Assignment

Chapter 2Artificial Intelligence Concepts, Drivers, Major Technologies, and Business Applications

Learning Objectives • Understand the concepts of artificial intelligence (AI)

• Become familiar with the drivers, capabilities, and benefits of AI

• Describe human and machine intelligence

• Describe the major AI technologies and some derivatives

• Discuss the manner in which AI supports decision making

• Describe AI applications in accounting

• Describe AI applications in banking and financial services

• Describe AI in human resource management

• Describe AI in marketing

• Describe AI in production-operation management

Artificial intelligence (AI), which was a curiosity for generations, is rapidly developing into a

major applied technology with many applications in a variety of fields. OpenAI’s (an AI research

institution described in Chapter 14) mission states that AI will be the most significant technology

ever created by humans. AI appears in several shapes and has several definitions. In a crude way,

it can be said that AI’s aim is to make machines exhibit intelligence as close as possible to what

people exhibit, hopefully for the benefit of humans. The latest developments in computing

technologies drive AI to new levels and achievements. For example, IDC Spending

Guide (March 22, 2018) forecasted that worldwide spending on AI will reach $19.1 billion in

2018. It also predicted annual double-digit spending growth for the near future. According

to Sharma (2017), China expects to be the world leader in AI, with a spending of $60 billion in

2025. For the business value of AI, see Greig (2018).

In this chapter, we provide the essentials of AI, its major technologies, its support for decision

making, and a sample of its applications in the major business functional areas.

The chapter has the following sections:

1. 2.1 Opening Vignette: INRIX Solves Transportation Problems 74 2. 2.2 Introduction to Artificial Intelligence 76 3. 2.3 Human and Computer Intelligence 83 4. 2.4 Major AI Technologies and Some Derivatives 87 5. 2.5 AI Support for Decision Making 95 6. 2.6 AI Applications in Accounting 99 7. 2.7 AI Applications in Financial Services 101 8. 2.8 AI in Human Resource Management (HRM) 105 9. 2.9 AI in Marketing, Advertising, and CRM 107

10. 2.10 AI Applications in Production-Operation Management (POM) 110

2.1 Opening Vignette: INRIX Solves Transportation Problems

The Problem

Traffic congestion is an ever-increasing problem in many large metropolitan areas. Drivers may

spend several hours on the roads each day. In addition, air pollution is increasing, and more

accidents are occurring.

The Solution

INRIX corporation (inrix.com) enables drivers to get real-time traffic information. They can

download the INRIX-XD Traffic App for iOS and Android. The information provided is

generated by a predictive analysis of massive data obtained from consumers and the

environment (e.g., road construction, accidents). Information sources include:

• Traffic data collected by helicopters, drones, and so on, which include real-time traffic

flow and accident information.

• Information provided by participating delivery companies and over 100 million

anonymous volunteer drivers, who have GPS-enabled smartphones, all reporting in real

time.

• Information provided by traffic congestion reports (e.g., delays due to road

maintenance).

INRIX processes the collected information with proprietary analytical tools and formulas, some

of which are AI-based. The processed information is used to generate traffic predictions. For

example, it creates a picture of anticipated traffic flows and delays for the next 15 to 20 minutes,

few hours, and few days for many locations. These predictions enable drivers to plan their

optimal routes. As of 2018, INRIX had offered global coverage in 45 countries and in many

major cities, and the company analyzed traffic information from over 100 sources. This service is

combined with digital maps. In Seattle, for example, traffic information is disseminated via

smartphones and color codes on billboards along the freeways. Smartphones also display

estimated times for the roads to be either clear or jammed. As of 2018, the company had covered

over 5,000,000 miles of highways worldwide, delivering upon request the best recommended

routes to use, all in real time.

The INRIX system provides information (or recommendations) for decisions such as:

• Optional routes for delivery vehicles and other travelers to take

• The best time to go to work or to other places from a given location

• Information for rerouting a trip to avoid encountering a traffic jam that just occurred

• Fees to be paid on highways, which are based on traffic conditions and time of the day

The technologies used to collect data are:

• Closed-circuit TV cameras and radar that monitor traffic conditions

• Public safety reports and traffic information

• Information about freeway access and departure flows

• Technologies that measure toll collection queues

• Magnetic sensing detectors embedded under the road surface (expensive)

• Smartphones and other data collection devices that gather data for INRIX

The information is processed by several AI techniques such as expert systems;

see Chapter 12 and different analytical models (such as simulation).

Several of the sources of information are connected to the company via the Internet of Things

(IoT) (Chapter 13). According to its Web site, INRIX has partnered with Clear Channel Radio to

broadcast real-time traffic data directly to vehicles via Ln Carr or via portable navigation

systems, broadcast media, and wireless and Internet-based services. Clear Channel’s Total

Traffic Network is available in more than 125 metropolitan areas in four countries

(inrix.com/press-releases/2654/). In 2018, the system was installed in over 275 million cars and

data collection devices. The system collects real-time traffic information from these devices.

The Results

In addition to being used by individual drivers, the processed information is shared by

organizations and city planners for making planning decisions. Also, less traffic congestion has

been recorded in participating cities, which results in less pollution, fewer road accidents, and

increased productivity by happier employees who spend less time commuting.

The INRIX Traffic App (available for download at inrix.com/mobile-apps) is suitable for all

smartphones; it supports 10 languages, including English, French, and Spanish. For the free

INRIX traffic features, see inrixtraffic.com/features. For interesting case studies,

see inrix.com/case-studies.

As of 2016, INRIX had released an improved traffic app that uses both AI and crowdsourcing

(Chapter 11) to support drivers’ decisions as to the best route to take (Korosec, 2016). The AI

technology analyzes drivers’ historical activities to infer their future activities.

NOTE:

Popular smartphone apps, such as Waze and Moovit, provide navigation and data collection

similar to INRIX.

Sources: Based on inrix.com, Gitlin (2016), Korosec (2016), and inrix.com/mobile-apps (all accessed June 2018).

Questions for the Opening Vignette

1. Explain why traffic may be down while congestion is up (see the London case at inrix.com/uk-highways-agency/).

2. How does this case relate to decision support? 3. Identify the AI elements in this system.

4. Identify developments related to AI by viewing the company’s press releases from the most recent four months at inrix.com/press-releases. Write a report.

5. According to Gitlin (2016), INRIX’s new mobile traffic app is a threat to Waze. Explain why.

6. Go to sitezeus.com/data/inrix and describe the relationship between INRIX and Zeus. View the 2:07 min. video at sitezeus.com/data/inrix/. Why is the system in the video

called a “decision helper”?

What We Can Learn from the Vignette

The INRIX case illustrates to us how the collection and analysis of a very large amount of

information (Big Data) can improve vehicles’ mobility in large cities. Specifically, by collecting

information from drivers and other sources instead of only from expensive sensors, INRIX has

been able to optimize mobility. This has been achieved by supporting decisions made by drivers

and by analyzing traffic flows. INRIX is also using applications from the IoT to connect vehicles

and devices with its computing system. This application is one of the building blocks of smart

cities (see Chapter 13). The analysis of the collected data is done by using powerful algorithms,

some of which are applications of AI.

2.2 Introduction to Artificial Intelligence

We would all like to see computerized decision making being simpler, easier to use,

more intuitive, and less threatening. And indeed, efforts have been made over time to

simplify and automate several tasks in the decision-making process. Just think of the

day that refrigerators will be able to measure and evaluate their contents and place

orders for goods that need replenishment. Such a day is not too far in the future, and

the task will be supported by AI.

CIO Insight projected that by 2035, intelligent computer technologies will result in $5–

$8.3 trillion in economic value (see cioinsight.com/blogs/how-ai-will-impact-the-global-

economy.html). Among the technologies listed as intelligent ones are the IoT, advanced

robotics, and self-driven vehicles, all described in this book. Gartner, a leading

technology consulting firm, listed the following in its 2016 and 2017 Hype Cycles for

Emerging Technologies: expert advisors, natural language questions and answering,

commercial drones, smart workspaces, IoT platforms, smart data discovery, general-

purpose machine intelligence, and virtual personal assistants. Most are described or

cited in this book (see also Greengard, 2016). For the history of AI, see Zarkadakis

(2016) and en.wikipedia.org/wiki/History_of_artificial_intelligence.

Definitions

Artificial intelligence has several definitions (for an overview see Marr 2018); however, many

experts agree that AI is concerned with two basic ideas: (1) the study of human thought

processes (to understand what intelligence is) and (2) the representation and duplication of those

thought processes in machines (e.g., computers, robots). That is, the machines are expected to

have humanlike thought processes.

One well-publicized definition of AI is “the capabilities of a machine to imitate intelligent

human behavior” (per Merriam-Webster Dictionary). The theoretical background of AI is based

on logic, which is also used in several computer science innovations. Therefore, AI is considered

a subfield of computer science. For the relationship between AI and logic,

see plato.stanford.edu/entries/logic-ai.

A well-known early application of artificial intelligence was the chess program hosted at IBM’s

supercomputer (Deep Blue). The system beat the famous world champion, Grand Master Garry

Kasparov.

AI is an umbrella term for many techniques that share similar capabilities and characteristics. For

a list of 50 unique AI technologies, see Steffi (2017). For 33 types of AI,

see simplicable.com/new/types-of-artificial-intelligence.

Major Characteristics of AI Machines

There is an increasing trend to make computers “smarter.” For example, Web 3.0 supposes to

enable computerized systems that exhibit significantly more intelligence than Web 2.0. Several

applications are already based on multiple AI techniques. For example, the area of machine

translation of languages is helping people who speak different languages to collaborate as well as

to buy online products that are advertised in languages they do not speak. Similarly, machine

translation can help people who know only their own language to converse with people speaking

other languages and to make decisions jointly in real time.

Major Elements of AI

As described in Chapter 1, the landscape of AI is huge, including hundreds or more components.

We illustrate the foundation and the major technologies in Figure 2.1. Notice that we divide them

into two groups: Foundations, and Technologies and Applications. The major technologies will

be defined later in this chapter and described throughout this book.

Figure 2.1 The Functionalities and Applications of Artificial Intelligence.

AI Applications

The technologies of AI are used in the creation of a large number of applications.

In Sections 2.6–2.10, we provide a sampler of applications in the major functional areas of

business.

Example Smart or intelligent applications include those that can help machines to answer customers’

questions asked in natural languages. Another area is that of knowledge-based systems which

can provide advice, assist people to make decisions, and even make decisions on their own. For

example, such systems can approve or reject buyers’ requests to purchase online (if the buyers

are not preapproved or do not have an open line of credit). Other examples include the automatic

generating of online purchasing orders and arranging fulfillment of orders placed online. Both

Google and Facebook are experimenting with projects that attempt to teach machines how to

learn and support or even make autonomous decisions. For smart applications in enterprises,

see Dodge (2016), Finlay (2017), McPherson (2017), and Reinharz (2017). For how AI solutions

are used to facilitate government services, see BrandStudio (2017).

AI-based systems are also important for innovation and are related to the areas of analytics and

Big Data processing. One of the most advanced projects in this area is IBM Watson Analytics

(see Chapter 6). For comprehensive coverage of AI, including definitions and its history,

frontiers, and future, see Kaplan (2016). NOTE:

In January 2016, Mark Zuckerberg, the CEO of Facebook, announced publicly that his goal for

2016 was to build an AI-based assistant to help with his personal and business activities and

decisions. Zuckerberg was teaching a machine to understand his voice and follow his basic

commands as well as to recognize the faces of his friends and business partners. Personal

assistants are used today by millions of people (see Chapter 12).

Example: Pitney Bowes Is Getting Smarter with AI Pitney Bowes Inc. is a U.S.-based global business solutions provider in areas such as product

shipments, location intelligence, customer engagement, and customer information management.

The company powers billions of physical and digital transactions annually across the connected

and borderless world of commerce.

Today, at Pitney Bowes, shipping prices are determined automatically based on the dimensions,

weight, and packaging of each package. The fee calculations create data that are fed into AI

algorithms. The more data processed, the more accurate are the calculations (a machine-learning

characteristic). The company estimates a 25 percent improvement in calculations achieved from

their algorithms. This gives Pitney Bowes an accurate base for pricing, better customer

satisfaction, and improved competitive advantage.

Major Goals of AI

The overall goal of AI is to create intelligent machines that are capable of executing a variety of

tasks currently done by people. Ideally, AI machines should be able to reason, think abstractly,

plan, solve problems, and learn.

Some specific goals are to:

• Perceive and properly react to changes in the environment that influence specific business

processes and operations.

• Introduce creativity in business processes and decision making.

Drivers of AI

The use of AI has been driven by the following forces:

• People’s interest in smart machines and artificial brains

• The low cost of AI applications versus the high cost of manual labor (doing the same work)

• The desire of large tech companies to capture competitive advantage and market share of

the AI market and their willingness to invest billions of dollars in AI

• The pressure on management to increase productivity and speed

• The availability of quality data contributing to the progress of AI

• The increasing functionalities and reduced cost of computers in general

• The development of new technologies, particularly cloud computing

Benefits of AI

The major benefits of AI are as follows:

• AI has the ability to complete certain tasks much faster than humans.

• The consistency of the completed AI work can be much better than that of humans. AI

machines do not make mistakes.

• AI systems allow for continuous improvement projects.

• AI can be used for predictive analysis via its capability of pattern recognition.

• AI can manage delays and blockages in business processes.

• AI machines do not stop to rest or sleep.

• AI machines can work autonomously or be assistants to humans.

• The functionalities of AI machines are ever increasing.

• AI machines can learn and improve their performance.

• AI machines can work in environments that are hazardous to people.

• AI machines can facilitate innovations by human (i.e., support research and development

[R&D]).

• No emotional barriers interfere with AI work.

• AI excels in fraud detection and in security facilitations.

• AI improves industrial operations.

• AI optimizes knowledge work.

• AI increases speed and enables scale.

• AI helps with the integration and consolidating of business operations.

• AI applications can reduce risk.

• AI can free employees to work on more complex and productive jobs.

• AI improves customer care.

• AI can solve difficult problems that previously were unsolved (Kharpal, 2017).

• AI increases collaboration and speeds up learning.

These benefits facilitate competitive advantage as reported by Agrawal (2018). NOTE:

Not all AI systems deliver all these benefits. Specific systems may deliver only some of them.

The capability of reducing costs and increasing productivity may result in large increases in

profit (Violino, 2017). In addition to benefiting individual companies, AI can dramatically

increase a country’s economic growth, as it is doing in Singapore.

Figure 2.2 Cost of Human Work versus the Cost of AI Work.

Examples of AI Benefits

The following are typical benefits of AI in various areas of applications:

1. The International Swabs and Derivatives Association (ISDA) uses AI to eliminate tedious activities in contract procedures. For example, by using optical character

recognition (OCR) integrated with AI, ISDA digitizes contracts and then defines,

extracts, and archives the contracts.

2. AI is starting to revolutionize business recruitment by (1) conducting more efficient and fairer candidate screening, (2) making better matches of candidates to jobs, and (3)

helping safeguard future talent pipelines for organizations. For details, see SMBWorld

Asia Editors (2017) and Section 2.8.

3. AI is redefining management. According to Kolbjørnsrud et al. (2016), the following five practices result from the use of AI:

▪ It can perform routine administrative tasks.

▪ Managers can focus on the judgment portions of work.

▪ Intelligent machines are treated as colleagues (i.e., managers trust the advice

generated by AI). In addition, there is people–machine collaboration

(see Chapter 11).

▪ Managers concentrate on creative abilities that can be supported by AI machines.

▪ Managers are developing social skills, which are needed for better collaboration,

leadership, and coaching.

4. Accenture Inc. developed AI-powered solutions using natural language processing (NLP) and image recognition to help blind people in India improve the way that they

can experience the world around them. This enables them to have a better life, and

those who work can work better, faster, and do jobs that are more challenging.

5. Ford Motor Credit uses machine learning to spot overlooked borrowers. In addition, it uses machine learning to help its underwriters better understand loan applicants. The

program helps the productivity of both underwriters and overlooked applicants.

Finally, the system predicts potential borrowers’ creditworthiness, thus minimizing

losses for Ford.

6. Alastair Cole uses data collected from several sources with IBM Watson to predict what customers are expecting from the company. The generated data are used for

supporting more efficient business decisions.

7. Companies are building businesses around AI. There are many examples of start-ups or existing companies that are attempting to create new businesses.

Two areas in which large benefits have already been reaped are customer experience and

enjoyment. According to a global survey reported by CMO Innovation Editors (2017), 91

percent of top-performing companies deployed AI solutions to support customer experience.

Some Limitations of AI Machines

The following are the major limitations of AI machines:

• Lack human touch and feel

• Lack attention to non-task surroundings

• Can lead people to rely on AI machines too much (e.g., people may stop to think on their

own)

• Can be programmed to create destruction (see discussion in Chapter 14)

• Can cause many people to lose their jobs (see Chapter 14)

• Can start to think by themselves, causing significant damage (see Chapter 14)

Some of the limitations are diminishing with time. However, risks exist. Therefore, it is

necessary to properly manage the development of AI and try to minimize risks.

What AI Can and Cannot Do

The limitations just identified constrain the capabilities of commercial AI. For example, it could

cost too much to be commercially used. Ng (2016) provides an assessment of what AI was able

to do by 2016. This is important for two reasons: (1) executives need to know what AI can do

economically and how companies can use it to benefit their business and (2) executives need to

know what AI cannot economically do.

AI is already transforming Web search, retailing and banking services, logistics, online

commerce, entertainment, and more. Hundreds of millions of people use AI on their smartphones

and in other ways. However, according to Ng (2016), applications in these areas are based on

how simple input is converted to simple output as a response; for example, in automatic loan

approval, the input is the profile of the applicant and the output will be an approval or rejection.

Applications in these areas are normally fully automated. Automated tasks are usually repetitive

and done by people with short periods of training. AI machines depend on data that may be

difficult to get (e.g., belong to someone else) or inaccurate. A second barrier is the need for AI

experts, who are difficult to find and/or expensive to hire. For other barriers, see Chapter 14.

Three Flavors of AI Decisions

Staff (2017) divided the capabilities of AI systems into three levels: assisted, autonomous, and

augmented.

Assisted Intelligence

This is equivalent mostly to weak AI, which works only in narrow domains. It requires clearly

defined inputs and outputs. Examples are some monitoring systems and low-level virtual

personal assistants (Chapter 12). Such systems and assistants are used in our vehicles for giving

us alerts. Similar systems can be used in many healthcare applications (e.g., monitoring,

diagnosing).

Autonomous AI

These systems are in the realm of the strong AI but in a very narrow domain. Eventually, a

computer will take over many tasks, automating them completely. Machines act as experts and

have absolute decision-making power. Pure robo-advisors (Chapter 12) are examples of such

machines. Autonomous vehicles and robots that can fix themselves are also good examples.

Augmented Intelligence

Most of the existing AI applications, which are between assisted and autonomous, are referred to

as augmented intelligence (or intelligence augmentation). Their technology can augment

computer tasks to extend human cognitive abilities (see Chapter 6 on cognitive computing),

resulting in high performance, as described in Technology Insight 2.1.

Technology Insight 2.1 Augmented Intelligence The idea of combining the performance of people and machines is not new. In this section, we

discuss combining (augmenting) human abilities with powerful machine intelligence—not

replacing people, which autonomous AI does, but extending human cognitive abilities. The result

is the ability of humans to solve more complex problems, as in the opening vignette to Chapter 1.

Computers have provided data to help people solve problems for which no solution had been

available. Padmanabhan (2018) specifies the following differences between traditional and

augmented AI:

1. Augmented machines extend human thinking capabilities rather than replace human decision making. These machines facilitate creativity.

2. Augmentation excels in solving complex human and industry problems in specific domains in contrast with strong, general AI machines, which are still in development.

3. In contrast with a “black box” model of some AI and analytics, the augmented intelligence provides insights and recommendations, including explanations.

4. In addition, augmented technology can offer new solutions by combining existing and discovered information in contrast to assisted AI that identifies problems or symptoms and

suggests predetermined known solutions.

Padmanabhan (2018) and many others believe that at the moment, augmented AI is the best

option to deal with practical problems and transform organizations to be “smarter.”

In contrast with autonomous AI, which describes machines with a wide range of cognitive

abilities (e.g., driverless cars), augmented intelligence has only a few cognitive abilities.

Examples of Augmented Intelligence Staff (2017) provides the following areas for which AI is useful:

• CYBERCRIME FIGHTING. For example, AI can identify forthcoming attacks and

suggest solutions.

• E-COMMERCE DECISIONS. AI marketing tools can make testing results 100 times

faster, and adapt the layout and response functions of a Web site to users. Machines also

make recommendations, and marketers can accept or reject them.

• HIGH-FREQUENCY STOCK MARKET TRADING. This process can be done either

completely autonomously or in some cases with human control and calibration.

Questions for Discussion 1. What is the basic premise of augmented intelligence? 2. List the major differences between augmented intelligence and assisted AI applications. 3. What are some benefits of augmented intelligence? 4. How does the technology relate to cognitive computing?

Artificial Brain

The artificial brain is a people-made machine that is desired to be as intelligent, creative, and

self-aware as humans. To date, no one has been able to create such a machine;

see artificialbrains.com. A leader in this area is IBM. IBM and the U.S. Air Force have built a

system equivalent to 64 million artificial neurons that aims to reach 10 billion neurons by 2020.

Note that a human brain contains about 100 billion neurons. The system tries to imitate a

biological brain and be energy efficient. IBM’s project is called TrueNorth or BlueBrain, and it

learns from humans’ brains. Many believe that it will be a long and slow process for AI

machines to be as creative as people (e.g., Dormehl, 2017).

Section 2.2 Review Questions

1. Define AI. 2. What are the major aims and goals of AI? 3. List some characteristics of AI. 4. List some AI drivers. 5. List some benefits of AI applications. 6. List some AI limitations. 7. Describe the artificial brain. 8. List the three flavors of AI and describe augmentation.

2.3 Human and Computer Intelligence

AI usage is growing rapidly due to its increased capabilities. To understand AI, we

need to first explore the meaning of intelligence.

What Is Intelligence?

Intelligence can be considered to be an umbrella term and is usually measured by an IQ test.

However, some claim that there are several types of intelligence. For example, Dr. Howard

Gardner of Harvard University proposed the following types of intelligence:

• Linguistic and verbal

• Logical

• Spatial

• Body/movement

• Musical

• Interpersonal

• Intrapersonal

• Naturalist

Thus, intelligence is not a simple concept.

Content of Intelligence

Intelligence is composed of reasoning, learning, logic, problem-solving ability, perception, and

linguistic ability.

Obviously, the concept of intelligence is not simple.

Capabilities of Intelligence

To understand what artificial intelligence is, it is useful to first examine those abilities that are

considered signs of human intelligence:

• Learning or understanding from experience

• Making sense out of ambiguous, incomplete, or even contradictory messages and

information

• Responding quickly and successfully to a new situation (i.e., using the most correct

responses)

• Understanding and inferring in a rational way, solving problems, and directing conduct

effectively

• Applying knowledge to manipulate environments and situations

• Recognizing and judging the relative importance of different elements in a situation

AI attempts to provide some, hopefully all, of these capabilities, but in general, it is still not

capable of matching human intelligence.

How Intelligent Is AI?

AI machines have demonstrated superiority over humans in playing complex games such as

chess (beating the world champion), Jeopardy! (beating the best players), and Go (a complex

Chinese game) whose top players were beaten by a computer using the well-known program,

Google’s DeepMind (see Hughes, 2016). Despite these remarkable demonstrations (whose cost

is extremely high), many AI applications still show significantly less intelligence than humans.

Comparing Human Intelligence with AI

Several attempts have been made to compare human intelligence with AI. There is difficulty in

doing so because it is a multidimensional situation. A comparison is presented in Table 2.1.

Table 2.1 Artificial Intelligence versus Human Intelligence

Area AI Human

Execution Very fast Can be slow

Emotions Not yet Can be

positive or

negative

Computation

speed

Very fast Slow, may

have trouble

Imagination Only what is

programmed for

Can expand

existing

knowledge

Answers to

questions

What is in the

program

Can be

innovative

Area AI Human

Flexibility Rigid Large, flexible

Foundation A binary code Five senses

Consistency High Variable, can

be poor

Process As modeled Cognitive

Form Numbers Signals

Memory Built in, or

accessed in the

cloud

Use of

content and

scheme

memory

Brain Independent Connected to

a body

Creativity Uninspired Truly creative

Area AI Human

Durability Permanent, but

can get obsolete

if not updated

Perishable,

but can be

updated

Duplication,

documentation,

and dissemination

Easy Difficult

Cost Usually low and

declining

Maybe high

and

increasing

Consistency Stable Erratic at

times

Reasoning

process

Clear, visible Difficult to

trace at times

Perception By rules and

data

By patterns

Area AI Human

Figure missing

data

Usually cannot Frequently

can

For additional comparisons and who had the advantage in which area,

see www.dennisgorelik.com/ai/ComputerintelligenceVsHumanIntelligence.htm.

Measuring AI

The Turing Test is a well-known attempt to measure the intelligence level of AI machines.

Turing Test: The Classical Measure of Machine Intelligence

Alan Turing designed a test known as the Turing Test to determine whether a computer exhibits

intelligent behavior. According to this test, a computer can be considered smart only when a

human interviewer asking the same questions to both an unseen human and an unseen computer

cannot determine which is which (see Figure 2.3). Note that this test is limited to a question-and-

answer (Q&A) mode.

Figure 2.3 A Pictorial Representation of the Turing Test

To pass the Turing Test, a computer needs to be able to understand a human language (NLP), to

possess human intelligence (e.g., have a knowledge base), to reason using its stored knowledge,

and to be able to learn from its experiences (machine learning).

NOTE:

The $100,000 Leobner prize is waiting for the person or persons who develop software that is

truly intelligent (i.e., passing the Turing Test).

NOTE:

The $100,000 Leobner prize is waiting for the person or persons who develop software that is

truly intelligent (i.e., passing the Turing Test).

Other Tests

Over the years, there have been several other proposals of how to measure machine intelligence.

For example, improvements in the Turing Test appear in several variants. Major U.S. universities

(e.g., University of Illinois, Massachusetts Institute of Technology [MIT], Stanford University)

are engaged in studying the IQ of AI. In addition, there are several other measuring tests. Let’s

examine one test in Application Case 2.1.

Application Case 2.1 How Smart Can a Vacuum Cleaner Be?

If you do not know it, vacuum cleaners can be smart. Some of you may use the Roomba from

iRobot. This vacuum cleaner can be left alone to clean floors, and it exhibits some intelligence.

However, in smart homes (Chapter 13), we expect to see even smarter vacuum cleaners. One is

Roboking Turbo Plus from LG in Korea. Researchers at South Korea’s Seoul National

University Robotics and Intelligent System Lab studied the Roboking and verified that its deep-

learning algorithm makes it as intelligent as a six- or seven-year-old child. If we have self-

driving cars, why can’t we have a self- driving vacuum cleaner, which is much simpler than a

car. The cleaner needs only to move around an entire room. To do so, the machine needs to “see”

its location in a room and identify obstacles in front of it. Then the cleaner’s knowledge base

needs to find what is the best thing to do (given worked in the past). This is basically what many

AI machines’ sensors, knowledge bases, and rules do. In addition, the AI machine needs to learn

from its past experience (e.g., what it should not do because it did not work in the past).

Roboking is equipped with LG’s Deep Thin QTM AI program, which enables the vacuum

cleaner to figure out the nature of an encountered obstacle. The program tells it to go around

furniture, wait for a dog to move, or stop. So, how intelligent is the machine? To answer this

question, the Korean researchers developed 100 metrics and tested vacuum cleaners that were

boasted as autonomous. The performance of the tested cleaners was divided into three levels

based on their performance regarding the 100 metrics. The levels were as intelligent as a dolphin,

as intelligent as an ape, and as intelligent as a six-to-seven-year-old child. The study confirmed

that Roboking performed tasks at the upper level of machine intelligence.

Sources: Compiled from Fuller (2017) and webwire.com/ViewPressRel.asp?aId=211017 news dated July 18, 2017.

Questions for Case 2.1

1. How did the Korean researchers determine the performance of the vacuum cleaners? 2. If you own (or have seen) the Roomba, how intelligent do you think it is? 3. What capability can be generated by the deep learning feature? (You need to do some

research.)

4. Find recent information about LG’s Roboking. Specifically, what are the newest improvements to the product?

In conclusion, it is difficult to measure the level of intelligence of humans as well as that of

machines. Doing so depends on the circumstances and the metrics used. Regardless of the

determination of how intelligent a machine is, AI exhibits a large number of benefits as

described earlier.

It is important to note that the capabilities of AI are increasing with time. For example, an

experiment at Stanford University (Pham, 2018) found that AI programs at Microsoft and

Alibaba Co. have scored higher than hundreds of individual people at reading comprehension

tests. (Of course, these are very expensive AI programs.) For a discussion of AI versus human

intelligence, see Carney (2018).

Section 2.3 Review Questions

1. What is intelligence? 2. What are the major capabilities of human intelligence? Which are superior to that of AI

machines?

3. How intelligent is AI? 4. How can we measure AI’s intelligence? 5. What is the Turing Test and what are its limitations? 6. How can one measure the intelligence level of a vacuum cleaner?

2.4 Major AI Technologies and Some Derivatives

The AI field is very broad; we can find AI technologies and applications in hundreds of

disciplines ranging from medicine to sports. Press (2017) lists 10 top AI technologies

similar to what is covered in this book. Press also provides the status of the life cycle

(ecosystem phase) of the technologies. In this section, we present some major AI

technologies and their derivatives as related to business. The selected list is illustrated

in Figure 2.4. Figure 2.4 The Major AI Technologies

Intelligent Agents

An intelligent agent (IA) is an autonomous, relatively small computer software program that

observes and acts upon changes in its environment by running specific tasks autonomously. An

IA directs an agent’s activities to achieve specific goals related to the changes in the surrounding

environment. Intelligent agents may have the ability to learn by using and expanding the

knowledge embedded in them. Intelligent agents are effective tools for overcoming the most

critical burden of the Internet information overload and making computers more viable decision

support tools. Interest in using intelligent agents for business and e-commerce started in the

academic world in the mid-1990s. However, only since 2014, when the capabilities of IA

increased remarkably, have we started to see powerful applications in many areas of business,

economics, government, and services.

Initially, intelligent agents were used mainly to support routine activities such as searching for

products, getting recommendations, determining products’ pricing, planning marketing,

improving computer security, managing auctions, facilitating payments, and improving inventory

management. However, these applications were very simple, using a low level of intelligence.

Their major benefits were increasing speed, reducing costs, reducing errors, and improving

customer service. Today’s applications, as we will see throughout this chapter, are much more

sophisticated.

Example 1: Virus Detection Program A simple example of an intelligent software agent is a virus detection program. It resides in a

computer, scans all incoming data, and removes found viruses automatically while learning to

detect new virus types and detection methods.

Example 2 Allstate Business Insurance is using an intelligent agent to reduce call center traffic and provide

help to human insurance agents during the rate-quoting process with business customers. In these

cases, rate quotes can be fairly complicated. Using this system, agents can quickly answer

questions posted by corporate customers, even if the agents are not fully familiar with the related

issue.

Intelligent agents are also utilized in e-mail servers, news filtering and distribution, appointment

handling, and automated information gathering.

Machine Learning

At this time, AI systems do not have the same learning capabilities that humans have; rather,

they have simplistic (but improving) machine learning (modeled after human learning methods.

The machine-learning scientists try to teach computers to identify patterns and make connections

by showing the machines a large volume of examples and related data. Machine learning also

allows computer systems to monitor and sense their environmental activities so the machines can

adjust their behavior to deal with changes in the environment. The technology can also be used

to predict performance, to reconfigure programs based on changing conditions, and much more.

Technically speaking, machine learning is a scientific discipline concerned with the design and

development of algorithms that allow computers to learn based on data coming from sensors,

databases, and other sources. This learning is then used for making predictions, recognizing

patterns, and supporting decision makers. For an overview, see Alpaydin (2016) and Theobald

(2017).

Machine-learning algorithms (see Chapter 5 for description and discussion) are used today by

many companies. For an executive guide to machine learning, see Pyle and San Jose (2015).

The process of machine learning involves computer programs that learn as they face new

situations. Such programs collect data and analyze them and then “train” themselves to arrive at

conclusions. For example, by showing examples of situations to a machine-learning program, the

program can find elements not easily visible without it. A well-known example is that of

computers detecting credit card fraud.

Application Case 2.2 illustrates how machine learning can improve companies’ business

processes.

Application Case 2.2 How Machine Learning Is Improving Work in Business

The following examples of using machine learning are provided by Wellers, et al. (2017), who

stated that “today’s leading organizations are using machine learning-based tools to automate

decision processes. . . .”

1. IMPROVING CUSTOMER LOYALTY AND RETENTION. Companies mine customers’ activities, transactions, and social interactions and sentiments to predict

customer loyalty and retention. Companies can use machine learning, for example, to

predict people’s desire to change jobs and then employers can make attractive offers to

keep the existing employees or to lure potential employees who work elsewhere to

move to new employers.

2. HIRING THE RIGHT PEOPLE. Given an average of 250 applicants for a good job in certain companies, an AI-based program can analyze applicants’ resumes and find

qualified candidates who did not apply but placed their resume online.

3. AUTOMATING FINANCE. Incomplete financial transactions that lack some data (e.g., order numbers) require special attention. Machine-learning systems can learn

how to detect and correct such situations, very quickly and at minimal cost. The AI

program can take the necessary corrective action automatically.

4. DETECTING FRAUD. Machine-learning algorithms use pattern recognition to detect fraud in real time. The program is looking for anomalies, and then it makes inferences

regarding the type of detected activities to look for fraud. Financial institutions are the

major users of this program.

5. PROVIDING PREDICTIVE MAINTENANCE. Machine learning can find anomalies in the operation of equipment before it fails. Thus, corrective actions are

done immediately at a fraction of a cost to repair equipment after it fails. In addition,

optimal preventive maintenance can be done (see Opening Vignette Chapter 1).

6. PROVIDING RETAIL SHELF ANALYSIS. Machine learning combined with machine vision can analyze displays in physical stores to find whether items are in

proper locations on the shelves, whether the shelves are properly stocked, and whether

the product labels (including prices) are properly shown.

7. MAKING OTHER PREDICTIONS. Machine learning has been used for making many types of predictions ranging in areas from medicine to investments. An example

is Google Flights, which predicts delays that have not been flagged yet by the airlines.

Source: Compiled from Wellers, et al. (2017) and Theobald (2017).

Questions for Case 2.2

1. Discuss the benefits of combining machine learning with other AI technologies. 2. How can machine learning improve marketing? 3. Discuss the opportunities of improving human resource management. 4. Discuss the benefits for customer service.

According to Taylor (2016), the “increased computing power, coupled with other improvements

including better algorithms and deep neural networks for image processing, and ultra-fast in-

memory databases like SAP HANA, are the reasons why machine learning is one of the hottest

areas of development in enterprise software today.” Machine-learning applications are also

expanding due to the availability of Big Data sources, especially those provided by the IoT

(Chapter 13). Machine learning is basically learning from data.

There are several methods of machine learning. They range from neural networks to case-based

reasoning. The major ones are presented in Chapter 5.

Deep Learning

One subset, or refinement, of machine learning is called deep learning. This technology, which

is discussed in Chapter 6, tries to mimic how the human brain works. Deep learning uses

artificial neural technology and plays a major role in dealing with complex applications that

regular machine learning and other AI technologies cannot handle. Deep learning (DL) delivers

systems that not only think but also keep learning, enabling self-direction based on fresh data

that flow in. DL can tackle previously unsolvable problems using its powerful learning

algorithms.

For example, DL is a key technology in autonomous vehicles by helping to interpret road signs

and road obstacles. DL is also playing critical roles in smartphones, robotics, tablets, smart

homes, and smart cities (Chapter 13). For a discussion of these and other applications, see Mittal

(2017). DL is mostly useful in real-time interactive applications in the areas of machine vision,

scene recognition, robotics, and speech and voice processing. The key is continuous learning. As

long as new data arrive, learning occurs.

Example Cargill Corp. offers conventional analytics, and DL-based analytics help farmers to do more

profitable work. For example, farmers can produce better shrimp at lower cost. DL is used

extensively in stock market analysis and predictions. For details, see Smith

(2017) and Chapter 6.

Machine and Computer Vision

The definitions of machine vision vary because several different computer vision systems

include different hardware and software as well as other components. Generally speaking, the

classical definition is that the term machine vision includes “the technology and methods used to

provide imaging-based automated inspection and analysis for applications such as robot

guidance, process control, autonomous vehicles, and inspection.” Machine vision is an important

tool for the optimization of production and robotic processes. A major part of machine vision is

the industrial camera, which captures, stores, and archives visual information. This information

is then presented to users or computer programs for analysis and eventually for automatic

decision making or for support of human decision making. Machine vision can be confused with

computer vision because sometimes the two are used as synonyms, but some users and

researchers treat them as different entities. Machine vision is treated more as an engineering

subfield, while computer vision belongs to the computer science area.

Computer Vision

Computer vision, according to Wikipedia, “is an interdisciplinary field that deals with how

computers can be made for gaining high-level understanding from digital images or videos. From

the perspective of engineering, it seeks to automate tasks that the human visual system can do.”

Computer vision acquires or processes, analyzes, and interprets digital images and produces

meaningful information for making decisions. Image data can take several formats, such as

photos or videos, and they can come from multidimensional sources (e.g., medical scanners).

Scene and item recognitions are important elements in computer vision. The computer vision

field plays a vital role in the domains of safety, security, health, and entertainment. Computer

vision is considered a technology of AI, which enables robots and autonomous vehicles to see

(refer to the description in Chapter 6). Both computer vision and machine vision automate many

human tasks (e.g., inspection). These tasks can deal with one image or a sequence of images. The

major benefit of both technologies is lowering the costs of performing tasks, especially those that

are repetitive and make the human eyes tired. The two technologies are also combined

with image processing that facilitates complex applications, such as in visual quality control.

Another view shows them as being interrelated based on image processing and sharing a variety

of contributing fields.

An applied area of machine vision is scene recognition, which is performed by computer vision.

Scene recognition enables recognition and interpretation of objects, scenery, and photos.

Example of Application Significant illegal logging exists in many countries. To comply with the laws in the United

States, Europe, and other countries, it is necessary to examine wood in the field. This requires

expertise. According to the U.S. Department of Agriculture, “the urgent need for such field

expertise, training and deploying humans to identify processed wood in the field [i.e., at ports,

border crossings, weigh-stations, airports, and other points of entry for commerce] is

prohibitively expensive and difficult logistically. The machine vision wood identification project

(MV) has developed a prototype machine vision system for wood identification.” Similarly, AI

computer vision combined with deep learning is used to identify illegal poachers of animals

(see USC, 2018).

Another example of this application is facial recognition in several security applications, such as

those used by the Chinese police that employ smart glasses to identify (via facial recognition)

potential suspects. In 2018, the Chinese police identified a suspect who attended a pop concert.

There were 60,000 people in the crowd. The person was recognized at the entrance gate where a

camera took his picture; see the video at youtube.com/watch?v=Fq1SEqNT-7c. In 2018, US

Citizenship and Immigration Services identified people that used false passports in the same

manner.

Video Analytics

Applying computer vision techniques to videos enables the recognition of patterns (e.g., for

detecting fraud) and identifying events. This is a derivative application of computer vision.

Another example is one in which, by letting computers view TV shows, it is possible to train the

computers to make predictions regarding human interactions and the success of advertising.

Robotic Systems

Sensory systems, such as those for scene recognition and signal processing, when combined with

other AI technologies, define a broad category of integrated, possibly complex, systems,

generally called robotics (Chapter 10). There are several definitions of robots, and they are

changing over time. A classical definition is this: “A robot is an electromechanical device that is

guided by a computer program to perform manual and/or mental tasks.” The Robotics Institute of

America formally defines a robot as “a programmable multifunctional manipulator designed to

move materials, parts, tools, or specialized devices through variable programmed motions for the

performance of a variety of tasks.” This definition ignores the many mental tasks done by

today’s robots.

An “intelligent” robot has some kind of sensory apparatus, such as a camera, that collects

information about the robot’s surroundings and its operations. The collected data are interpreted

by the robot’s “brain,” allowing it to respond to the changes in the environment.

Robots can be fully autonomous (programmed to do tasks completely on their own, even repair

themselves), or can be remotely controlled by a human. Some robots known

as androids resemble humans, but most industrial robots are not this type. Autonomous robots

are equipped with AI intelligent agents. The more advanced smart robots are not only

autonomous but also can learn from their environment, building their capabilities. Some robots

today can learn complex tasks by watching what humans do. This leads to better human–robot

collaboration. The Interactive Group at MIT is experimenting with this capability by teaching

robots to make complex decisions. For details, see Shah (2016). For an overview of the robot

revolution, see Waxer (2016).

Example: Walmart Is Using Robots to Properly Stock Shelves The efficiency of Walmart stores depends on appropriately stocking their shelves. Using manual

labor for checking what is going on is expensive and may be inaccurate. As of late 2017, robots

were supporting the company’s stocking decisions.

At Walmart, the 2-foot-tall robots use a camera/sensor to scan the shelves to look for misplaced,

missing, or mispriced items. The collected information and the interpretation of problems are

done by these self-moving robots. The results are transmitted to humans who take corrective

actions. The robots carry out their tasks faster and frequently more accurately than humans. The

company experimented with this in 50 stores in 2018. Preliminary results are significantly

positive and are also expected to increase customer satisfaction. The robots will not cause

employees to lose their jobs.

Robots are used extensively in e-commerce warehouses (e.g., tens of thousands are used

by Amazon.com). They also are used in make-to-order manufacturing as well as in mass

production (e.g., cars), lately of self-driven vehicles. A new generation of robots is designed to

work as advisors, as described in Chapter 12. These robots are already advising on topics such as

investments, travel, healthcare, and legal issues. Robots can serve as front desk receptionists and

even can be used as teachers and trainers.

Robots can help with online shopping by collecting shopping information, matching buyers and

products, and conducting price and capability comparisons. These are known as shopbots (e.g.,

see igi-global.com/dictionary/shopbot/26826). Robots can carry goods for shoppers in open air

markets. Walmart is experimenting now with robotic shopping carts (Knight, 2016). For a video

(4:41 min.), see businessinsider.com/personal-robots-for-shopping-and-e-commerce-2016-

9?IR=T. The Japanese company SoftBank opened a cellphone store in Tokyo entirely staffed by

robots, each named Pepper. Each robot is mobile (on wheels) and can approach customers.

Initially, communication with customers was done by entering information into a tablet attached

to each Pepper. A major issue with robots is their trend to take human jobs. For a discussion of

this topic, see Section 14.6.

Natural Language Processing

Natural language processing (NLP) is a technology that gives users the ability to communicate

with a computer in their native language. The communication can be in written text and/or in

voice (speech). This technology allows for a conversational type of interface in contrast with

using a programming language that consists of computer jargon, syntax, and commands. NLP

includes two subfields:

• NATURAL LANGUAGE UNDERSTANDING that investigates methods of enabling

computers to comprehend instructions or queries provided in ordinary English or other

human languages.

• NATURAL LANGUAGE GENERATION that strives to have computers produce

ordinary spoken language so that people can understand the computers more easily. For

details and the history of NLP,

see en.wikipedia.org/wiki/Natural_language_processing and Chapter 6.

NLP is related to voice-generated data as well as text and other communication forms.

Speech (Voice) Understanding

Speech (voice) understanding is the recognition and understanding of spoken languages by a

computer. Applications of this technology have become more popular. For instance, many

companies have adopted this technology in their automated call centers. For an interesting

application, see cs.cmu.edu/~./listen.

Related to NLP is machine translation of languages, which is done by both written text (e.g.,

Web content) and voice conversation.

Machine Translation of Languages

Machine translation uses computer programs to translate words and sentences from one language

to another. For example, Babel Fish Translation, available at babelfish.com, offers more than 25

different combinations of language translations. Similarly, Google’s Translate

(translate.google.com) can translate dozens of different languages. Finally, users can post their

status on Facebook in several languages.

Example: Sogou’s Travel Translator This Chinese company introduced, in 2018, an AI-powered portable travel device. Chinese

people are now traveling to other countries in increasing numbers (200 million expected in 2020

versus 122 million in 2016). The objective of the device is to enable Chinese tourists to plan trips

(so they can read Web sites like Trip Advisor, available in English). The AI-powered portable

travel device enables tourists to read menus, street signs, and communicate with native speakers.

The device, which is using NLP and image recognition, is connected to Sogou search (a search

engine). In contrast with the regular Chinese-English dictionaries, this device is structured

specifically for travelers and their needs.

Knowledge and Expert Systems and Recommenders

These systems, which are presented in Chapter 12, are computer programs that store knowledge,

which their applications use to generate expert advice and/or perform problem solving.

Knowledge-based expert systems also help people to verify information and make certain types

of automated routine decisions.

Recommendation systems (Chapter 12) are knowledge-based systems that make shopping and

other recommendations to people. Another knowledge system is chatbots (see Chapter 12).

Knowledge Sources and Acquisition for Intelligent Systems

For many intelligent systems to work, it is necessary for them to have knowledge. The process of

acquiring this knowledge is referred to as knowledge acquisition. This activity can be complex

because it is necessary to make sure what knowledge is needed. It must fit the desired system. In

addition, the sources of the knowledge need to be identified to ensure the feasibility of acquiring

the knowledge. The specific methods of acquiring the knowledge need to be identified and if

expert(s) are the source of knowledge, their cooperation must be ensured. In addition, the method

of knowledge representation and reasoning from the collected knowledge must be taken into

account, and knowledge must be validated and be consistent.

Given this information, it is easy to see that the process of knowledge acquisition

(see Figure 2.5) can be very complex. It includes extracting and structuring knowledge. It has

several methods (e.g., observing, interviewing, scenario building, and discussing), so specially

trained knowledge engineers may be needed for knowledge acquisition and system building. In

many cases, teams of experts with different skills are created for knowledge acquisition.

Knowledge can be generated from data, and then experts may be used to verify it. The acquired

knowledge needs to be organized in an activity referred to as knowledge representation.

Figure 2.5 Automated Decision-Making Process

Figure 2.5 Full Alternative Text

Knowledge Representation

Acquired knowledge needs to be organized and stored. There are several methods of doing this,

depending on what the knowledge will be used for, how the reasoning from this knowledge will

be done, how users will interact with the knowledge, and more. A simple way to represent

knowledge is in the form of questions and matching answers (Q&A).

Reasoning from Knowledge

Perhaps the most important component in an intelligent system is its reasoning feature. This

feature processes users’ requests and provides answers (e.g., solutions, recommendations) to the

user. The major difference among the various types of the intelligent technologies is the type of

reasoning they use.

Chatbots

Robots come in several shapes and types. One type that has become popular in recent years is the

chatbot. A chatbot, which will be presented in Chapter 12, is a conversional robot that is used for

chatting with people. (A “bot” is short for “robot.”) Depending on the purpose of the chat, which

can be done in writing or by voice, bots can be in the form of intelligent agents that retrieve

information or personal assistants that provide advice. In either case, chatbots are usually

equipped with NLP that enables conversations in natural human languages rather than in a

programmed computer language. Note that Google has rolled out six different voices to its

Google’s Assistant.

Emerging AI Technologies

Several new AI technologies are emerging. Here are a few examples:

• EFFECTIVE COMPUTING. These technologies detect the emotional conditions of

people and suggest how to deal with discovered problems

• BIOMETRIC ANALYSIS. These technologies can verify an identity based on unique

biological traits that are compared to stored ones (e.g., facial recognition).

Cognitive Computing

Cognitive computing is the application of knowledge derived from cognitive science (the study

of the human brain) and computer science theories in order to simulate the human thought

processes (an AI objective) so that computers can exhibit and/or support decision-making and

problem-solving capabilities (see Chapter 6). To do so, computers must be able to use self-

learning algorithms, pattern recognition, NLP, machine vision, and other AI technologies. IBM

is a major proponent of the concept by developing technologies (e.g., Watson) that support

people in making complex decisions. Cognitive computing systems learn to reason with purpose,

and interact with people naturally. For details, see Chapter 6 and Marr (2016).

Augmented Reality

Augmented reality (AR) refers to the integration of digital information with the user

environment in real time (mostly vision and sound). The technology provides people real-world

interactive experience with the environment. Therefore, information may change the way people

work, learn, play, buy, and connect. Sophisticated AI programs may include machine vision,

scene recognition, and gesture recognition. AR is available on iPhones as ARKit. (Also

see Metz, 2017.)

These AR systems use data captured by sensors (e.g., vision, sound, temperature) to augment and

supplement real-world environments. For example, if you take a photo of a house with your

cellphone, you can immediately get the publicly available information about its configuration,

ownership, and tax liabilities on your cellphone.

Section 2.4 Review Questions

1. Define intelligent agents and list some of their capabilities. 2. Prepare a list of applications of intelligent agents. 3. What is machine learning? How can it be used in business? 4. Define deep learning. 5. Define robotics and explain its importance for manufacturing and transportation. 6. What is NLP? What are its two major formats? 7. Describe machine translation of languages. Why it is important in business? 8. What are knowledge systems? 9. What is cognitive computing? 10. What is augmented reality?

2.5 AI Support for Decision Making

Almost since the inception of AI, researchers have recognized the opportunity of using

it for supporting the decision-making process and for completely automating decision

making. Jeff Bezos, the CEO of Amazon.com, said in May 2017 that AI is in a golden

age, and it is solving problems that were once in the realm of science fiction (Kharpal,

2017). Bezos also said that Amazon.com is using AI in literally hundreds of

applications, and AI is really of amazing assistance. Amazon.com has been using AI, for

example, for product recommendations for over 20 years. The company also uses AI for

product pricing, and as Bezos said, to solve many difficult problems. And indeed, since

its inception, AI has been related to problem solving and decision making. AI

technologies allow people to make better decisions. The fact is that AI can:

• Solve complex problems that people have not been able to solve. (Note that solving

problems frequently involves making decisions.)

• Make much faster decisions. For example, Amazon makes millions of pricing and

recommendation decisions, each in a split second.

• Find relevant information, even in large data sources, very fast.

• Make complex calculations rapidly.

• Conduct complex comparisons and evaluations in real time.

In a nutshell, AI can drive some types of decisions many times faster and more

consistently than humans can. For details, watch the video

at youtube.com/watch?v=Dr9jeRy9whQ/. The nature of decision making, especially

nonroutine ones, as noted in Chapter 1, is complex. We discussed in Chapter 1 the fact

that there are several types of decisions and several managerial levels of making them,

and we looked at the typical process of making decisions. Making decisions, many of

which are used for problem solving, requires intelligence and expertise. AI’s aim is to

provide both. As a result, it is clear that using AI to facilitate decision making involves

many opportunities, benefits, and variations. For example, AI can successfully support

certain types of decision making and fully automate others.

In this section, we discuss some general issues of AI decision support. The section also

distinguishes between support of decision making and fully automating decision making.

Some Issues and Factors in Using AI in Decision Making

Several issues determine the justification of using AI and its chance of success. These include:

• The nature of the decision. For example, routine decisions are more likely to be fully

automated, especially if they are simple.

• The method of support, what technology(ies) is (are) used. Initially, automated decision

supports were rule-based. Practically, expert systems were created to generate solutions to

specific decision situations in well-defined domains. Another popular technology mentioned

earlier was “recommender,” which appeared with e-commerce in the 1990s. Today, there is

an increased use of machine learning and deep learning. A related technology is that of

pattern recognition. Today, attention is also given to biometric types of recognition.

For example, research continues to develop an AI machine that will interview people at airports,

asking one or two questions, and then determining whether they are telling the truth. Similar

algorithms can be used to vet refugees and other types of immigrants.

• COST-BENEFIT AND RISK ANALYSES. These are necessary for making large-scale

decisions, but computing these values may not be simple with AI models due to difficulties

in measuring costs, risks, and benefits. For example, as we cited earlier, researchers used

100 metrics to measure the intelligence level of vacuum cleaners.

• USING BUSINESS RULES. Many AI systems are based on business or other types of

rules. The quality of automated decisions depends on the quality of these rules. Advanced

AI systems can learn and improve business rules.

• AI ALGORITHMS. There is an explosion in the number of AI algorithms that are the

basis for automated decisions and decision support. The quality of the decisions depends on

the input of the algorithms, which may be affected by changes in the business environment.

• SPEED. Decision automation is also dependent on the speed within which decisions need

to be made. Some decisions cannot be automated because it takes too much time to get all

the relevant input data. On the other hand, manual decisions may be too slow for certain

circumstances.

AI Support of the Decision-Making Process

Much AI support can be applied today to the various steps of the decision-making process. Fully

automated decisions are common in routine situations and will be discussed in the next section.

Here we follow the steps in the decision-making process described in Chapter 1.

Problem Identification

AI systems are used extensively in problem identification typically in diagnosing equipment

malfunction and medical problems, finding security breaches, estimating financial health, and so

on. Several technologies are used. For example, sensor-collected data are used by AI algorithms.

Performance levels of machines are compared to standards, and trend analysis can point to

opportunities or troubles.

Generating or Finding Alternative Solutions

Several AI technologies offer alternative solutions by matching problem characteristics with best

practices or proven solutions stored in databases. Both expert systems and chatbots employ this

approach. They can generate recommended solutions or provide several options from which to

choose. AI tools such as case-based reasoning and neural computing are used for this purpose.

Selecting a Solution

AI models are used to evaluate proposed solutions, for example, by predicting their future impact

(predictive analysis), assessing their chance of success, or predicting a company’s reply to action

taken by a competitor.

Implementing the Solutions

AI can be used to support the implementation of complex solutions. For example, it can be used

to demonstrate the superiority of proposals and to assess resistance to changes.

Applying AI to one or more of the decision-making processes and steps enables companies to

solve complex real-world problems, as shown in Application Case 2.3.

Application Case 2.3 How Companies Solve Real-World Problems Using Google’s Machine-

Learning Tools

The following examples were extracted from Forrest (2017):

Google’s Cloud Machine Learning Engine and Tensor Flow allow unique access to machine

learning tools without the need for PhD-educated data scientists.

The following companies use Google’s tools to solve the listed problem.

1. AXA INTERNATIONAL. This global insurance company uses machine learning to predict which drivers would be more likely to cause major accidents. The analysis

provides prediction accuracy of 78 percent. This prediction is used to determine

appropriate insurance premiums.

2. AIRBUS DEFENSE & SPACE. Detecting clouds in satellite imagery was done manually for decades. Using machine learning, the process has been expedited by 40

percent, and the error rate has been reduced from 11 percent to 3 percent.

3. PREVENTING OVERFISHING GLOBALLY. A government agency previously monitored only small sample regions globally to find fishing violators. Now, using

satellite AIS positioning, the agency can watch the entire ocean. Using machine

learning, the agency can track all fishing vessels to find violators.

4. DETECTING CREDIT CARD FRAUD IN JAPAN. SMFG, a Japanese financial services company, uses Google’s machine learning (a deep learning application) to

monitor fraud related to credit card use, with an 80–90 percent accuracy of detection.

The detection generates an alarm for taking actions.

5. KEWPIE FOOD OF JAPAN. This company detected defective potato cubes manually using a slow and expensive process. Using Google AI tools enables it to

automatically monitor video feeds and alert inspectors to remove defective potatoes.

Source: Condensed and compiled from Forrest (2017).

Questions for Case 2.3

1. Why use machine learning for predictions? 2. Why use machine learning for detections? 3. What specific decisions were supported in the five cases?

Automated Decision Making

As the power of AI technologies increases, so does its ability to fully automate more and more

complex decision-making situations.

Intelligent and Automated Decision Support

As early as 1970, there were attempts to automate decision making. These attempts were

typically done with the use of rule-based expert systems that provided recommended solutions to

repetitive managerial problems. Examples of decisions made automatically include the

following:

• Small loan approvals

• Initial screening of job applicants

• Simple restocking

• Prices of products and services (when and how to change them)

• Product recommendation (e.g., at Amazon.com)

The process of automated decision making is illustrated in Figure 2.5. The process starts with

knowledge acquisition and creation of a knowledge repository. Users submit questions to the

system brain, which generates a response and submits it to the users. In addition, the solutions

are evaluated so that the knowledge repository and the reasoning from it can be improved.

Complex situations are forwarded to humans’ attention. This process is especially used in

knowledge-based systems. Note that the process in Figure 2.5 for knowledge acquisition

illustrates automatic decision making as well. Companies use automated decision making for

both their external operations (e.g., sales) and internal operations (e.g., resource allocation,

inventory management). An example follows.

Example: Supporting Nurses’ Diagnosis Decisions

A study conducted in a Taiwanese hospital (Liao, et al., 2015) investigated the use of AI to

generate nursing diagnoses and compared them to diagnoses generated by humans. Diagnoses

required comprehensive knowledge, clinical experience, and instinct. The researchers used

several AI tools, including machine learning, to conduct data mining and analysis to predict the

probable success of automated nursing diagnoses based on patient characteristics. The results

indicated an 87 percent agreement between the AI and human diagnosis decisions.

Such technology can be used in places that have no human nursing staff as well as by nursing

staff who want to verify the accuracy of their own diagnostic predictions. The system can

facilitate the training of nursing staff as well.

Automated decisions can take several forms, as illustrated in Technology Insight 2.2.

Technology Insight 2.2 Schrage’s Models for Using AI to Make

Decisions Schrage (2017) of MIT’s Sloan School has proposed the following four models for AI to make

autonomous business decisions:

1. THE AUTONOMOUS ADVISOR. This is a data-driven management model that uses AI algorithms to generate best strategies and instructions on what to do and makes specific

recommendations. However, only humans can approve the recommendations (e.g., proposed

solutions).

Schrage provided an example in which an American retailing company replaced an entire

merchandising department with an AI machine, ordering employees to obey directives from

it. Obviously, resistance and resentment followed. To ensure compliance, the company had to

install monitoring and auditing software.

2. THE AUTONOMOUS OUTSOURCE. Here, the traditional business process outsourcing model is changed to a business process algorithm. To automate this activity, it is necessary to

create crystal-clear rules and instructions. It is a complex scenario since it involves resource

allocation. Correct predictability and reliability are essential.

3. PEOPLE–MACHINE COLLABORATION. Assuming that algorithms can generate optimal decisions in this model, humans need to collaborate with the brilliant, but

constrained, fully automated machines. To ensure such collaboration, it is necessary to train

people to work with the AI machines (see the discussion in Chapter 14). This model is used

by tech giants such as Netflix, Alibaba, and Google.

4. COMPLETE MACHINE AUTONOMY. In this model, organizations fully automate entire processes. Management needs to completely trust AI models, a process that may take years.

Schrage provides an example of a hedge fund that trades very frequently based on a

machine’s recommendations. The company uses machine learning to train the trading

algorithms.

Implementing these four models requires appropriate management leadership and collaboration

with data scientists. For suggestions of how to do so, consult Schrage (2017), who has written

several related books. Kiron (2017) discusses why managers should consider AI for decision

support.

An interesting note is that some competition among companies will actually occur among data-

driven autonomous algorithms and related business models.

Questions for Discussion 1. Differentiate between the autonomous advisor and the people–machine collaboration

models.

2. In all four models, there are some degrees of people–machine interaction. Discuss. 3. Why it is easier to use model 4 for investment decisions than, for example, marketing

strategies?

4. Why is it important for data scientists to work with top management in autonomous AI machines?

Conclusion

There is little doubt that AI can change the decision-making process for businesses; for an

example, see Sincavage (2017). The nature of the change varies based on the circumstances. But,

in general, we expect AI to have a major impact for making better, faster, and more efficient

decisions. Note that, in some cases, an AI watchdog is needed to regulate the process

(see Sample, 2017, for details).

Section 2.5 Review Questions

1. Distinguish between fully automated and supported decision making. 2. List the benefits of AI for decision support. 3. What factors influence the use of AI for decision support? 4. Relate AI to the steps in the classical decision-making process. 5. What are the necessary conditions for AI to be able to automate decision making? 6. Describe Schrage’s four models.

2.6 AI Applications in Accounting

Throughout this book, we provide many examples of AI applications in business,

services, and government. In the following five sections, we provide additional

applications in the traditional areas of business: accounting; finance; human resource

management; marketing, advertising, and CRM; and production-operation

management.

AI in Accounting: An Overview

The CEO of SlickPie Accounting Software for small businesses, Chandi (2017), noticed trends

among professional accountants: their use of AI, including bots in professional routines,

increased. Chandi observed that the major drivers for this are perceived savings in time and

money and increased accuracy and productivity. The adoption has been rapid and it has been

followed by significant improvements. An example is the execution of compliance procedures,

where, for instance, Ernst & Young (EY) is using machine learning for detecting anomalous data

(e.g., fraudulent invoices).

AI in Big Accounting Companies

Major users of AI are the big tax and accounting companies as illustrated in Application

Case 2.4.

Application Case 2.4 How EY, Deloitte, and PwC Are Using AI

The big accounting companies use AI to replace or support human activities in tasks such as tax

preparation, auditing, strategy consulting, and accountancy services. They mostly use NLP,

robotic process automation, text mining, and machine learning. However, they use different

strategies as described by Zhou (2017):

• EY attempts to show quick, positive return on investment (ROI) on a small scale. The

strategy concentrates on business value. EY uses AI, for example, to review legal

documents related to leasing (e.g., to meet new government regulations).

• PricewaterhouseCoopers (PwC) favors small projects that can be completely functioning in

four weeks. The objective is to demonstrate the value of AI to client companies. Once

demonstrated to clients, the projects are refined. PwC demonstrates 70 to 80 such projects

annually.

• Deloitte Touche Tohmatsu Limited, commonly referred to as Deloitte, builds cases that

guide AI-based projects for both clients and internal use. The objective is to facilitate

innovation. One successful area is the use of NLP for review of large contracts that may

include hundreds of thousands of legal documents. The company reduced such review time

from six months to less than a month, and it reduced the number of employees who had

performed the review by more than 70 percent. Deloitte, like its competitors, is using AI to

evaluate potential procurement synergies for merger and acquisition decisions. Such

evaluation is a time-consuming task since it is necessary to check huge quantities of data

(sometime millions of data lines). As a result, Deloitte can finish such evaluation in a week

compared to the four to five months required earlier. Deloitte said that with AI, it is viewing

data in ways never even contemplated before (Ovaska-Few, 2017).

All big accounting companies use AI to assist in generating reports and to conduct many other

routine, high-volume tasks. AI has produced high-quality work, and its accuracy has become

better and better with time.

Sources: Compiled from Chandi (2017), Zhou (2017), and Ovaska-Few (2017).

Questions for Case 2.4

1. What are the characteristics of the tasks for which AI is used? 2. Why do the big accounting firms use different implementation strategies?

Accounting Applications in Small Firms

Small accounting firms also use AI. For example, Crowe Horwath of Chicago is using AI to

solve complex billing problems in the healthcare industry. This helps its clients to deal with

claims processing and reimbursements. The firm can now solve difficult problems that had

previously resisted solutions. Many other applications are used with the support of AI, ranging

from analyzing real estate contracts to risk analysis. It is only a question of time before even

smaller firms will be able to utilize AI as well.

Comprehensive Study of AI Use in Accounting

The ICAEW information technology (IT) faculty provides a free comprehensive study, “AI and

the Future of Accountancy.” This report (ICAEW, 2017) provides an assessment of AI use in

accounting today and in the future. The report sees the advantage of AI by:

• Providing cheaper and better data to support decision making and solve accounting

problems

• Generating insight from data analysis

• Freeing time of accountants to concentrate on problem solving and decision making

The report points to the use of the following:

• Machine learning for detecting fraud and predicting fraudulent activities

• Machine-learning and knowledge-based systems for verifying of accounting tasks

• Deep learning to analyze unstructured data, such as in contracts and e-mails

Job of Accountants

AI and analytics will automate many routine tasks done today by accountants (see discussion

in Chapter 14), many of whom may lose their jobs. On the other hand, accountants will need to

manage AI-based accounting systems. Finally, accountants need to drive AI innovation in order

to succeed or even survive (see Warawa, 2017).

Section 2.6 Review Questions

1. What are the major reasons for using AI in accounting? 2. List some applications big accounting firms use. 3. Why do big accounting firms lead the use of applied AI? 4. What are some of the advantages of using AI cited by the ICAEW report? 5. How may the job of the accountant be impacted by AI?

2.7 AI Applications in Financial Services

Financial services are much diversified, and so is AI usage in the area. One way to organize the

AI activities is by major segments of services. In this section, we discuss only two segments:

banking and insurance.

AI Activities in Financial Services

Singh (2017) observed the following activities that may be found across various types of

financial services:

• Extreme personalization (e.g., using chatbots, personal assistants, and robo investment

advisors) (Chapter 12)

• Shifting customer behavior both online and in brick-and-mortar branches

• Facilitating trust in digital identity

• Revolutionizing payments

• Sharing economic activities (e.g., person-to-person loans)

• Offering financial services 24/7 and globally (connecting the world)

AI in Banking: An Overview

Consultancy.uk (2017) provides an overview of how AI is transforming the banking industry. It

found AI applications mostly in IT, finance and accounting, marketing and sales, human resource

management (HRM), customer service, and operations. A comprehensive survey on AI in

banking was conducted in 2017, and a report is available for purchase (see Tiwan, 2017).

The key findings of this report are as follows:

• AI technologies in banking include all those listed in Section 2.7 and several other

analytical tools (Chapters 3 to 11 of this book).

• These technologies help banks improve both their front-office and back-office operations.

• Major activities are the use of chatbots to improve customer service and communicating

with customers (see Chapter 12), and robo advising is used by some financial institutions

(see Chapter 12).

• Facial recognition is used for safer online banking.

• Advanced analytics helps customers with investment decisions. For examples of this help,

see Nordrum (2017), E. V. Staff (2017), and Agrawal (2018).

• AI algorithms help banks identify and block fraudulent activities including money

laundering.

• AI algorithms can help in assessing the creditworthiness of loan applicants. (For a case

study of an application of AI in credit screening, see ai-toolkit.blogspot.com/2017/01/case-

study-artificial-intelligence-in.html.)

Illustrative AI Applications in Banking

The following are banking institutions that use AI:

• Banks are using AI machines, such as IBM Watson, to step up employee surveillance. This

is important in preventing illegal activities such as those that occurred at Wells Fargo, the

financial services and banking company. For details, see information-

management.com/articles/banks-using-algorithms-to-step-up-employee-surveillance.

• Banks use applications for tax preparation. H&R Block is using IBM Watson to review tax

returns. The program makes sure that individuals pay only what they owe. Using interactive

conversations, the machine attempts to lower people’s tax bills.

• Answering many queries in real time. For example, Rainbird Co. (rainbird.ai/) is an AI

vendor that trains machines to answer customers’ queries. Millions of customers’ questions

keep bank employees busy. Bots assist staff members to quickly find the appropriate

answers to queries. This is especially important in banks where turnovers of employees are

high. Also, there is knowledge degrading overtime, due to frequent changes in policies and

regulations.

Rainbird is integrated with IBM Watson, which is using AI capabilities and cognitive

reasoning to understand the nature of the queries and provide solutions. The program–

employee conversations are done via chatbots, which are deployed to all U.K. branches of

the banks served by Rainbird.

• At Capital One and several other banks, customers can talk with Amazon’s Alexa to pay

credit card bills and check their accounts.

• TD Bank and others (see Yurcan, 2017) experiment with Alexa, which provides machine

learning and augmented reality capabilities for answering queries.

• Bank Danamon uses machine learning for fraud detection and anti–money-laundering

activities. It also improves the customer experience.

• At HSBC, customers can converse with the virtual banking assistant, Olivia, to find

information about their accounts and even learn about security. Olivia can learn from its

experiences and become more useful.

• Santander Bank employs a virtual assistant (called Nina) that can transfer money, pay bills,

and do more. Nina can also authenticate its customers via an AI-based voice recognition

system. Luvo of RBS is a customer service and customer relationship management (CRM)

bot that answers customers’ queries.

• At Accenture, Collette is a virtual mortgage advisor that provides personalized advice.

• A robot named NaO can analyze facial expression and behavior of customers that enter the

branches of certain banks and determine their nationality. Then the machine selects a

matching language (Japanese, Chinese, or English) to interact with the customer.

IBM Watson can provide banks many other services ranging from crime fighting to regulatory

compliance as illustrated next.

Example: How Watson Helps Banks Manage Compliance and Supports Decision Making Government regulations place a burden on banks and other financial institutions. To comply with

regulations, banks must spend a considerable amount of time examining huge amounts of data

generated daily.

Developed by Promontory Financial Group (an IBM subsidiary), IBM Watson (Chapter 6)

developed a set of tools to deal with the compliance problem. The set of tools was trained by

using the knowledge of former regulators and examining data from over 200 different sources.

All in all, the program is based on over 60,000 regulatory citations. It includes three sets of

cognitive tools that deal with regulatory compliance. One of the tools deals with financial crimes,

flagging potential suspicious transactions and possible fraud. The second tool monitors

compliance, and the third one deals with the large volume of data. Watson is acting as a banking

financial consultant for these and other banking issues.

IBM’s tools are designed to assist financial institutions to justify important decisions. The AI

algorithms examine the data inputs and outputs in managerial decision making. For example,

when the program spots suspicious activity, it will notify the appropriate manager, who then will

take the necessary action. For details, see Clozel (2017).

Application Case 2.5 illustrates US Bank’s use of AI to improve customer service.

Application Case 2.5 US Bank Customer Recognition and Services

As of July 2017, US Bank has been able to automatically identify military service members and

veterans when they call or enter one of its branches This is not a simple task. The service

members are recognized by Einstein, an AI-based CRM service from Salesforce Inc.

(see Section 2.9).

What US Bank is trying to do is to recognize customers and understand their needs. Einstein

helps the bank gain a competitive advantage in doing so. Knowledge provided is important not

only for marketing and providing targeted professional financial services but also for greeting

customers on their birthdays or thanking them for using the bank’s services.

The bank now has considerable information about customers available to human agents in real

time. Such information helps customers when online and when at one of the bank’s actual

locations.

The AI application tells the rep all about the customer so the rep can offer appropriate services.

For example, if the customer needs insurance, the AI will detect this need and the rep will offer a

good alternative. It also offers information to an online customer: “Hello, Mary; I see you are

checking your mortgage payments. I have good news for you. . . .”

Source: Compiled from Crosman (2017) and Carey (2017).

Questions for Case 2.5

1. What are Einstein’s advantages to US Bank? 2. What are its advantages to customers? 3. What are the benefits of voice communication?

Insurance Services

Advancements in AI are improving several areas in the insurance industry, mostly in

issuing policies and handling claims.

According to Hauari (2017), the major objectives of the AI support are to improve

analysis results and enhance customer experience. Incoming claims are analyzed by AI,

and, depending on their nature, are sent to appropriate available adjusters. The

technologies used are NLP and text recognition (Chapters 6 and 7). The AI software can

help in data collection and analysis and in data mining old claims.

Agents previously spent considerable time asking routine questions from people

submitting insurance claims. AI machines, according to Beauchamp (2016), provide

speed, accuracy, and efficiency in performing this process. Then AI can facilitate the

underwriting process.

Similarly, claims processing is streamlined with the help of AI. It reduces processing

time (by up to 90 percent) and improves accuracy. Capabilities of machine-learning and

other AI programs can be shared in seconds in multi-office configurations, including

global settings.

have to go through a transformation and adapt to change. Companies and individual agents can

learn from early adopters. For how this is done at MetLife, see Blog (2017).

Example: Metromile Uses AI in Claim Processing Metromile is an innovator in vehicle insurance, using the pay-per-mile model. It operates in

seven U.S. states. In mid-2017, it started using AI-based programs to automate accident data,

process accident claims, and pay customer claims. The automated platform, according to Santana

(2017), is powered by a smart claim bot called AVA. It processes images forwarded by

customers, extracting the pertinent telematic data. The AI bot simulates the accidents’ major

points and makes a verification based on decision rules; authorization for payments provides for

successful verification. The process takes minutes. Only complex cases are sent to investigation

by human processors. Customers are delighted since they can get fast resolutions. While at the

moment AVA is limited to certain types of claims, its range of suitability is increasing with the

learning capabilities of machine learning and the advances in AI algorithms. NOTE:

A 2015 start-up, Lemonade (lemonade.com) provides an AI-based platform for insurance that

includes bots and machine learning. For details, see Gagliordi (2017).

Section 2.7 Review Questions

1. What are the new ways that banks interact with customers by using AI? 2. It is said that financial services are more personalized with AI support. Explain. 3. What back-office activities in banks are facilitated by AI? 4. How can AI contribute to security and safety? 5. What is the role of chatbots and virtual assistants in financial services? 6. How can IBM Watson help banking services? 7. Relate Salesforce Einstein to CRM in financial services. 8. How can AI help in processing insurance claims?

2.8 AI in Human Resource Management (HRM)

As in other business functional areas, the use of AI technologies is spreading rapidly in

HRM. And as in other areas, the AI services reduce cost and increase productivity,

consistency, and speed of execution.

AI in HRM: An Overview

Savar (2017) points to the following reasons for AI to transform HRM, especially in recruiting:

(1) reducing human bias, (2) increasing efficiency, productivity, and insight in evaluating

candidates, and (3) improving relationships with current employees.

Wislow (2017) sees the use of AI as a continuation of automation that supports HRM and keeps

changing it. Wislow suggests that such automation changes how HRM employees work and are

engaged. This change also strengthens teamwork. Wislow divided the impact of AI into the

following areas:

Recruitment (Talent Acquisition)

One of the cumbersome tasks in HRM, especially in large organizations, is recruiting new

employees. The fact is that many job positions are unfilled due to difficulties in finding the right

employees. At the same time, many qualified people cannot find the right jobs.

AI improves the recruiting process as illustrated in Application Case 2.6.

Application Case 2.6 How Alexander Mann Solutions (AMS) Is Using AI to Support the

Recruiting Process

Alexander Mann is a Chicago-based company that offers AI solutions to support the employee

recruitment process. The major objective is to help companies solve HRM problems and

challenges. The AI is used to:

1. Help companies evaluate applicants and their resumes by using machine learning. The result is the decision regarding which applicants to invite for an interview.

2. Help companies evaluate resumes that are posted on the Web. The AI software can use key words for the search related to the background of employees (e.g., training, years

of experience).

3. Evaluate the resumes of the best employees who currently work in a company and create, accordingly, desired profiles to be used when vacancies occur. These profiles

are then compared to those of applying candidates, and the top ones are ranked by their

fit to each job opening. In addition to the ranking, the AI program shows the fit with

each desired criterion. At this stage, the human recruiter can make a final selection

decision. This way, the selection process is faster and much more accurate.

The accuracy of the process solves the candidate volume problem, ensuring that qualified people

are not missed and poorly fit applicants are not selected.

Alexander Mann is also helping its clients to install chatbots that can provide candidates’

answers to questions related to the jobs and the working conditions at the employing company.

(For the recruiting chatbot, see Dickson, 2017). Sources: Compiled from Huang (2017), Dickson (2017), and alexandermannsolutions.com, accessed June 2018.

Questions for Case 2.6

1. What types of decisions are supported? 2. Comment on the human–machine collaboration. 3. What are the benefits to recruiters? To applicants?

4. Which tasks in the recruiting process are fully automated? 5. What are the benefits of such automation?

The use of chatbots to facilitate recruitment is also described by Meister (2017).

Companies that help recruiters and job seekers, especially LinkedIn, are using AI algorithms to

suggest matches to both recruiters and job seekers. Haines (2017) describes the process, noting

that a key benefit of this process is the removal of unconscious biases and prejudices of humans.

AI Facilities Training

The rapid technological developments make it necessary to train and retrain employees. AI

methods can be used to facilitate learning. For example, chatbots can be used as a source of

knowledge to answer learners’ queries. Online courses are popular with employees. AI can be

used to test progress, for example. In addition, AI can be used to personalize online teaching for

individuals and to design group lectures.

AI Supports Performance Analysis (Evaluation)

AI tools enable HR management to conduct performance analysis by breaking work into many

small components and by measuring the performance of each employee and team on each

component. The performance is compared to objectives, which are provided to employees and

teams. AI also can track changes and progress by combining AI with analytical tools.

AI Use in Retention and Attrition Detection

In order to keep employees from leaving, it is necessary for businesses to analyze and predict

how to make workers happy. Machine learning can be used to detect reasons why employees

leave companies by identifying influencing patterns.

AI in Onboarding

Once new employees are hired, the HR department needs help introducing them to the

organizational culture and operating processes. Some new employees require much attention. AI

helps HRM prepare customized onboarding paths that are best for the newcomers. Results

showed that those employees supported by AI-based plans tend to stay longer in organizations

(Wislow, 2017).

Using Chatbots for Supporting HRM

The use of chatbots in HRM is increasing rapidly. Their ability to provide current information to

employees anytime is a major reason. Dickson (2017) refers to the following chatbots: Mya, a

recruiting assistant, and Job Bot, which supports the recruitment of hourly workers. This bot is

also used as a plug-in to Craigslist. Another chatbot mentioned earlier is Olivia;

see olivia.paradox.ai/.

Introducing AI to HRM Operations

Introducing AI to HRM operations is similar to introducing AI to other functional areas.

Meister (2017) suggests the following activities:

1. Experiment with a variety of chatbots 2. Develop a team approach involving other functional areas 3. Properly plan a technology roadmap for both the short and long term, including shared

vision with other functional areas

4. Identify new job roles and modifications in existing job roles in the transformed environment

5. Train and educate the HRM team to understand AI and gain expertise in it

For additional information and discussion, see Essex (2017).

Section 2.8 Review Questions

1. List the activities in recruiting and explain the support provided by AI to each. 2. What are the benefits rewarded to recruiters by AI? 3. What are the benefits to job seekers? 4. How does AI facilitate training? 5. How is performance evaluation of employees improved by AI? 6. How can companies increase retention and reduce attrition with AI? 7. Describe the role of chatbots in supporting HRM.

2.9 AI in Marketing, Advertising, and CRM

Compared to other business areas, there are probably more applications of AI in marketing and

advertising. For example, AI-based product recommendations have been in use

by Amazon.com and other e-commerce companies for more than 20 years. Due to the large

number of applications, we provide only a few examples here.

Overview of Major Applications

Davis (2016) provides 15 examples of AI in marketing as listed with explanations by the authors

of this book and from Martin (2017). Also see Pennington (2018).

1. PRODUCT AND PERSONAL RECOMMENDATIONS. Starting with Amazon.com’s book recommendations for Netflix’s movies, AI-based

technologies are used extensively for personalized recommendations (e.g., see Martin,

2017).

2. SMART SEARCH ENGINES. Google is using RankBrain’s AI system to interpret users’ queries. Using NLP helps in understanding the products or services for which

online users are searching. This includes the use of voice communication.

3. FRAUD AND DATA BREACHES DETECTION. Application for this has covered credit/debit card use for many years, protecting Visa and other card issuers. Similar

technologies protect retailers (such as Target and Neiman Marcus) from hackers’

attacks.

4. SOCIAL SEMANTICS. Using AI-based technologies, such as sentiment analysis and image and voice recognitions, retailers can learn about customers’ needs and provide

targeted advertisements and product recommendations directly (e.g., via e-mail) and

through social media.

5. WEB SITE DESIGN. Using AI methods, marketers are able to design attractive Web sites.

6. PRODUCER PRICING. AI algorithms help retailers price products and services in a dynamic fashion based on the competition, customers’ requirements, and more. For

example, AI provides predictive analysis to forecast the impact of different price

levels.

7. PREDICTIVE CUSTOMER SERVICE. Similar to predicting the impact of pricing, AI can help in predicting the impact of different customer service options.

8. AD TARGETING. Similar to product recommendations, which are based on user profiles, marketers can tailor ads to individual customers. The AI machines attempt to

match different ads with individuals.

9. SPEECH RECOGNITION. As the trend to use voice in human–machine interaction is increasing, the use of bots by marketers to provide product information and prices

accelerates. Customers prefer to talk to bots rather than to key in dialogue.

10. LANGUAGE TRANSLATION. AI enables conversations between people who speak different languages. Also, customers can buy from Web sites written in languages they

do not speak by using GoogleTranslate, for example.

11. CUSTOMER SEGMENTATION. Marketers are segmenting customers into groups and then tailoring ads to each group. While less effective than targeting individuals,

this is more effective than mass advertising. AI can use data and text mining to help

marketers identify the characteristics of specific segments (e.g., by mining historical

files) as well as help tailor the best ads for each segment.

12. SALES FORECASTING. Marketers’ strategy and planning are based on sales forecasting. Such forecasting may be very difficult for certain products. Uncertainties

may exist in many situations such as in customer need assessment. Predictive analytics

and other AI tools can provide better forecasting than traditional statistical tools.

13. IMAGE RECOGNITION. This can be useful in market research (e.g., for identifying consumer preferences of one company’s products versus those of its competition). It

can also be used for detecting defects in producing and/or packaging products.

14. CONTENT GENERATION. Marketers continuously create ads and product information. AI can expedite this task and make sure that it is consistent and complies

with regulations. Also, AI can help in generating targeted content to both individuals

and segments of consumers.

15. USING BOTS, ASSISTANTS, AND ROBO ADVISORS. In Chapter 12, we describe how bots, personal assistants, and robo advisors help consumers of products

and services. Also, these AI machines excel in facilitating customer experience and

strengthen customer relationship management. Some experts call bots and virtual

personal assistants the “face of marketing.”

Another list is provided at en.wikipedia.org/wiki/Marketing_and_artificial_intelligence.

AI Marketing Assistants in Action

There are many ways that AI can be used in marketing. One way is illustrated in Application

Case 2.7 about Kraft Foods.

Application Case 2.7 Kraft Foods Uses AI for Marketing and CRM

The number of mobile users is growing rapidly as is the number of mobile shoppers. Kraft Foods

took notice of that. The company is adapting its advertising and sales to this trend. Mobile

customers are looking for brands and interacting with Kraft brands. Kraft Foods wanted to make

it easy for customers to interact with the company whenever and wherever they want. To achieve

this interaction goal, Kraft Foods created a “Food Assistant,” also known as Kraft Food

Assistant.

The Kraft Food Assistant

Kraft’s Food Assistant is an app for smartphones that allows customers to access more than 700

recipes. Thus, the consumer can browse easily for ideas. Customers enter a virtual store and open

the “recipe of the day.” The app tells the user all the ingredients needed for that recipe or for any

desired recipe. The Food Assistant also posts all the relevant coupons available for the

ingredients on users’ smartphone. Users need only to take the smartphone to a supermarket, scan

the coupons, and save on the ingredients. The recipe of the day is also demonstrated on a video.

Unique to this app is the inclusion of an AI algorithm that learns from users’ orders and can

infer, for example, the users’ family size. The more the AI learns about users, the more

suggestions it makes. For example, it tells users what to do with their leftover ingredients. In

addition, the more the Food Assistant learns about users, the more useful suggestions for recipes

and cooking it can offer. It is like the Netflix recommender. The more Kraft products that users

buy (the ingredients), the more advice they get. The Food Assistant also directs users to the

nearest store that has the recipes’ ingredients. Users can get assistance on how to prepare food in

20 minutes and on many cooking-related topics.

The AI is tracking consumers’ behavior. Information is stored on each user’s loyalty card. The

system makes inferences about what consumers like and targets related promotions to them. This

process is called behavioral pattern recognition, and is based on AI techniques such as

“collaborative filtering.” (See Chapter 12.)

AI assistants also can tweak messages to users, and they know if users are interested in their

topics. The assistant also knows whether customers are responding positively and whether they

are or are not motivated to try a new product or purchase more of what they previously

purchased. The Kraft AI Food Assistant actually is trying to influence and sometimes

to modify consumer behavior. Like other vendors, Kraft is using the information collected by the

AI assistant to forge and execute mobile and regular commerce strategies.

Using the information collected, Kraft and similar vendors can expand their mobile marketing

programs both online and in physical stores.

NOTE:

Users can interact with the system with voice powered by Nuance Communication. The system

is based on natural language processing.

Sources: Compiled from Celentano (2016), press releases at nuance.com, and kraftrecipes.com/media/iphoneassistant.aspx/, accessed

March 2018.

Questions for Case 2.7

1. Identify all AI technologies used in the Food Assistant. 2. List the benefits to the customers. 3. List the benefits to Kraft Foods. 4. How is advertising done? 5. What role is “behavioral pattern recognition” playing? 6. Compare Kraft’s Food Assistant to Amazon.com and Netflix recommendation

systems.

Customer Experiences and CRM

As described earlier, a major impact of AI technologies is changing customer experiences. A

notable example is the use of conversational bots. Bots (e.g., Alexa) can provide information

about products and companies and can provide advice and guidance (e.g., robo advisors for

investment; see Chapter 12). Gangwani (2016) lists the following ways to improve customers’

experiences:

1. Use NLP for generating user documentation. This capability also improves the customer–machine dialogue.

2. Use visual categorization to organize images (for example, see IBM’s Visual Recognition and Clarifai)

3. Provide personalized and segmented services by analyzing customer data. This includes improving shopping experience and CRM.

A well-known example of AI in CRM is Salesforce’s Einstein.

Example: Salesforce’s AI Einstein

Salesforce Einstein is an AI set of technologies (e.g., Einstein Vision for image

recognition) that is used for enhancing customer interactions and supporting sales. For

example, the system delivers dynamic sales dashboards to sales reps. It also tracks

performance and manages teamwork by using sales analytics. The AI product also can

provide predictions and recommendations. It supports Salesforce Customer Successful

Platform and other Salesforce products.

Einstein’s automatically prioritized sales leads make sales reps more productive when

dealing with sales leads and potential opportunities. The sales reps also get insights

about customers’ sentiments, competitors’ involvement, and other information.

For information and a demo, see salesforce.com/products/einstein/overview/. For features and

description of the product, see zdnet.com/article/salesforces-einstein-ai-platform-what-you-need-

to-know/. For additional features, see salesforce.com/products/einstein/features/.

Other Uses of AI in Marketing

The following show the diversity of AI technologies used in marketing:

• It is used to mimic the expertise of in-store salespeople. In many physical stores, humans

are not readily available to help customers who do not want to wait very long. Thus,

shopping is made easier when bots provide guidance. A Japanese store already provides all

services in a physical store by speaking robots.

• It provides lead generation. As seen in the case of Einstein, AI can help generate sales leads

by analyzing customers’ data. The program can generate predictions. Insights can be

generated by intelligent analytics.

• It can increase customer loyalty using personalization. For example, some AI techniques

can recognize regular customers (e.g., in banks). IBM Watson can learn about people from

their tweets.

• Salesforce.com provides a free e-book, “Everything You Need to Know about AI for CRM”

(salesforce.com/form/pdf/ai-for-crm.jsp).

• It can improve the sales pipeline. Narayan (2018) provides a process of how companies can

use AI and robots to do this. Specifically, robots convert unknown visitors into customers.

Robots use three stages: (1) prepare a list of target customers in the database, (2) send

information, ads, videos, and so on to prospects on the list created earlier, and (3) provide

the company sales department with a list of leads that successfully convert potential

customers to buyers.

Section 2.9 Review Questions

1. List five of the 15 applications of Davis (2016). Comment on each. 2. Which of the 15 applications relate to sales? 3. Which of the 15 applications relate to advertising? 4. Which of the 15 applications relate to customer service and CRM? 5. For what are the prediction capabilities of AI used? 6. What is the Salesforce’s Einstein? 7. How can AI be used to improve CRM?

2.10 AI Applications in Production-Operation Management

(POM)

The field of POM is much diversified, and its use of AI is evident today in many areas.

To describe all of them, we would need more than a whole book. In the remaining

chapters, we provide dozens of examples about AI applications in POM. Here, we

provide only a brief discussion regarding two related application areas: manufacturing

and logistics.

AI in Manufacturing

To handle ever-increasing labor costs, changes in customers’ requirements, increased global

competition, and government regulations (Chapter 1), manufacturing companies are using

elevated levels of automation and digitization. According to Bollard et al. (2017), companies

need to be more agile, and react quicker and more effectively. They also need to be more

efficient and improve customers’ (organizations’ and individuals’) experiences. Companies are

pressured to cut costs and increase quality and transparency. To achieve these goals, they need to

automate processes and make use of AI and other cutting-edge technologies.

Implementation Model

Bollard, et al. (2017) proposed a five-component model for manufacturing companies to use

intelligent technologies. This model includes:

• Streamlining processes, including minimizing waste, redesigning processes, and using

business process management (BPM)

• Outsourcing certain business processes, including going offshore

• Using intelligence in decision making by deploying AI and analytics

• Replacing human tasks with intelligent automation

• Digitizing customers’ experiences

Companies have used this model for a long time. Actually, robotics have been used since around

1960 (e.g., Unimate in General Motors). However, the robots were “dumb,” each usually doing

one simple task. Today, companies use intelligent robots for complex tasks, enabling make-to-

order products and mass customization. In other words, many mental and cognitive tasks are

being automated. These developments, involving AI and sensors, allow supporting or even

automating production decisions in real time.

Example When a sensor detects a defective product or a malfunction, the data are processed by an AI

algorithm. An action then takes place instantly and automatically. For example, a defective item

can be removed or replaced. AI can even make predictions about equipment failures before they

occur (see the opening vignette in Chapter 1). This real-time action saves a huge amount of

money for manufacturers. (This process may involve the IoT; see Chapter 13.)

Intelligent Factories

Ultimately, companies will use smart or intelligent factories (see Chapter 13). These factories

use complex software and sensors. An example of a lead supplier is General Electric, which

provides software such as OEE Performance Analyzer and Production Execution Supervisor.

The software is maintained in the “cloud” and it is provided as a “software-as-a-service.” GE

partners with Cisco and FTC to provide security, connectivity, and special analytics.

In addition to GE, well-known companies such as Siemens and Hitachi provide comprehensive

solutions. For an example, see Hitachi AI Technology’s Report (social-

innovation.hitachi/ph/solutions/ai/pdf/ai_en_170310.pdf).

Many small vendors are specializing in different aspects of AI for manufacturing. For example,

BellHawk Systems Corporation, which provides services to small companies, specializes in real-

time operations tracking (see Green, 2016).

Early successes were recorded by large companies such as Procter & Gamble and Toyota.

However, as time passes, medium-size and small companies can also afford AI services. For

additional information, see bellhawk.com.

Logistics and Transportation

AI and intelligent robots are used extensively in corporate logistics and internal and external

transportation, as well as in supply chain management. For example, Amazon.com is using over

50,000 robots to move items in its distribution centers (other e-commerce companies are doing

the same). Soon, we will see driverless trucks and other autonomous vehicles all over the world

(see Chapter 13).

Example: DHL Supply Chain DHL is a global delivery company (competing with FedEx and UPS). It has a supply chain

division that works with many business partners. AI and IoT are changing the manner by which

the company, its partners, and even its competitors operate. DHL is developing innovative

logistics and transportation business models, mostly with AI, IoT, and machine learning. These

models also help DHL’s customers gain a competitive advantage (and this is why the company

cannot provide details in its reports).

Several of the IoT projects are linked to machine learning, specifically in the areas of sensors,

communication, device management, security, and analysis. Machine learning in such cases

assists in tailoring solutions to specific requirements.

Overall, DHL concentrates on the areas of supply chains (e.g., identifies inventories and controls

them along the supply chain) and warehouse management. Machine learning and other AI

algorithms enable more accurate procurement, production planning, and work coordination.

Tagging and tracking items using Radio Frequency Identification (RFID) and Quick Response

(QR) code allow for item tracking along the supply chain. Finally, AI facilitates predictive

analytics, scheduling, and resource planning. For details, see Coward (2017).

Section 2.10 Review Questions

1. Describe the role of robots in manufacturing. 2. Why use AI in manufacturing? 3. Describe the Bollard et al. implementation model. 4. What is an intelligent factory? 5. How are a company’s internal and external logistics supported by AI technologies?

Chapter Highlights

• The aim of artificial intelligence is to make machines perform tasks intelligently,

possibly like people do.

• A major reason for using AI is to cause work and decision making to be easier to

perform. AI can be more capable (enable new applications and business models),

more intuitive, and less threatening than other decision support applications.

• A major reason to use AI is to reduce cost and/or increase productivity.

• AI systems can work autonomously, saving time and money, and perform work

consistently. They can also work in rural and remote areas where human expertise

is rare or not available.

• AI can be used to improve all decision-making steps.

• Intelligent virtual systems can act as assistants to humans.

• AI systems are computer systems that exhibit low (but increasing) levels of

intelligence.

• AI has several definitions and derivatives, and its importance is growing rapidly.

The U.S. government postulated that AI will be a “critical driver of the U.S.

economy” (Gaudin 2016).

• The major technologies of AI are intelligent agents, machine learning, robotic

systems, NLP and speech recognition, computer vision, and knowledge systems.

• Expert systems, recommendation systems, chatbots, and robo advisors are all based

on knowledge transferred to machines.

• The major limitations of AI are the lack of human touch and feel, the fear that it will

take jobs from people, and the possibility that it could be destructive.

• AI is not a match to humans in many cognitive tasks, but it can perform many

manual tasks quicker and at a lower cost.

• There are several types of intelligence, so it is difficult to measure AI’s capacity.

• In general, human intelligence is superior to that of machines. However, machines

can beat people in complex games.

• Machine learning is currently the most useful AI technology. It attempts to learn

from its experience to improve operations.

• Deep learning enables AI technologies to learn from each other, creating synergy in

learning.

• Intelligent agents excel in performing simple tasks considerably faster and more

consistently than humans (e.g., detecting viruses in computers).

• The major power of machine learning is a result of the machine’s ability to learn

from data and its manipulation.

• Deep learning can solve many difficult problems.

• Computer vision can provide understandings from images, including from videos.

• Robots are electromechanical computerized systems that can perform physical and

mental tasks. When provided with sensory devices, they can become intelligent.

• Computers can understand human languages and can generate text or voice in

human languages.

• Cognitive computing simulates the human thought process for solving problems

and making decisions.

• Computers can be fully automated in simple manual and mental tasks using AI.

• Several types of decision making are fully automated using AI; other types can be

supported.

• AI is used extensively in all functional business departments, reducing cost and

increasing productivity, accuracy, and consistency. There is a tendency to increase

the use of chatbots. They all support decision making well.

• AI is used extensively in accounting, automating simple transactions, helping deal

with Big Data, finding fraudulent transactions, increasing security, and assisting in

auditing and compliance.

• AI is used extensively in financial services to improve customer service, provide

investment advice, increase security, and facilitate payments among other tasks.

Notable applications are in banking and insurance.

• HRM is using AI to facilitate recruitment, enhance training, help onboarding, and

streamline operations.

• There is considerable use of AI in marketing, sales, and advertising. AI is used to

support product recommendation, help in search of products and services,

facilitate Web site design, support pricing decisions, provide language translation

in globe trade, assist in forecasting and predictions, and use chatbots for many

marketing and customer service activities.

• AI has been used in manufacturing for decades. Now it is applied to support

planning, supply chain coordination, logistics and transportation, and operation of

intelligent factories.

Key Terms

• artificial brain • artificial intelligence (AI) • augmented intelligence • chatbots • computer vision • deep learning • intelligent agent • machine learning • machine vision • natural language processing (NLP) • robot • scene recognition • shopbot • speech (voice) understanding

Questions for Discussion

1. Discuss the difficulties in measuring the intelligence of machines.

2. Discuss the process that generates the power of AI. 3. Discuss the differences between machine learning and deep learning. 4. Describe the difference between machine vision and computer vision. 5. How can a vacuum cleaner be as intelligent as a six-year-old child? 6. Why are NLP and machine vision so prevalent in industry? 7. Why are chatbots becoming very popular? 8. Discuss the advantages and disadvantages of the Turing Test. 9. Why is augmented reality related to AI? 10. Discuss the support that AI can provide to decision makers. 11. Discuss the benefits of automatic and autonomous decision making. 12. Why is general (strong) AI considered to be “the most significant technology ever

created by humans”?

13. Why is the cost of labor increasing, whereas the cost of AI is declining? 14. If an artificial brain someday contains as many neurons as the human brain, will it be

as smart as a human brain? (Students need to do extra research.)

15. Distinguish between dumb robots and intelligent ones. 16. Discuss why applications of natural language processing and computer vision are

popular and have many uses.

Exercises

1. Go to itunes.apple.com/us/app/public-transit-app-moovit/id498477945?mt=8. Compare Moovit operations to the operation of INRIX.

2. Go to sitezeus.com and view the 2:07 min. video. Explain how the technology works as a decision helper.

3. Go to Investopedia and learn about investors’ tolerance. Then find out how AI can be used to contain this risk, and write a report.

4. In 2017, McKinsey & Company created a five-part video titled “Ask the AI Experts: What Advice Would You Give to Executives About AI?” View the video and

summarize the advice given to the major issues discussed. (Note: This is a class

project.)

5. Watch the McKinsey & Company video (3:06 min.) on today’s drivers of AI at youtube.com/watch?v=yv0IG1D-OdU and identify the major AI drivers. Write a

report.

6. Go to the Web site of the Association for the Advancement of Artificial Intelligence aaai.org/home.html and describe its content. Compare it to that

of ai.sri.com and csail.mit.edu/.

7. Go to crosschx.com and find information about Olive. Explain how it works, what its limitations and advantages are, and which types of decisions it automates and which it

only supports.

8. Go to waze.com and moovitapp.com and find their capabilities. Summarize the help they can provide users.

9. Go to sentient.ai. Find its products that facilitate e-commerce. Write a report. 10. Go to artificialbrain.org and report the latest progress there. 11. Find recent information on research that is aimed to measure artificial intelligence.

Write a report.

12. Go to salesforce.com and find recent developments on AI Einstein. Why it is so popular?

13. Find the latest information on IBM Watson’s advising activities. Write a report. 14. Find information on the use of AI in iPhones. Explore the role of Edge AI. Write a

report.

15. Explore the AI-related products and services of Nuance Inc. (nuance.com). Explore the Dragon voice recognition product.

16. Go to the Netradyne report at cs_netradyne.com/ and read about the use of its product for road safety. Write a report.

17. Go to salesforce.com and investigate the capabilities of Gecko HRM. Relate it to Salesforce Einstein. Provide examples of two applications.

18. Enter McKinsey & Company and find in its Fifty Five “The Value AI Can Bring to Your Business” (mckinsey.com/featured-insights/artificial-intelligence/five-fifty-real-

world-ai). Then look for “Real-World AI.” Find the banking section and dive more

deeply into its content.

19. Find material on the impact of AI on advertising. Write a report. 20. Go to strategicsourceror.com/2018/03/giant-scale-supply-chains-can-make.html.

Summarize the use of AI.

References

• Agrawal, V. “How Successful Investors Are Using AI to Stay Ahead of the

Competition.” ValueWalk, January 28, 2018.

• Alpaydin, E. Machine Learning: The New AI (The MIT Press Essential Knowledge Series).

Boston, MA: MIT Press, 2016.

• Beauchamp, P. “Artificial Intelligence and the Insurance Industry: What You Need to

Know.” The Huffington Post, October 27, 2016.

• Blog. “Welcome to the Future: How AI Is Transforming Insurance.” Blog.metlife.com, October

1, 2017.

• Bollard, A., et al. “The next-generation operating model for the digital world.” McKinsey &

Company, March 2017.

• BrandStudio. “Future-Proof: How Today’s Artificial Intelligence Solutions Are Taking

Government Services to the Next Frontier.” Washington Post, August 22, 2017.

• Carey, S. “US Bank Doubles Its Conversion Rate for Wealth Customers Using Salesforce

Einstein.” Computerworld UK, November 10, 2017.

• Carney, P. “Pat Carney: Artificial Intelligence versus Human Intelligence.” Vancouver Sun,

April 7, 2018.

• Celentano, D. “Kraft Foods iPhone Assistant Appeals to Time Starved Consumers.” The

Balance, September 18, 2016.

• Chandi, N. “How AI is Reshaping the Accounting Industry.” Forbes.com, July 20, 2017.

• Clozel, L. “IBM Unveils New Watson tools to Help Banks Manage Compliance,

AML.” American Banker, June 14, 2017.

• Consultancy.uk. “How Artificial Intelligence Is Transforming the Banking Industry.” September

28, 2017. consultancy.uk/news/14017/how-artificial-intelligence-is-transforming-the-banking-

industry/ (accessed June 2018).

• Coward, J. “Artificial Intelligence Is Unshackling DHL’s Supply Chain Potential.” IoT Institute,

April 18, 2017. ioti.com/industrial-iot/artificial-intelligence-unshackling-dhls-supply-chain-

potential (accessed June 2018).

• Crosman, P. “U.S. Bank Bets AI Can Finally Deliver 360-Degree View.” American Banker, July

20, 2017.

• Davis, B. “15 Examples of Artificial Intelligence in Marketing.” Econsultancy, April 19, 2016.

• Dickson, B. “How Artificial Intelligence Optimizes Recruitment.” The Next Web, June 3, 2017.

• Dodge, J. “Artificial Intelligence in the Enterprise: It’s On.” Computerworld, February 10, 2016.

• Dormehl, L. Thinking Machines: The Quest for Artificial Intelligence—and Where It’s Taking Us

Next. New York, NY: Tarcher-Perigee, 2017.

• Essex, D. “AI in HR: Artificial Intelligence to Bring Out the Best in

People.” TechTargetEssential Guide, April 2017.

• E. V. Staff. “Artificial Intelligence Used to Predict Short-Term Share Price Movements.” The

Economic Voice, June 22, 2017.

• Finlay, S. Artificial Intelligence and Machine Learning for Business: A No-Nonsense Guide to

Data Driven Technologies. 2nd ed. Seattle, WA: Relativistic, 2017.

• Forrest, C. “7 Companies That Used Machine Learning to Solve Real Business Problems.” Tech

Republic, March 8, 2017.

• Fuller, D. “LG Claims Its Roboking Vacuum Is As Smart As a Child.” Androidheadlines.com,

July 18, 2017.

• Gagliordi, N. “Softbank Leads $120M Investment in AI-Based Insurance Startup

Lemonade.” ZDNET, December 19, 2017.

• Gangwani, T. “3 Ways to Improve Customer Experience Using A.I.” CIO Contributor Network,

October 12, 2016.

• Gaudin, S. “White House: A.I. Will Be Critical Driver of U.S. Economy.” Computerworld,

October 12, 2016.

• Gitlin, J. M. “Watch Out, Waze: INRIX’s New Traffic App Is Coming for You.” Ars Technica,

March 30, 2016. arstechnica.com/cars/2016/watch-out-waze-inrixs-new-traffic-app-is-coming-

for-you/ (accessed June 2018).

• Greengard, S. “Delving into Gartner’s 2016 Hype Cycle.” Baseline, September 7, 2016.

• Greig, J. “Gartner: AI Business Value Up 70% in 2018, and These Industries Will Benefit the

Most.” Tech Republic, April 25, 2018.

• Haines, D. “Is Artificial Intelligence Making It Easier and Quicker to Get a New

Job?” Huffington Post UK, December 4, 2017.

• Hauari, G. “InsurersLeverage AI to Unlock Legacy Claims Data.” Information Management,

July 3, 2017.

• Huang, G. “Why AI Doesn’t Mean Taking the ‘Human’ Out of Human Resources.” Forbes.com,

September 27, 2017.

• Hughes, T. “Google DeepMind’s Program Beat Human at Go.” USA Today, January 27, 2016.

• ICAEW. “Artificial Intelligence and the Future of Accountancy.” artificial-intelligence-

report.ashx/, 2017.

• Kaplan, J. Artificial Intelligence: What Everyone Needs to Know. London, UK: Oxford

University Press, 2016.

• Kharpal, A. “A.I. Is in a ‘Golden Age’ and Solving Problems That Were Once in the Realm of

Sci-Fi, Jeff Bezos Says.” CNBC News, May 8, 2017.

• Kiron, D. “What Managers Need to Know About Artificial Intelligence?” MITSloan

Management Review, January 25, 2017.

• Knight, W. “Walmart’s Robotic Shopping Carts Are the Latest Sign That Automation Is Eating

Commerce.” Technology Review, June 15, 2016.

• Kolbjørnsrud, V., R. Amico, and R. J. Thomas. “How Artificial Intelligence Will Redefine

Management.” Harvard Business Review, November 2, 2016.

• Korosec, K. “Inrix Updates Traffic App to Learn Your Daily Habits.” Fortune Tech, March 30,

2016.

• Liao, P-H., et al. “Applying Artificial Intelligence Technology to Support Decision-Making in

Nursing: A Case Study in Taiwan.” Health Informatics Journal, June 2015.

• Marr, B., “The Key Definitions of Artificial Intelligence That Explain Its Importance.” Forbes,

February 14, 2018.

• Marr, B. “What Everyone Should Know About Cognitive Computing.” Forbes.com, March 23,

2016.

• Martin, J. “10 Things Marketers Need to Know about AI.” CIO.com, February 13, 2017.

• McPherson, S.S. Artificial Intelligence: Building Smarter Machines. Breckenridge, CO: Twenty-

First Century Books, 2017.

• Meister, J. “The Future of Work: How Artificial Intelligence Will Transform the Employee

Experience.” Forbes.com, November 9, 2017.

• Metz, C. “Facebook’s Augmented Reality Engine Brings AI Right to Your Phone.” Wired, April

19, 2017.

• Mittal, V. “Top 15 Deep Learning Applications That Will Rule the World in 2018 and

Beyond.” Medium.com, October 3, 2017.

• Narayan, K. “Leverage Artificial Intelligence to Build your Sales Pipeline.” LinkedIn, February

14, 2018.

• Ng, A. “What Artificial Intelligence Can and Can’t Do Right Now.” Harvard Business Review,

November 9, 2016.

• Nordrum, A. “Hedge Funds Look to Machine Learning, Crowdsourcing for Competitive

Advantage.” IEEE Spectrum, June 28, 2017.

• Ovaska-Few, S. “How Artificial Intelligence Is Changing Accounting.” Journal of Accountancy,

October 9, 2017.

• Padmanabhan, G. “Industry-Specific Augmented Intelligence: A Catalysts for AI in the

Enterprise.” Forbes, January 4, 2018.

• Pennington, R. “Artificial Intelligence: The New Tool for Accomplishing an Old Goal in

Marketing.” Huffington Post, January 16, 2018.

• Press, G. “Top 10 Hot Artificial Intelligence (AI) Technologies.” Forbes, January 23, 2017.

• Pyle, D., and C. San José. “An Executive’s Guide to Machine Learning.” McKinsey &

Company, June 2015.

• Reinharz, S. An Introduction to Artificial Intelligence: Professional Edition: An Introductory

Guide to the Evolution of Artificial Intelligence. Kindle Edition. Seattle, WA: Simultaneous

Device Usage (Amazon Digital Service), 2017.

• Sample, I. “AI Watchdog Needed to Regulate Automated Decision-Making, Say Experts.” The

Guardian, January 27, 2017.

• Santana, D. “Metromile Launches AI Claims Platform.” Digital Insurance, July 25, 2017.

• Savar, A. “3 Ways That A.I. Is Transforming HR and Recruiting.” INC.com, June 26, 2017.

• Schrage, M. “4 Models for Using AI to Make Decisions.” Harvard Business Review, January 27,

2017.

• Shah, J. “Robots Are Learning Complex Tasks Just by Watching Humans Do Them.” Harvard

Business Review, June 21, 2016.

• Sharma, G. “China Unveils Multi-Billion Dollar Artificial Intelligence Plan.” International

Business Times, July 20, 2017. ibtimes.co.uk/china-unveils-multi-billion-dollar-artificial-

intelligence-plan-1631171/ (accessed January 2018).

• Sincavage, D. “How Artificial Intelligence Will Change Decision-Making for

Businesses.” Business 2 Community, August 24, 2017.

• Singh, H. “How Artificial Intelligence Will Transform Financial Services.” Information

Management, June 6, 2017.

• SMBWorld Asia Editors. “Hays: Artificial Intelligence Set to Revolutionize

Recruitment.” Enterprise Innovation, August 30, 2017.

• Smith, J. Machine Learning: Machine Learning for Beginners. Can Machines Really Learn Like

Humans? All About Artificial Intelligence (AI), Deep Learning and Digital Neural Networks.

Kindle Edition. Seattle, WA: Amazon Digital Service, 2017.

• Staff. “Assisted, Augmented and Autonomous: The 3 Flavours of AI Decisions.” Software and

Technology, June 28, 2017. tgdaily.com/technology/assisted-augmented-and-autonomous-the-3-

flavours-of-ai-decisions

• Steffi, S. “List of 50 Unique AI Technologies.” Hacker Noon.com, October 18, 2017.

• Taylor, P. “Welcome to the Machine – Learning.” Forbes BrandVoice, June 3,

2016. forbes.com/sites/sap/2016/06/03/welcome-to-the-machine-

learning/#3175d50940fe (accessed June 2017).

• Theobald, O. Machine Learning for Absolute Beginners: A Plain English Introduction. Kindle

Edition. Seattle, WA, 2017.

• Tiwan, R. “Artificial Intelligence (AI) in Banking Case Study Report 2017.” iCrowd Newswire,

July 7, 2017.

• USC. “AI Computer Vision Breakthrough IDs Poachers in Less Than Half a Second.” Press

Release, February 8, 2018.

• Violino, B. “Most Firms Expect Rapid Returns on Artificial Intelligence

Investments.” Information Management, November 1, 2017.

• Warawa, J. “Here’s Why Accountants (Yes, YOU!) Should Be Driving AI Innovation.” CPA

Practice Advisor, November 1, 2017.

• Waxer, C. “Get Ready for the BOT Revolution.” Computerworld, October 17, 2016.

• Wellers, D., et al. “8 Ways Machine Learning Is Improving Companies’ Work

Processes.” Harvard Business Review, May 31, 2017.

• Wislow, E. “5 Ways to Use Artificial Intelligence (AI) in Human Resources.” Big Data Made

Simple, October 24, 2017. bigdata-madesimple.com/5-ways-to-use-artificial-intelligence-ai-in-

human-resources/.

• Yurcan, B. “TD’s Innovation Agenda: Experiments with Alexa, AI and Augmented

Reality.” Information Management, December 27, 2017.

• Zarkadakis, G. In Our Own Image: Savior or Destroyer? The History and Future of Artificial

Intelligence. New York, NY: Pegasus Books, 2016.

• Zhou, A. “EY, Deloitte and PwC Embrace Artificial Intelligence for Tax and

Accounting.” Forbes.com, November 14, 2017.