Journal of Data, Information and Management (2020) 2:67–74

https://doi.org/10.1007/s42488-020-00023-1

ORIGINAL ARTICLE

AI in operations management: applications, challenges

and opportunities
Ali K. Dogru1 · Burcu B. Keskin2

Received: 25 September 2019 / Accepted: 29 January 2020 / Published online: 21 February 2020
© Springer Nature Switzerland AG 2020

Abstract
We have witnessed unparalleled progress in artificial intelligence (AI) and machine learning (ML) applications in the
last two decades. The AI technologies have accelerated advancements in robotics and automation, which have significant
implications on almost every aspect of businesses, and especially supply chain operations. Supply chains have widely
adopted smart technologies that enable real-time automated data collection, analysis, and prediction. In this study, we review
recent applications of AI in operations management (OM) and supply chain management (SCM). Specifically, we consider
the innovations in healthcare, manufacturing, and retail operations, since collectively, these three areas represent a majority
of the AI innovations in business as well as growing problem areas. We discuss primary challenges and opportunities for
utilizing AI in those industries. We also discuss trending research topics with significant value potential in these areas.

Keywords Artificial Intelligence (AI) · Machine Learning (ML) · Robotics · Automation · Operations Management
(OM) · Supply Chain Management (SCM)

1 Introduction Turing (1950), is considered to be the first philosophical

severe argument to judge AI. According to this test, if
Artificial intelligence (AI) is a broad scientific discipline at a human observer cannot distinguish a conversation with
the intersection of computer sciences, statistics, operations a machine from a conversation with a human, then the
management, mathematics, humanities, social sciences, and machine is assumed to be intelligent. The first applications
philosophy. AI aims to develop nonbiological systems, have focused on developing computer programs that
such as machines and computers, to perform tasks that can compete with human players at structural games,
typically require human intelligence. Machine learning such as checkers, chess, and Go. For instance, Arthur
(ML), on the other hand, is a sub-discipline of AI that Samuel’s checker’s program in 1959 was intelligent enough
focuses on applications of statistical learning. ML aims to challenge an amateur human gamer (Samuel 1959).
to develop algorithms that are capable of automatically However, the term AI is first coined by John McCarthy,
learning associations in large amounts of data in real-time during the Dartmouth Conference of 1956, which marks the
and improving with experience without being explicitly birth of AI as a scientific discipline (McCarthy et al. 2006).
programmed. The AI research dates back to the 1940s when Since then, AI research has attracted significant attention,
scientists started discussing the possibility of creating a particularly after three milestones. The first was Deep Blue
thinking machine. The Turing test, which was proposed by by IBM, the first chess engine that could beat a human world
chess champion, Garry Kasparov, in 1997 (Campbell et al.
2002). The second milestone was AlphaGo by Deepmind,
 Ali K. Dogru
[email protected] an AI Go engine that beaten a professional human Go
player in 1915 (Silver et al. 2018). The third milestone
1
was AlphaZero by Deepmind; an AI chess engine that
College of Business and Economic Development,
School of Management, University of Southern
achieved a superhuman level of play after playing against
Mississippi, Hattiesburg, MS 39406, USA itself for only 24 hours and then defeating the best brute
2 Culverhouse College of Business Administration,
force dynamic programming chess engines in 2017 (Silver
ISM Department, The University of Alabama, Tuscaloosa, et al. 2018). These turning points have served as catalysts to
AL 35487, USA unprecedented advancements in artificial intelligence (AI),
68 J. of Data, Inf. and Manag. (2020) 2:67–74

and its machine learning (ML) applications in the last two 2 AI in healthcare operations
decades, attracting considerable investment for R&D of AI
technologies by high-tech companies. For example, in 2016, Healthcare is one of the areas that AI has a high value poten-
companies invested $26 to $39 billion in AI, which is more tial. Various applications have been developed for online
than three times as much as in 2013 (Bughin et al. 2017). platforms to increase the efficiency of clinical operations,
Supply chains have largely benefited from these growing such as appointment/surgery scheduling, diagnostic image
interest and investments in AI technologies. The advanced analysis, and prognosis and prediction of specific ailments
information systems, such as cloud storage, internet of (Panch et al. 2018). For instance, IBM AI uses mammogra-
things, wireless technologies, and affordable sensors, acted phy images and medical records, including iron deficiency,
as enablers of the evolution of business operations. Now, it thyroid function, biopsies, lab tests and codes from other
is possible that business processes and supply chains, even diagnoses, to predict breast cancer a year in advance with
across multiple firms and different geographies, are seam- 87% accuracy that is comparable to a human radiologist
lessly connected. Smart devices, point of sale data, mobile (Chorev 2019). Sugar.IQ, a diabetes management app that is
applications (apps), and sensors have enabled businesses powered by ML algorithms, analyzes blood glucose levels,
to gather demographic, geographic, and behavioral con- insulin data, hypoglycemia episodes, and food logs to give
sumer data in real-time, which is critical for product/service personalized insights and predictions on blood sugar events
development and capacity planning. AI applications have of users (Latts 2019).
improved the productivity of operations via robotics and One promising AI application is Migraine.AI, a person-
automation and have allowed marketing to better understand alized migraine management application that predicts what
and predict customer demand. Responsiveness has gained causes a user’s migraine and optimizes their medication
importance in the supply chain world as customers have pro- window (MigraineAI 2019). The application provides these
gressively demanded customized products and personalized customized predictions by integrating medical and genetic
services at short notice and at reduced lead times. profiles of patients, environmental triggers, and various
In this study, we review recent applications of AI in patient behaviors, including medication usage and sleep pat-
operations management (OM) / supply chain management terns. It only uses smartphone sensors to passively monitor
(SCM). Specifically, we consider the innovations in and process real-time data, which consists of 400 distinct
healthcare, manufacturing, and retail operations, since data points. Currently, Migraine.AI research team is run-
collectively, these three areas represent a majority of the ning clinical trials partnering with prestigious institutions
business innovations as well as problem areas. Analyzing such as Oxford University, Garmin, and Georgetown Uni-
more than 400 use cases across 19 industries, a recent versity. In March 2019, Berkeley HAAS Business School
McKinsey report estimates the potential value of AI as $3.5 chose Migraine.AI as the most impactful healthcare start-up
to $5.8 trillion, of which $0.2-$0.3 billion in healthcare of 2019.
services, $0.7 to $1.0 billion in manufacturing (automotive Our interview with Ozcan Cikmaz, the CEO of the start-
assembly, advanced electronics & semiconductors, and high up, points to four major challenges in using AI in healthcare.
technology combined), and $0.4-$0.8 billion in retail (Chui First, ML algorithms require an extensive amount of
et al. 2018). It is estimated that ML applications may personal data for performance and customization of health
significantly reduce healthcare expenditures (up to $300 services. However, collecting, storing, and protecting vast
billion per year), increase online sales (by up to % 30), amounts of identifiable personal data poses a serious
reduce inventory (by up to %50), decrease delivery times challenge. Cikmaz notes that besides acquiring data from
(by up to %30), and improve efficiency of production human subjects, technology companies have to invest a
(by up to %10) (Bughin et al. 2017). Nevertheless, the considerable amount of capital to ensure proper storage
current adoption of AI technologies varies by industry and security of personal data. The second difficulty is
remarkably. While the automotive industry and high-tech related to creating an efficient user interface. Practical
companies take the lead in AI adoption, adoption rates applications need to minimize the data collection time by
by retail and healthcare sectors, where AI may provide asking the smallest number of questions to capture as
the most significant value, are relatively moderate and much useful information as possible. Furthermore, a user
low, respectively (Bughin et al. 2017). Each sector has should be able to access any point in the application via
its dynamics and concerns regarding the adoption of AI. the smallest number of clicks. Migraine.AI, for instance,
In this study, we discuss sector-specific challenges and asks a patient only one-minute of their daily time to collect
opportunities for adopting AI-enabled technologies. We also information and does not even require patients to type
discuss recent research trends with high value potentials in their answers. Cikmaz highlights that there are currently
these areas. mobile applications in the market that require as much
J. of Data, Inf. and Manag. (2020) 2:67–74 69

as 30 minutes to retrieve user data. Third, he emphasizes sensor-enabled physical assets and designed to monitor
that even though AI is quite technical and originate from performance and help reduce costly unplanned equipment
computer sciences and statistics, implementation of AI outages. The convergence of advanced technologies such
for practical purposes require a broader collaborative and as sensors, cloud services, big data, and machine learning
interdisciplinary teamwork. He believes that OM/OR plays has brought this idea to fruition. For example, Chevron
a crucial role in this regard. The fourth challenge is Corp. and Unilever PLC both use digital twin technology
related to customization and potential bias. It is essential for predictive maintenance (Castellanos 2019; Smith 2019).
to represent diversity well in the data sets on which the In Unilever’s plants, the devices send real-time information
ML algorithms are trained because ML algorithms are only on temperature, motor speed, and other production variables
as good as the data set they are trained on Panch et al. into the cloud. Algorithms take in the data and use advanced
(2018). Any bias against an underrepresented group in a analytics to map out the best operational conditions.
data set will result in a biased automated decision. For Workers on-site track product quality with handheld
example, an appointment scheduling software can make devices, modeling solutions to problems, and sharing data
racially-biased scheduling decisions, if objective function with colleagues in other locations. On the other hand, the
of the algorithm is not tailored to reduce disparities among blockchain can be used across the whole manufacturing
different racial groups (Travers 2019). Preventing those industry to address all different types of projects and
biases is particularly important for AI applications in stakeholders. For instance, a particular use of blockchain
healthcare. Cikmaz states that most migraine management is highlighted by Olsen and Tomlin (2019). In a possible
applications fail simply because they do not factor in the future scenario, each kiwifruit is automatically tagged with
unique characteristics of women. He notes that applications a micro-RFID by the producer in New Zealand during the
need to be tailored according to the nuances of female picking process, which then helps the retailer in China
physiology to improve personal recommendations and to track the full history of kiwis’ journey within the
prevent against algorithmic biases. supply chain including the picking time, loading, travel and
unloading times, temperature changes the kiwis exposed
to during transportation, with the blockchain technology.
3 AI in manufacturing operations Olsen and Tomlin (2019) also note that this blockchain
technology allows for smart contracts, which dynamically
The digitization of manufacturing initiated a significant adapts to the changes in external events (i.e., transporter’s
transformation in manufacturing operations. This transition payments depends on temperature extremes that kiwi fruits
is so compelling that it is being called Industry 4.0. The are exposed to during the trip).
fourth industrial revolution starts with the adoption of Schmitz et al. (2019) point out five major roadblocks that
computers and automation in the third industrial revolution the manufacturing sector needs to overcome to benefit from
and enhances it with smart and autonomous systems fueled Industry 4.0 technologies fully. First, the leading companies
by data and ML. In a sense, Industry 4.0 optimizes the even in the frontier of Industry 4.0 are grappling with how
use of data, computers, algorithms, and automation. For to upskill their current workforce to take on new work
example, Big River Steel LLC in Arkansas is using AI that responsibilities made possible by Internet 4.0 and to recruit
is powered by deep learning and neural networks to create new employees with the right skills. Adopting technical
a “self-learning mill” to spot problems during production, expertise takes time and money. Second, companies are
ensure quality and conserve energy (Murawski 2019). The unwilling to make a vast amount of financial investments
data to train the algorithms are captured by thousands for Industry 4.0 technologies due to perceived high cost
of sensors. The AI system collects data on equipment and risk of scaling. Third, the return of investments in
conditions such as pressure and temperature. It also tracks Industry 4.0 technologies is relatively long. The report
scrap-metal conditions, customer delivery schedules, and suggests taking the minimum viable product approach by
electricity costs. The technology helps the mill produce implementing small to minimize the upfront cost of scaling.
different grades of steel with lower operating costs, a Fourth, small scale pilot implementations of concept
competitive boost. Industry 4.0 technologies usually manifest unclear business
Some of the technologies that make Industry 4.0 a value since managers fail to estimate the potential impact
reality are autonomous robots, industrial Internet of Things, of the pilot on the manufacturing processes and define
additive manufacturing (3D-printing), cloud computing, success criteria accurately. Finally, due to a lack of focus,
augmented reality, digital twins, and blockchain. We expand companies often struggle to determine which case of a pilot
on two of these technologies with examples: digital twins technology to prioritize. The report recommends value-
and blockchain. Digital twins are software models of backed prioritization to keep the pilot programs focused.
70 J. of Data, Inf. and Manag. (2020) 2:67–74

4 AI in retail operations an already complicated stochastic appointment scheduling

problem by increasing uncertainty. Some studies aim to
AI is also widely used in retail operations. Web-based mitigate inherent high uncertainty in customer preferences
shopping enables online retailers to collect vast amounts by developing data-driven models and methodologies.
of data regarding shopping habits, basket composition, and For instance, Pan et al. (2017) use power consumption
even browsing patterns. These data provide information behavior of customers to predict customer attendance and
regarding future offerings, promotions, and help retailers Klein et al. (2017) proposes a differentiated time slot
manage their supply chain operations, inventory fulfillment, pricing scheme to strategically direct customer choices to
and distribution systems. Walmart, for instance, uses an help find more efficient distribution plans. Second, the
ML algorithm with image recognition capabilities called perishability of grocery items adds to the complexity of
“Eden” to classify produces according to their freshness the problem, imposing further restrictions on inventory
and to predict when they will go bad (Weber and Schütte availability, transportation, and appointment scheduling
2019). Home Depot uses deep learning to predict stock-out problems. Serving overtime, for instance, is not possible
probabilities of inventory items and to automate inventory (i.e., a vehicle cannot deliver after a cutoff time). Moreover,
replenishment decisions, whereas the Kroger Company uses non-attendance of the customer at the time of the delivery
deep learning to find the shortest path to navigate in results in the broken cold chain for certain groceries.
a warehouse to complete a customer order (Weber and Third, last-mile delivery is an urban phenomenon; therefore,
Schütte 2019). various complicating factors need to be considered while
Some traditional retailers are offering home grocery solving the CVRPTW, including rush hours and traffic
delivery to compete with online retailers. E-groceries, such intensity in certain districts. ML algorithms may play a
as AmazonFresh, Instacart, Peapod, and Tesco, deliver key role in overcoming these difficulties. For example, AI
grocery items at the door of their online customers. Creating can be used to better estimate time and location-dependent
an efficient e-grocery distribution plan requires multiple travel time in urban areas and ML algorithms can be used
hard optimization problems to be solved jointly, including a to cluster customers based on their geographic location,
stochastic appointment scheduling problem with customer shopping habits (i.e., frequency and quantity), day and
preferences, a capacitated vehicle routing problem with time preferences regarding the delivery, which may help
time windows (CVRPTW) and an associated inventory facilitate finding solutions for hard appointment scheduling
management problem. This problem structure is not only and vehicle routing problems.
limited to e-groceries but also applies to other last-mile The e-grocery distribution problem becomes more inter-
services as well, such as home health care delivery, esting when the delivery is implemented by an autonomous
appliance delivery, and installation services, and other vehicle. For example, George Mason University uses
attended home services including gas, electricity, and water autonomous robots on wheels to deliver food to students on-
(Bruck et al. 2018). Though one of the most expensive, demand (Holley 2019). We already know that Google and
polluting, and least efficient supply chain operations, Tesla have been running tests for their self-driving cars for
attended home services provide numerous advantages to a while. Recently, UPS also joined the wagon by invest-
consumers such as a wide selection of products, price ing on TuSimple, a self-driving trucking start-up, to run
comparison among e-groceries, and easy access to goods tests on long-haul trucks that operate autonomously with
which are not produced locally (Ehmke and Mattfeld 2012). a human operator on board to monitor and intervene if
Customers also save time and fuel by not driving to the retail necessary (Vartabedian 2019). Similarly, FedEx currently
store. tests a platooning technology, an assistant self-driving tech-
The last-mile delivery of grocery items is particularly nology that enables trucks to follow each other closely to
challenging because of three main reasons. First, home save fuel (Vartabedian 2019). In this context, autonomous
delivery service requires the attendance of the customer vehicles are likely to change a typical CVRPTW problem
at the time of delivery. If a customer is not present, a formulation. Constraints that impose restrictions on driver’s
new delivery needs to be scheduled, which may potentially working hours and breaks, for instance, need to be adapted
disrupt a previously scheduled delivery. Appointment or completely removed. Potentially, new routes need to be
scheduling literature shows that customers are more likely defined for secure autonomous vehicle delivery. Even more
to attend when their preferences are considered (Schectman interestingly, if drones are to be used for delivery, routes
et al. 2008; Gupta and Wang 2008; Bowser et al. need to be redefined in a three-dimensional space, consid-
2010; Feldman et al. 2014; Dogru and Melouk 2019). ering heights of buildings, power lines, and numerous air
However, considering customer choices further complicates traffic restrictions.
J. of Data, Inf. and Manag. (2020) 2:67–74 71

5 Challenges and opportunities critical areas where human judgment is essential. For exam-
ple, it is well-documented that there is a physician shortage
5.1 Displacement effect vs. reinstatement effect of primary care services, particularly in rural areas. It is
also believed that the demand for services such as caring for
AI, automation, and robotics have significant impacts on the elderly is likely to increase, especially in countries with
human resources in supply chains. Acemoglu and Restrepo aging populations where life expectancy increases, whereas
(2018) state that these new technologies increasingly fertility rates decrease (Panch et al. 2018). These empirical
replace labor, creating a strong “displacement effect” in the findings may help health institutions restructure HR plan-
economy. They argue that if the displacement effect is not ning, education, and employment to mitigate the negative
balanced with a more powerful countervailing force called impacts of displacement effect.
“reinstatement effect” (creation of new tasks for labor),
productivity gains in the economy will not be sustainable 5.2 Accountability and trust
and balanced economic growth will not be achieved.
Sustainable economic growth requires a prosperous middle Despite their immense benefits, AI applications currently
class with relatively high and steady incomes. High operate in the grey legal area that has not been regulated
unemployment leads to a decrease in consumption, which well. Inadequate regulation creates an asymmetric environ-
in turn hits supply chains that produce goods and provide ment, in which profit-oriented business organizations are
services for the middle class. Then, what new tasks can incentivized to accelerate deployment of AI technologies
supply chains create to maintain labor? while spending minimal time for tests that are vital to pro-
Even though AI technologies are likely to go on to tect consumers from numerous harms and biases (Whittaker
replace some tasks that have been previously implemented et al. 2018). Moreover, inadequate regulation makes it dif-
purely by human or by human-computer collaboration, such ficult to find a legal entity on the provider end to keep
as pattern recognition, forecasting, logistics and decision responsible, if a consumer is harmed by an automated tech-
making (Nilsson 2009), (Acemoglu and Restrepo 2018) nology. Whom should we keep accountable, for instance,
emphasize that there are many inherently human tasks that if an autonomous vehicle is involved in a car accident,
are difficult to automate, including “complex reasoning, or a patient is misdiagnosed by AI receiving unnecessary
judgment, analogy-based learning, abstract problem solv- chemotherapy treatment? In this context, AI Now Institute
ing, and a mixture of physical activity, empathy, and com- has released three reports since 2016, compiling a long
munication.” For example, Daugherty and Wilson (2017) list of negative impacts of AI on consumers. These reports
identify some new tasks for AI companies. These new job make many ethical and practical recommendations to reg-
titles include ulate technology companies and other entities that produce
and use AI technologies better. The reports emphasize the
• “trainers” who train the AI algorithms for better
importance of sector-specific regulation and transparency
performance,
and accountability of the full stack supply chain of AI
• “explainers” who explains the AI decision to non-
technologies (Whittaker et al. 2018).
technical professionals, and
Most ML algorithms, mainly based on neural networks,
• “sustainers” who monitor performance and make AI
operate as black boxes, which make millions of iterative
sustainable.
calculations on big data sets to optimize model parameters
Similarly, recent advancements in AI in diagnostic radiol- that eventually classify or predict based on a defined
ogy image analysis likely have a displacement effect on objective function (Daugherty and Wilson 2017; Panch et al.
radiologists. Panch et al. (2018) argue that health systems 2018). Even though the algorithm performance can be tested
need to revise the distribution of radiology teams in light of and validated, data scientists often cannot provide detailed
these AI progresses in the discipline. Jha and Topol (2016) explanations of the inner workings of machine learning
suggest that radiology and pathology should be consoli- (Daugherty and Wilson 2017). To address this issue, the
dated as one specialty named as “information specialist,” EU Parliament adopted General Data Protection Regulation
whose primary responsibility will be to oversee and ana- (GDPR) in 2016, granting the consumers the right to
lyze the outputs of AI technologies. Furthermore, Rajpurkar request an explanation on a decision made by an automated
et al. (2017) state that AI technologies may help human system (EUGDPR 2016; Panch et al. 2018). This step is
radiologists to focus their limited attention to higher value- particularly essential to regulate technology companies that
added areas by reading the image and highlighting essential develop applications directly linked to the well-being of
areas to the radiologist. To fully benefit from the advantages their consumers. Its extraterritorial applicability aims to
of using AI, healthcare institutions must determine those regulate companies that use EU citizen’s data regardless of
72 J. of Data, Inf. and Manag. (2020) 2:67–74

where the company is located, fining the breaching entities processor of data, leaving consumers in a vulnerable posi-
up to 4% of annual global turnover or 20 Million Euros tion. Subcontracting also increases the likelihood of data
(whichever is greater) (EUGDPR 2016). breaches and enables illegal use of personal data for other
There also exist concerns against utilizing AI, particu- unrelated business purposes.
larly in the healthcare sector. Patients fear that AI technolo- One way to increase control over data is to create inter-
gies violate their privacy, and practitioners are suspicious operable information technologies. Firms and institutions
about the accuracy of automated decisions. Major mistrust usually have inharmonious data technologies and proce-
regarding AI is partly due to the asymmetric environment dures. For instance, hospitals and clinics typically have
that victimizes the patients via data breaches and incor- patient portals. These patient portals usually do not speak
rect medical practices but fails to penalize entities that to each other. Therefore, when patients move to new loca-
create or use them to offer healthcare services. To ensure tions, they usually give their written consent and ask clinics
trust, developers and researchers need to convince clinical to share patient information via mail or email. During this
decision-makers that proposed ML algorithms are capable exchange, patient information may be lost or stolen by third
of delivering targeted health outcomes by running real- parties. However, as Panch et al. (2018) points out, a recent
world clinical trials (Panch et al. 2018). Patients also need study by Google reveals that deep learning can be applied
to be convinced that in exchange for sharing personal data, to combine patient data collected and stored in different for-
they will be receiving improved health services. Govern- mats by three hospitals, without transforming them into a
mental regulations like GDRP and interdisciplinary studies common format first. There is a need for more of these types
and reports by research institutes like AI Now Institute of AI applications to ensure interoperability of database
would help better protect consumers from potential hazards management systems and information security.
and ensure that entities that develop and use AI technolo-
gies kept accountable. Ensuring accountability would also
build public trust regarding AI technologies, which in turn 6 Overall future outlook
would benefit supply chains that rely on AI technologies to
increase the efficiency of their operations. AI, robotics, and automation appear to dominate industries
including machinery, automotive and consumer electronics,
5.3 Customization and information security as well as service sector including healthcare, retail,
education, criminal justice, housing, and banking in the
Increasing demand for customized products and personal- near future. These technologies will go on to provide
ized services require supply chains to collect, store, and businesses and consumers with numerous advantageous
process vast amounts of identifiable personal data, both and opportunities such as improved efficiency, increased
structured (name, geographic location data, and unique customization, better quality, and shorter lead times.
identification number, etc.), and unstructured (posts, tweets, However, there are also challenges to overcome. These
pictures, and videos, etc.). Supply chains need to invest a challenges include but not limited to re-calibration of
considerable amount of capital on information technolo- division of labor by creating new tasks for human,
gies to ensure proper data storage and security. Due to developing secure hybrid systems that human and robots
transaction cost economics, supply chains usually outsource can work simultaneously in tandem, improving regulation
those capabilities by contracting out with technology firms, to ensure accountability of automated systems and to build
rather than purchasing. Unlike big companies, small tech- trust regarding AI technologies, and developing strategies
nology firms usually lack expensive data infrastructures. and technologies to protect personal, institutional and
Since adopting large data warehouses and computational corporate data better. The unprecedented advancements in
power is expensive, these small technology firms usually AI that already surprise us today are likely to accelerate
pay to big companies to obtain specific IT capabilities. shortly, as companies increasingly invest in research and
Additionally, big data requires a different approach to pro- development of AI pilot programs.
cess data. Since data is big, it is more efficient to send Colleges and research institutions play a crucial role
computation on data, instead of sending data on compu- in adapting to the structural changes imposed by AI and
tation. In other words, data collected by small technology robotics technologies. Historically, our education systems
firms, which are outsourced by supply chains, are usually are designed to foster area specialization. Not surprisingly,
stored and processed in data warehouses owned by big tech- AI has long been considered in the computer programming
nology companies. Multiple layers of subcontracting create and statistical learning realm. Despite its technicality, AI has
blurred lines between the owner, collector, controller, and gained a broader meaning in the last two decades. Whittaker
J. of Data, Inf. and Manag. (2020) 2:67–74 73

et al. (2018) recommends that college AI programs must Feldman J, Liu N, Topaloglu H, Ziya S (2014) Appointment
also expand beyond science and engineering disciplines scheduling under patient preference and no-show behavior. Oper
Res 62(4):794–811
in a way that humanities and social sciences are also
Gupta D, Wang L (2008) Revenue management for a primary-care
included. In this regard, colleges must take the lead in clinic in the presence of patient choice. Oper Res 56(3):576–592
bringing this interdisciplinary perspective to the field, Holley P (2019) George mason students have a new dining option:
since the industry urgently needs social and technical Food delivery by robots. https://www.washingtonpost.com/
technology/2019/01/22/george-mason-students-have-new-dining-
skills that can work in analytical teams composed of
option-food-delivered-by-robots/. Accessed 17 September 2019
experts from distinct disciplines harmoniously. Workplaces Jha S, Topol EJ (2016) Adapting to artificial intelligence: radiologists
have already changed: Cubicles have given way to and pathologists as information specialists. Jama 316(22):2353–
open spaces and relaxed social study environments, in 2354
Klein R, Neugebauer M, Ratkovitch D, Steinhardt C (2017)
which workers with different expertise can work together,
Differentiated time slot pricing under routing considerations in
exchange information, and produce creative ideas. Now, it attended home delivery. Transp Sci 53(1):236–255
is time to change our mindsets and perspectives regarding Latts L (2019) Alleviating the burden of diabetes with ai.
AI, automation, and robotics, starting with academic https://www.ibm.com/blogs/think/2019/01/alleviating-the-burden-
of-diabetes-with-ai/. Accessed 17 September 2019
curriculum, education, and research.
McCarthy J, Minsky ML, Rochester N, Shannon CE (2006) A
proposal for the dartmouth summer research project on artificial
intelligence, august 31, 1955. AI Mag 27(4):12–12
References MigraineAI (2019) Migraine ai official website. http://migraine.ai.
Accessed 12 September 2019
Murawski J (2019) Ai runs smart steel plant. https://www.wsj.com/
Acemoglu D, Restrepo P (2018) Artificial intelligence automation and articles/ai-runs-smart-steel-plant-11563183000. Accessed 17
work. Tech. rep., National Bureau of Economic Research September 2019
Bowser DM, Utz S, Glick D, Harmon R (2010) A systematic Nilsson NJ (2009) The quest for artificial intelligence. Cambridge
review of the relationship of diabetes mellitus, depression, and University Press
missed appointments in a low-income uninsured population. Arch Olsen TL, Tomlin B (2019) Industry 4.0: opportunities and challenges
Psychiatr Nurs 24(5):317–329 for operations management. Manuf Serv Oper Manag
Bruck BP, Cordeau JF, Iori M (2018) A practical time slot Pan S, Giannikas V, Han Y, Grover-Silva E, Qiao B (2017) Using
management and routing problem for attended home services. customer-related data to enhance e-grocery home delivery. Ind
Omega 81:208–219 Manag Data Syst 117(9):1917–1933
Bughin J, Hazan E, Ramaswamy S, Chui M, Allas T, Dahlstrom P, Panch T, Szolovits P, Atun R (2018) Artificial intelligence, machine
Henke N, Trench M (2017) Artificial intelligence: the next digital learning and health systems. J Glob Health 8:2
frontier? https://www.mckinsey.com/mgi. Accessed 5 December Rajpurkar P, Irvin J, Zhu K, Yang B, Mehta H, Duan T, Ding D, Bagul
2019 A, Langlotz C, Shpanskaya K et al (2017) Chexnet: radiologist-
Campbell M, Hoane AJ Jr, Hsu Fh (2002) Deep blue. Artif Intell level pneumonia detection on chest x-rays with deep learning.
134(1–2):57–83 arXiv:171105225
Castellanos S (2019) Digital twins concept gains traction among Samuel AL (1959) Some studies in machine learning using the game
enterprises. https://blogs.wsj.com/cio/2018/09/12/digital-twins- of checkers. IBM J Res Dev 3:210–229
concept-gains-traction-among-enterprises/. Accessed 17 September Schectman JM, Schorling JB, Voss JD (2008) Appointment adherence
2019 and disparities in outcomes among patients with diabetes. J Gen
Chorev M (2019) Ai models predict breast cancer with radiologist- Intern Med 23(10):1685
level accuracy. https://www.ibm.com/blogs/research/2019/06/ Schmitz C, Tschiesner A, Janse C, Hallerstede S, Garms F (2019)
ai-models-radiologist-level-accuracy/. Accessed 18 September Indusry 4.0: capturing value at scale in discrete manufacturing.
2019 https://www.mckinsey.com/industries/advanced-electronics/our-
Chui M, Manyika J, Miremadi M, Henke N, Chung R, Nel P, Malhotra insights/capturing-value-at-scale-in-discrete-manufacturing-with-
S (2018) Notes from the ai frontier: insights from hundreds of industry-4-0. Accessed 10 December 2019
use cases. https://www.mckinsey.com/mgi. Accessed 5 December Silver D, Hubert T, Schrittwieser J, Antonoglou I, Lai M, Guez A,
2019 Lanctot M, Sifre L, Kumaran D, Graepel T et al (2018) A general
Daugherty P, Wilson HJ (2017) Process reimagined: together, people reinforcement learning algorithm that masters chess, shogi, and go
and ai are reinventing business processes from the ground up. through self-play. Science 362(6419):1140–1144
https://www.accenture.com/ acnmedia/pdf-76/accenture-process- Smith J (2019) Unilever uses virtual factories to tune up its
reimagined.pdf#zoom=50. Accessed 18 September 2019 supply chain. https://www.wsj.com/articles/unilever-uses-virtual-
Dogru AK, Melouk SH (2019) Adaptive appointment scheduling for factories-to-tune-up-its-supply-chain-11563206402. Accessed 17
patient-centered medical homes. Omega 85:166–181 September 2019
Ehmke JF, Mattfeld DC (2012) Vehicle routing for attended home Travers M (2019) Medical scheduling software makes black patients
delivery in city logistics. Procedia-Soc Behav Sci 39:622– wait longer in waiting rooms than white patients. https://www.
632 forbes.com/sites/traversmark/2019/12/03/medical-scheduling-
EUGDPR (2016) European union general data protection regulation. software-makes-black-patients-wait-longer-in-waiting-rooms-than-
https://eugdpr.org/. Accessed 18 September 2019 white-patients/. Accessed 9 December 2019
74 J. of Data, Inf. and Manag. (2020) 2:67–74

Turing A (1950) Computing machinery and intelligence. Mind Whittaker M, Crawford K, Dobbe R, Fried G, Kaziunas E, Mathur V,
59(236):433 West SM, Richardson R, Schultz J, Schwartz O (2018) Ai now
Vartabedian M (2019) Ups ventures invest in self-driving trucking report 2018. https://ainowinstitute.org/AI Now 2018 Report.pdf.
startup. https://www.wsj.com/articles/ups-ventures-invests-in-self- Accessed 10 December 2019
driving-trucking-startup-11565874002. Accessed 17 September
2019
Weber F, Schütte R (2019) A domain-oriented analysis of the impact Publisher’s note Springer Nature remains neutral with regard to
of machine learning—the case of retailing. Big Data Cogn Comput jurisdictional claims in published maps and institutional affiliations.
3(1):11