chatbots – and AI in general – represent a new model, and an investment that can improve
CX, not just after the first project, but continually over time.

Conclusion – Moving Forward

Bringing AI into the contact center via chatbots is an important step for leveraging today’s technology to
address the needs of today’s customers. However, there is more to consider than trying to automate
one part of the CX, or determining whether to build a chatbot in-house or go with a partner.

Customer expectations are changing, and it’s not enough to take a project-driven approach where a
chatbot is built like a point solution to address a specific need. This white paper takes a broader view by
presenting a strategic approach that takes a holistic view of the full CX. Our view is based on the reality
that AI cannot yet fully automate CX, and since the broad range of needs for customer care cannot be
addressed entirely by agents, contact centers face a fundamental challenge.

Chatbots present a viable solution by providing a better form of automation, but for the foreseeable
future, contact centers will need to support a mix of automated self-service and assisted service with
live agents. The strategy, then, needs to be about finding the right balance between automation and
agents, with the overall goal being to improve CX. Those outcomes cannot be achieved with a narrow,
tactical focus on chatbots.

To move this strategy forward, you’ll need the right type of partner. This will not be easy given the
complexity of the chatbot landscape, and how quickly the ecosystem is evolving. At minimum, your
chatbot partner’s value proposition should align with the strategy outlined herein, and one indication
will be how realistic their messaging is. This white paper should provide enough background to
determine the hype volume, and if too high, you need to move on.

AI in the Contact Center J Arnold & Associates Page 15

Another indication will be how well they understand the role of chatbots for customer care, along with a
demonstrated ability to integrate those chatbots with your contact center platform. Unless your
platform will soon be changing, that capability will be critical, especially if you view chatbots as an
ongoing journey for improving CX. Over time, customers will expect more forms of intelligent
automation, and they’ll expect to engage with AI across every channel. As such, no matter what your
starting point is with chatbots, there will be a lot of ground to cover, and that’s the big picture to keep in
mind for finding the right partner.

J Arnold & Associates, an independent technology analyst practice, produced this white paper, which was
sponsored by Cisco. The contents herein reflect our conclusions drawn from ongoing research about the
contact center market, AI, and chatbots, along with industry-based interviews specific to this white
paper. For more information, please contact us by email: [email protected].

AI in the Contact Center J Arnold & Associates Page 16