Expert Systems With Applications: Youssra Riahi, Tarik Saikouk, Angappa Gunasekaran, Ismail Badraoui

Expert Systems With Applications 173 (2021) 114702
Contents lists available at ScienceDirect
Expert Systems With Applications

journal homepage: www.elsevier.com/locate/eswa
Review
Artificial intelligence applications in supply chain: A descriptive

bibliometric analysis and future research directions
Youssra Riahi a, Tarik Saikouk b, *, 1, Angappa Gunasekaran c, Ismail Badraoui d
a
Department of Supply Chain Management, Rabat Business School, International University of Rabat, Rabat, Morocco
b
Excelia Business School, CERIIM, La Rochelle, France
c
School of Business and Public Administration, California State University, Bakersfield, USA
d
Department of Supply Chain Management, Rabat Business School, International University of Rabat, Rabat, Morocco
A R T I C L E I N F O A B S T R A C T
Keywords: Today’s supply chains are very different from those of just a few years ago, and they continue to evolve within an
Artificial intelligence extremely competitive economy. Dynamic supply chain processes require a technology that can cope with their
Supply chain increasing complexity. In recent years, several functional supply chain applications based on artificial intelli
Bibliometric analysis
gence (AI) have emerged, yet very few studies have addressed the applications of AI in supply chain processes.
Systematic literature review
Classification
Machine learning, natural language processing, and robotics are all potential enablers of supply chain trans
formation. Aware of the potential advantages of AI implementation in supply chains and of the paucity of work
done regarding it, we explore what researchers have done so far with respect to AI and what needs further
exploration. We reviewed 136 research papers published between 1996 and 2020 from the Scopus database and
provided a classification of the research material according to four critical structural dimensions (level of ana
lytics, AI algorithms or techniques, sector or industry of application, and supply chain processes). This study is
the first attempt to study the AI applications in SC from a process perspective and provides a decisional
framework for adequate use of AI techniques in the different SC processes.
1. Introduction operate—and to create more agile and resilient supply chains—these

linkages, processes, and activities require monitoring, forecasting, pre
Supply chains play a crucial role in quality customer experiences, diction, and optimization. In recent years, applications based on artifi
cost control, and a company’s agility in the face of market opportunities cial intelligence (AI) have emerged in several different fields including
and uncertainties. Companies seek speed, reliability, and traceability supply chains (Borges et al., 2020).
whilst considering their cost imperatives, deadlines, and inventory AI enables systems to make resourceful decisions and execute tasks
optimization (Collin et al., 2009; Zhang et al., 2016). Supply chain automatically without human intervention.
managers should monitor and avoid incidents and factors that disrupt Companies exploit AI and machine learning to gain insights into
the supply process—from the most common incidents (delivery delays, various areas, including warehousing, logistics, and supply chain ma
quality defects, etc.) to major events (political unrest, natural disasters, nagement. AI definition tend to differ based on the perspective from
suppliers’ financial instability, etc.). These factors could cause supply which we wish to define it. A restrictive definition of AI can encompass
chain complexity in already uncertain environments. every machine or equipment that uses computational abilities to mimic
According to Mentzer et al. (2001), “a supply chain is the network of human intelligence. There are many definitions of AI depending on what
organizations that are involved, through upstream and downstream AI achieves, and are generally classified into four categories on the basis
linkages, in the different processes and activities that produce value in of the factor of a thinking human being and in terms of rational behavior
the form of products and services delivered to the ultimate consumer.” (i) systems that think and act like a human being
To work well in the complex environments in which supply chains and (ii) systems that think and act rationally.From a general
* Corresponding author.
E-mail addresses: youssra.riahi@uir.ac.ma (Y. Riahi), saikoukt@excelia-group.com (T. Saikouk), agunasekaran@csub.edu (A. Gunasekaran), Ismail.badraoui@uir.
ac.ma (I. Badraoui).
1
ORCID:0000-0001-9674-4722.
https://doi.org/10.1016/j.eswa.2021.114702
Received 19 November 2020; Received in revised form 17 January 2021; Accepted 6 February 2021
Available online 12 February 2021
0957-4174/© 2021 Elsevier Ltd. All rights reserved.
Y. Riahi et al. Expert Systems With Applications 173 (2021) 114702
perspective, AI can be defined as the ability of a system to reproduce a technology that has been applied for many years but is still far from
human intelligence, with the ideal characteristic to rationalize and take being fully utilized in SC. Yet, only recently have technological de
actions that have the highest tendency of achieving a specific goal velopments shown that AI has many applications (Min, 2010).
(Čerka et al., 2015). In addition to its increasing adoption in industry, AI shows a growing
AI makes it possible to implement predictive approaches that allow and extensive presence in the scholarly discourse, and this presence can
rapid assessment and more effective minimization of the risks or be noticed in many fields. Although research regarding AI is being
disruptive events that could occur throughout the supply chain conducted, there is a need to explore the contribution of AI to the field of
it also enables users to distinguish patterns in the supply chain. Using SC.
algorithms, AI can clearly and quickly identify relevant supply chain There is no systematic reviews entirely devoted to the AI applications
data to develop models that enable managers to better understand how in SC. Usually, researchers focuses on AI for supply chain management
each process works and to discover areas for improvement (Ni et al., (SCM), thus restricting their research to the managerial aspect and
2020). In this new approach of using AI to improve the supply chain and excluding the process perspective in which a researcher can consider the
search for optimization, AI enables companies to constantly learn about applications of AI in the different processes of the SC. It is also important
areas that require improvement, to identify factors that affect perfor to note that even with the limited focus of researcher, only few studies
mance, and to predict performance. considered AI applications in SCM (Min, 2010; Ngai et al., 2014). The
AI in the supply chain context remains an innovation with a full work of Min (Min, 2010) was the first work in which the author con
potential which we have yet to understand. The extant literature is ducted a review on AI applications in SCM. The author selected 28 ar
limited in terms of AI applications in the supply chain field. With this ticles related to AI tools and SCM activities, and extracted seven AI tools
study, we aim to provide a comprehensive view of AI applications in and reviewed their applications in eight SCM activities. Ngai et al.
supply chains, which will serve as a reference for future researchers and (2014); reviewed 77 articles from 1994 and attempted to explore. the
practitioners. We address the knowledge gap by answering the following applications of seven AI techniques in the SCM of textile and apparel
research questions: industry. The results of this review claimed that the gaps between AI
RQ1: What are the general research trends of AI applications in techniques and SCM were caused by the unbalanced applications of
supply chains? certain ML algorithms like NNs (Ni et al., 2020). The articles identified
RQ2: What are the supply chain outcomes that AI achieves? above did presented some insights of the AI applications in SCM, but
RQ3: What level of AI drives such outcomes? they all failed to produce a systematic review. This remark is formulated
Using these research questions as a starting point, we conduct a with consideration that these research works are specifically devoted to
systematic review combined with a bibliometric analysis of the extant SCM. To conclude, it is important to mention that even from a mana
literature to identify prior research exploring AI in the supply chain gerial perspective in which research studies have been conducted, the
context. We perform a descriptive analysis to answer the research research works still have gaps and failed to conduct a systematic liter
questions, the results of which academics and practitioners can use as a ature review. The applications of AI in SC from a process perspective is
reference to track the evolution of research and to envision the potential still not covered in the existing literature and this work is the first
future trends of AI applications in the supply chain domain. Since in the attempt.
literature, there was no systematic examination on the AI application, in
particular in SC from a process perspective, the exploration of the 3. Methods
existing research works was deemed necessary to contribute to the body
of knowledge and to address the existing gaps that should be addressed Based on the study objectives, we followed a specific methodology
in future research. for material collection, tool choice for the analysis, and answering the
The findings of this research intend to constitute a starting point for research questions. For this purpose, we performed a systematic litera
further investigations regarding the application of AI in the different ture review (SLR) combined with a bibliometric analysis (BA). Unsys
supply chain processes and areas, and provide a framework for practi tematic narrative literature reviews tend to be biased, and often lack
tioners in the field for a successful implementation of AI techniques. The rigor, while a systematic review, on the other hand, uses a systematic
analysis allowed to identify the different AI techniques usage patterns method to summarize evidence on research questions with a detailed
according to SC processes, which in turn allowed to develop a decisional and comprehensive plan of study (Tawfik et al., 2019; Phulwani et al.,
framework for AI techniques choice according to the different SC pro 2020). Combining a systematic literature review with a bibliometric
cesses. The rest of the paper is structured as follows analysis enhances the contribution of the systematic literature review by
Section 2 presents a review of the literature. Section 3 outlines the an objective assessment of scientific literature, by increasing the rigor,
adopted methodology. Section 4 presents the findings and the analysis, and by alleviating researcher bias (Phulwani et al., 2020).
followed by the classification results in Section 5. Section 6 discusses the
findings, and Section 7 presents the developed framework. Section 8 3.1. Material collection
explains the implications of the study, followed by Section 9 in which
the limitations are listed, and finally Section 10 as a conclusion. We collected the data inputs of this research from the Scopus data
base, whose coverage is almost 60% greater than that of Web of Science
2. Review of literature (WoS) (Zhao & Strotmann, 2015). Several previous studies have
compared different literature databases for literature reviews and bib
As globalization continues, businesses are looking ahead to optimize liometric analysis and showed the Scopus database covers different
their supply chain and develop their capabilities, which brings many fields of science, technology, social sciences and more (Fahimnia et al.,
changes to the market, demand, data availability, management and 2015). The Scopus database is also more comprehensive compared to
more. These changes require the concerned business organizations to Web-of-Science database which would include only ISI indexed journals,
follow up in order to maintain their competitiveness by avoiding de which in turns limits the number of articles (Yong-Hak, 2013). Finally,
mand uncertainty, disruptions, and financial risk. Thus, many capabil Scopus provides comprehensive author and institution profiles, in
ities should be developed within SCs to ensure the highest rankings in addition to exhaustive data related to each document, ensuring high
the global competition (Giannakis & Papadopoulos, 2016). Since the accuracy and completeness. All the articles and journals included in the
world has been moving towards a digital future over the years, one of the study were written in English. Since systematic literature reviews rely
most outstanding technologies that could assist in the enhancement of heavily on the judicious choice of keyword, we decided that the
SCs capabilities, thus of its transformation, is artificial intelligence. AI is following keywords had to be in the title, in the keywords, or in both:
2
“supply chain” and “artificial intelligence.” This ensured that the obtained from the International Journal of Production Research (23%),
retrieved articles strongly correlated to the different elements of AI and Expert Systems with Applications (19%), and the International Journal of
supply chain processes. To limit the number of papers selected for re Production Economics (15%). A detailed journal-wise distribution of the
view, we excluded conference contributions, articles published in trade 136 articles is listed in Appendix 1.
journals, books, and book chapters from the search process. We also
based our selection on the ABS ranking by considering only ABS ranking
3 journals, since the third level comprises highly cited journals that 3.2. Tool and process
publish original and well-executed research papers that are research-
oriented and meet high standards for scholarly rigour which is We used RStudio mapping software for the descriptive bibliometric
compatible with the aim of our study to review the literature. It is also analysis. The R environment provides numerous packages related to
the most adopted ranking in the research field (Skjølsvik et al., 2017). bibliometrics. As presented in Section 3.1, we performed the data
Avoiding ABS ranking 2 journals is due to the orientation of the articles collection by extracting research articles from the Scopus database. Aria
that are usually practitioner oriented. In addition, this category pub and Cuccurullo (2017) explained how bibliometric analyses can be
lishes original research of only acceptable standard of quality. The same performed using R. Aria and Cuccurullo (2017) pointed out that R,
can be said about ABS level 1 ranking journals that publish research of a which is an open-source software with numerous statistical capabilities,
modest standard of quality and rigour. In March 2020, the output of this is an ideal option for scientific computing. In our analysis, we followed a
literature search was a set of 136 papers published in the timespan of standard five-stage science mapping workflow: study design, data
1996–2020. All the fields were selected during the process of filtering, collection, data analysis, data visualization, and interpretation (Zupic &
because the aim of the study is to explore the applications of AI in SC in Čater, 2014).
all domains and sectors. Most of the articles selected for review are All study designs involve defining research question(s) and selecting
a suitable bibliometric method to answer them. Three broad types of
Records identified through Additional records identified

Identification
database searching through other sources

(n = 1478) (n = 10)
Records after duplicates removed

(n = 538)
Screening
Records screened Records excluded

(n = 502)
(Not employing AI)
(n = 36)
Full-text articles Full-text articles

assessed for eligibility excluded, with reasons
(n = 152)
Studies addressed in
Eligibility
the conference, book

chapter, book reviews
were excluded.
Only ABS Ranking 3
Journals other than
and above journals
English were excluded.
were considered for
review. (n = 22)
Included
Studies included for

the review
(n = 136)
Fig. 1. Material Collection Process.
3
research questions can be answered using bibliometrics: (i) the cognitive Table 1
content and intellectual structure of a topic or field, (ii) the conceptual Methodological Table: Research Questions, Data Involved, Tools Applied, and
structure of a topic or field, and (iii) social networks based on specific Data Sources.
social factors (authors, countries, etc.). In this paper, we answer the Research Data Representation Tool for the Source
three research questions. Although bibliometric methods usually reveal Question Analysis
the structure of a field in a different way compared to what traditional What are the Author Co-occurrence biblioNetwork Scopus
literature reviews reveal, they are not an alternative to reading research general keywords network ()
articles. Researchers with in-depth knowledge of the field are well- research Paper
trends of AI classification
positioned to better interpret the findings (see Fig. 1).
applications
Data analysis and data visualization are often software-assisted and in supply
involve multiple substeps (Aria & Cuccurullo, 2017) supported by the chains?
recommended science mapping workflow depicted in Fig. 2; this can be What are the Author Co-occurrence biblioNetwork Scopus
performed using the Bibliometrix package. supply chain keywords, network ()
outcomes abstracts, titles Historical direct histNetwork()
that AI citation histPlot()
3.3. Approach achieves? Bibliographic network
data frame
Paper
The topic “AI and supply chain” as a subject for review is abstract; classification
therefore, analyzing published work and extant literature as the main What level of AI Author Co-occurrence biblioNetwork Scopus
source of analysis seemed compelling (Jauch et al., 1980). Given that the drives such keywords, network ()
study aim is to review the existing literature related to AI and supply outcomes? abstracts, titles Paper
classification
chains, we performed a descriptive systematic literature review, as
Mayring (2014) suggested. We adopted a methodology featuring a
three-step iterative approach: (i) material collection, (ii) descriptive the authors are focused on, namely, the topic, approach, and typology of
analysis, and (iv) material evaluation. We chose this method based on the paper. In contrast, indexed keywords are specified by academic
our research questions (Table 1). Similar to the review methodologies search systems to categorize papers coherently and accurately (Pozzi &
that Seuring and Müller (2008) and Gao et al. (2017) adopted, we Strozzi, 2018) based on vocabulary, antonyms, and synonyms and tak
deduced the dimensions involved in our review process from the state- ing into consideration diverse spellings, plurals, and term updates.
of-the-art supply chain literature before analyzing the content from We approached RQ2 by analyzing a historiographic map, a type of
the selected papers. graph that Garfield (2004) proposed. It is meant to illustrate a chrono
We addressed RQ1 by analyzing author keyword co-occurrence logical network map of highly significant direct citations originating
networks. Author keywords are the keywords the author(s) of a paper from a bibliographic set. We also answered the research question by
have specified. They reflect the scope of the study and the dimensions
Fig. 2. Bibliometric and Recommended Science Mapping Workflow.
4
classifying the papers. We addressed RQ3 using author keyword co- reference framework. Unlike other models, the SCOR model provides a
occurrence networks. unique framework that links business processes, metrics, best practices,
and technology features in a unified structure to support exchange
3.4. Review classification framework among supply chain participants, and consequently, to improve effi
ciency and to promote activities for improving supply chain manage
We also performed a classification of the material collected accord ment. The processes defined in the SCOR model provide a set of
ing to four critical structural dimensionslevel of analytics, AI algorithms predefined descriptions for the activities that most companies perform
or techniques, sector or industry of application, and supply chain pro to effectively run their supply chains.
cesses (Appendix 2). The dimensions used for classifying the research
papers could be derived deductively or inductively, allowing traceability 4. Results and Analysis
and inter-subject verifiability as compared to other qualitative inter
pretive methods (Seuring & Gold, 2012). Before performing the classi We present the results of this study as a descriptive analysis con
fication of the content material, we based our choice of dimensions on sisting of the annual production of articles, most frequent sources, most
the existing literature review and we selected the most representative productive authors, most relevant sources, and top manuscripts per ci
dimensions with respect to our research questions and the aim of our tations. We also visualized various types of network mappings pertain
study to gain a descriptive insight into the literature. ing to word, authorship, and citation, and we produced a historiographic
diagram of the most powerful articles from the past to the present. In
3.4.1. Level of data analytics addition, we present the conceptual structure of the field.
We classified the papers into three analytic levels: descriptive, pre
dictive, and prescriptive. Descriptive analytics is performed to assess
historical data and extract trends; it helps users to identify problems and 4.1. Descriptive analysis
areas for improvement and is very useful for reporting using mathe
matics and statistical tools and techniques. Predictive analytics helps 4.1.1. Year-wise publication details
users to predict future events or patterns based on historical data by Fig. 3 illustrates the number of published papers related to AI and
using different statistical and AI techniques. Prescriptive analytics ex supply chain. The trend in the publications on AI and supply chain re
tends beyond the historical perception of descriptive analytics and the flects increased focus on the topic since 2008. The selected 136 papers
future outcomes of predictive analytics to produce with recommenda (133 articles and 3 reviews) were published in 23 journals. A list of each
tions through multi-criteria decision-making (MCDM) techniques, journal’s contribution is indexed in Appendix 1.
simulation, and optimization (Riahi & Riahi, 2018).
4.1.2. Country co-authorship analysis
3.4.2. AI algorithms or techniques The country co-authorship analysis revealed a pattern of cooperation
We exhaustively analyzed the papers to extract the AI algorithm or among authors from different countries (see the network depicted in
technique that each author used to conduct their review or research. We Fig. 2—the nodes represent authors). The focal country’s interaction
extracted any algorithm or technique that was cited in the article or used with other countries is represented by the node size, which is propor
to solve a problem. We were careful to specify the exact algorithm or tional to the total number of countries in which the focal country has
technique in the classification, as referencing the category to which the worked. The links represent co-authorships, and the link width repre
algorithms belonged could be vague. Identifying each algorithm or sents the total number of collaborations between the focal country and
technique improved our understanding of how various supply chain other countries. A network depicting the collaboration among authors
problems were approached in terms of the level of AI involved in solving can be obtained by using the general formula:
the problem. This information can be used to establish a correspondence Bcoll = A × A (1)
between certain types of problems and the algorithms that are best
suited for their resolution. where A is a Document × Author matrix, element bij indicates how many
collaborations exist between authors i and j and the diagonal element bii
3.4.3. Sector or industry of application is the number of documents authored or co-authored by researcher i.
We classified the papers based on the sector or industry in which AI The country-wise collaboration network between 1996 and 2020 is
was used to address a supply chain problem. This dimension reflects the shown in Fig. 4. The US and the UK produced the highest number of
fields in which AI for supply chains is gaining rapid traction as well as papers in collaboration with many different countries. Countries such as
those that still lack a clear perspective for AI adoption in their supply France, India, and China collaborated with many countries but produced
chain. It also enabled us to detect which sectors remained far from comparatively low numbers of papers. According to the betweenness
adopting AI to transform their supply chain, and thus, need more centrality analysis, the US, the UK, and France were the top countries
attention and research to assist them in moving toward and through this linking the remaining countries in the collaboration network
revolutionary approach. This dimension can help companies to under (betweenness centrality measures the frequency with which a node is on
stand the importance and the potential of AI to transform their supply the shortest path between two other nodes).
chain.
4.1.3. Citation analysis of published articles
3.4.4. Supply chain processes There are two main co-citation analyses categories (document co-
Classifying the selected papers based on supply chain processes gave citation analysis and author co-citation analysis) that can be used to
us insight into the processes in which AI techniques are applied and into visualize the structure of an entire domain of study. For this study, we
the supply chain management activities that AI techniques support. used document co-citation analysis to explore the intellectual structure
The classification based on this dimension also revealed the main of AI and supply chains. We created a historiographic map for the
sources of data that can be collected and used to improve the different analysis. Garfield (2004) proposed the historiographic map as a graph
supply chain processes. With this classification, we aimed to empirically that sequentially represents the most significant direct citations from a
represent AI’s contribution in supply chain processes to indicate bibliographic set. Fig. 5 depicts the chronological mapping of the most
research activities that could be carried out in the future. relevant citations from the selected studies.
To determine the supply chain processes to be considered, we used Co-citation analysis determines the occurrence of authors, refer
the supply chain operations reference (SCOR) model, Version 11 as a ences, or journals in a third document; this analysis is based on the
5
Articles
15 15 15
16 14
14
11 11
12 10
10 8
7 7
8
5 5
6 4
4 2 2
1 1 1 1 1
2
0
1995 2000 2005 2010 2015 2020 2025
Fig. 3. Year-Wise Distribution of Papers.
Fig. 4. Collaboration Network of Countries.
implication that co-cited articles might address the same topic. The existing linkages among various papers in the field of AI and supply
following general formula can be used to generate a co-citation network: chain by mapping out the historical evolution of the field, thus helping
scholars to identify the most important work therein.
Bcocit = A × A (2)
Examining the number of citations reflects the quality of a document
where A is a Document × Cited reference matrix, element bij indicates how (Liao et al., 2018). Swaminathan et al.’s (1998) paper ranked first owing
many co-citations exist between documents i and j, and the main diag to the number of citations (584), showing the popularity and influence
onal of Bcocit contains the number of documents where a reference is of this paper in the supply chain field. This work established a frame
cited in our data frame (diagonal element bii is the number of local ci work that reduces the effort involved in modeling various supply chain
tations of the reference i). reengineering alternatives and in assessing their efficiency by means of
The direction of the arrows in the historiographic map depicted in simulations that consider a multitude of assumptions about un
Fig. 5 represents the chronological change in research trends from the certainties regarding supply, demand, and process in a supply chain. The
past. For example, Giannoccaro & Pontrandolfo (2002) described an second most cited article (298) was by Wolfert et al. (2017). The authors
approach composed of three techniques: (i) Markov decision processes reviewed state-of-the-art uses of big data in smart farming and identified
(MDP), and (ii) an AI algorithm to solve MDPs (which is based on [iii] the associated socio-economic concerns. The study reported that inte
simulation modeling to manage inventory decisions at all stages of the grating big data into smart farming not only enhanced production but
supply chain in an integrated manner). Kadadevaramath et al. (2012) also impacted food supply chain integrity. Smart farming can benefit
wrote about modeling and optimizing a three-echelon supply chain from AI and new technologies by introducing robots, enabling the use of
network using particle swarm optimization AI algorithms. This pro predictive insights, and making adequate operational decisions based on
moted research dealing with green logistics and swarm intelligence real-time data.
using different algorithms. As shown in Fig. 5, starting from 1998, there
was a substantial growth in interest in the field. Swaminathan et al. 4.1.4. Semantic network analysis
(1998) laid the foundation for many other studies about the historical Semantic network analysis is defined as a method for analyzing the
citation network map. Our historical direct citation analysis displays the relationships between words in individual texts. In this study, we
examined the frequency distributions of words and the co-occurrences of
6
Fig. 5. Historical Direct Citation Network from the Historical Network.
the keywords the authors used. We used co-word analysis to detect co- 4.1.5. Authors’ productivity
occurrences of keywords or terms retrieved from the title, abstract, or In terms of author performance, K. L. Choy (h-index; 6, g-index: 8)
body of a document. We then examined the co-occurrence networks of published the highest number of studies. Choy, the most productive
the authors’ keywords, generated using the following formula: author, published an initial paper in 2002 and published eight papers
between 2002 and 2018. Although A. De Meyer published only four
Bcoc = A × A (3)
papers between 2015 and 2018, this author is one of the 10 most pro
where A is a Document × Word matrix and Word is alternatively the ductive authors in the revise. H. C. W. Lau has not published anything
authors’ keywords, keywords plus, or the terms extracted from the titles since 2011. From these results, we deduced a strong positive correlation
or abstracts of the papers. Element bij indicates how many co- between the number of citations and author productivity. The 10 most
occurrences exist between words i and j. The diagonal element bii is productive authors did not publish any papers in 2020, and only two of
the number of documents containing the word i. these authors published in 2019. Fig. 7 shows the authors’ productivity
We divided the 50 most frequently occurring keywords into four over time based on the number of papers and the annual total citations
clusters (Fig. 6). The radius of the circle shows the frequency of the (see Figs. 8–10).
words, and the color reflects a specific cluster. The largest cluster is in
purple and consists of 14 keywords. The highest value of betweenness in 4.2. Conceptual structure of the field
the purple cluster is related to the enable process (according to the SCOR
model) (166,871). The keywords pertaining to AI in the other clusters Author’s keywords co-occurrence analysis reflets clusters of key
include “artificial intelligence” (86,110), “artificial neural networks” words. They are considered as themes, whose density and centrality can
(3,085), and “machine learning” (27,062). In short, this means that be used in classifying themes and mapping in a two-dimensional dia
many supply chain studies deal with AI. gram called a thematic map. The keywords are distributed in four
quadrants which reflects the maturity level of the theme.
Fig. 6. Co-Occurrence Network Based on Authors’ Keywords.
7
Fig. 7. Top Authors’ Productivity Over Time.
Fig. 8. Thematic Map.
The upper-right quadrant is composed of motor-themes which are relatively new or under development. It is until 2006 that this field have
characterized by both high centrality and density. This means that these known an increase in term of research. In addition, with the increasing
themes are developed and important for the research field. The lower- VUCA (Volatility, uncertainty, complexity and ambiguity) aspect of the
right quadrant refers to the basic and transversal. They are character environment in which the supply chain operates nowadays, especially
ized by high centrality and low density. These themes are important for during crisis like the coronavirus pandemic in 2020, SCs requires
a research field and refers to general topics transversal to the different developed technologies such as AI allowing prediction, forecasting,
research areas of the field. The lower-left quadrant is composed of assistance in decision-making and assistance in the various processes of
emerging or disappearing themes. They have both low centrality and the SC. Consequently, AI adoption in SC is far from being a mature field
density meaning that are weakly developed and marginal. The upper- as it is still being researched and built. For this reason, we can notice that
left quadrant contains very specialized or niche themes. They have the upper right quadrant in the thematic map is empty.
well developed internal links (high density) but unimportant external The upper-left quadrant is composed of a theme related to inventory
links and so are of only limited importance for the field (low centrality) management through the principles of systems dynamics. As this
(Corte et al., 2019). quadrant refers to a very specialized niche, the AI technique is easily
Since the field of AI in SC is gaining more attention over the years, identifiable, which is ANN in this case. This theme is very related to the
which is justified in Fig. 3, the themes related to the field are still field of AI for SC, however, is of only limited importance for the field.
8
Predictive processing techniques. To take advantage of this important asset, new

analytics big data analytical techniques are required to extract the knowledge that
4% can increase supply chain efficiency (Hashem et al., 2014). We reviewed
all the selected papers based on the level of analytics: descriptive, pre
scriptive, and predictive. We found that the majority of the articles used
descriptive analytics (57%; 77 out of 136 papers) followed by pre
scriptive analytics (39%; 53 out of 136 papers). A very small number of
Prescriptive studies used predictive analytics (4%; 6 out of 136 papers).
analytics Descriptive
39% analytics 5.1.1. Descriptive analytics
57% A large number of papers selected for the review used descriptive
analytics (77 out of 136 papers) to analyze the correlation between AI
and supply chains. They used descriptive analytics to examine the use of
new technologies and AI to enhance agriculture supply chain perfor
mance (Lezoche et al., 2020). This type of analytics was also used to
Fig. 9. Level of analytics distribution.
investigate the impact of AI adoption on waste reduction and quality
improvement (Fröhling et al., 2010; Georgiadis & Athanasiou, 2013;
Make Return Zhou & Piramuthu, 2013). Many studies that demonstrated the benefit
of emerging technologies such as RFID (Radio Frequency Identification)
5% 2%
(Werthmann et al., 2017; Zhou & Piramuthu, 2013) to transform supply
chains used descriptive analytics.
Source
5.1.2. Prescriptive analytics
10%
Enable Thirty-nine percent of the selected papers used prescriptive ana
39% lytics. Most used multi-criteria decision-making (MCDM) techniques to
Deliver design decision support systems for supplier selection (Scott et al.,
14% 2015). A combination of MCDM and the analytic hierarchy process
(AHP) was also used for the same objective (Dweiri et al., 2016). Data
envelopment analysis (DEA) was used in research on reducing green
house gas emissions in closed-loop supply chains (CLSC) (Neto et al.,
Plan 2010).
30%
5.1.3. Predictive analytics
Only six of the 136 selected papers used predictive analytics. Its
Fig. 10. Supply chain processes distribution.
application covered the prediction of adopting interorganizational sys
tems (IOS) for a well-integrated and collaborative supply chain network
Again, since the field of AI & SC is still relatively new, the pertaining (Chong & Bai, 2014) and for developing predictive early warning and
themes are still under development and profound research, which jus proactive control systems in food supply chain networks to reduce
tifies the emptiness of the lower-left quadrant. However, we can notice performance deviations (Li et al., 2010). Finally, it was applied in a
that the theme of logistics, and more precisely facility location and co predictive decision support system to manage the supply chain of
ordination can be considered a basic theme and an emerging theme. The packaged fresh and highly perishable products through sales forecasting
probability of this theme going towards disappearance because of the (Dellino et al., 2018).
growing interest in this theme as it is justified by the paper’s classifi
cation in the Appendix. The lower right quadrant constitutes basic 5.2. AI-driven supply chain
themes, in our analysis it is about supply chain management and opti
mization, which is a transversal theme related to all the areas of the AI & With reference to the classification of the selected papers based on
SC. This theme pertains to the enable process, for which the full po the AI algorithm or technique used, 9% of the papers compared different
tential of AI can be invested to manage and organize all the activities of AI algorithms. The most used technique was genetic algorithms (14
the SC. papers)—these decreased both the bullwhip effect (BWE) and the cash
flow bullwhip (CF-BWE) across supply chains (Badakhshan et al., 2020).
5. Classification Results Genetic algorithms were also used for risk assessment in food supply
chains (Chang et al., 2017). The second most used technique was arti
We adopted four critical structural dimensions to organize the clas ficial neural networks (ANNs) (5 papers) as a tool for enhancing logistics
sification framework for our research and answer the research questions: workflow responsiveness (Lee at al., 2011). One study proposed a
level of analytics, types of AI algorithms used in the supply chain field, comparative forecasting methodology for uncertain customer demands
applications of AI across the different sectors and industries, and AI use in a multi-level supply chain structure via neural techniques (Efendigil
across supply chain processes. The classification results are presented in et al., 2009). Many AI techniques were used, which justifies the fact that
Appendix 2. even if we could identify the most used techniques, the number of ar
ticles using them remains limited. Almost every paper used a new AI
5.1. Data-driven supply chain technique, such as decision trees, intelligent agents, bio-inspired algo
rithms, and particle swarm intelligence. It is important to note that many
A large volume of data is constantly generated due to the growing use papers did not explicitly use AI; however, they reviewed a set of algo
of smart equipment, sensors, and IoT in supply chains. Various tech rithms or initiated its use in the respective field of the study.
nologies such as IoT and cloud computing are driving traditional supply
chains to become smart supply chains. The data generated are usually
too large and complex to be processed using conventional data-
9
5.3. Sector or industry of AI application in supply chains literature (41%), descriptive analytics was used in 31 papers, prescrip
tive analytics was used in 21 papers, and predictive analytics was used in
AI’s emergence in the supply chain context has caused radical three papers.
change in the organization of work processes. Any sector can benefit It is important to emphasize that the use of predictive analytics in the
from the right integration of AI in its processes and become a more literature was very limited (2%) and distributed across the plan process
proactive, predictive, automated, and personalized sector. Our review (Dellino et al., 2018; Nikolopoulos et al., 2016; Yu et al., 2019) and the
highlights that AI has been used in diverse sectors—food (Di Giacomo & enable process (Chong & Bai, 2014; Li et al., 2010; Rodger, 2014). This
Patrizi, 2010; Li et al., 2010), retail (Salmi & Holmström, 2004; Van Aart might be due to the nature of the two processes, which are focused on
et al., 2004), automotive (Kellner et al., 2019; Werthmann et al., 2017), ensuring a balance between supply and demand, making predictive
and manufacturing (Kristianto et al., 2017; Kumar et al., 2017) represent analytics the best option for predicting demand and managing the
the four major sectors that integrated AI into supply chains. AI was also supply chain through business rules and supply chain network man
used in supply chains pertaining to healthcare (Ghandforoush & Sen, agement. Thus, predictive analytics is the best fit because prediction
2010; Nabelsi & Gagnon, 2017), oil (Kallestrup et al., 2014; Repoussis enables the supply chain to be more proactive, improves operational
et al., 2009; Yu et al., 2019), remanufacturing (Fröhling et al., 2010; productivity, and fosters strategic decision-making.
Georgiadis & Athanasiou, 2013; Zhou & Piramuthu, 2013), and several The findings from Table 3 indicate that descriptive and prescriptive
other sectors such as agriculture, bioenergy, green supply chain, and analytics were the most used techniques for analyzing the different is
aerospace. sues within the supply chain and the likelihood of AI adoption to solve
It is important to note that the majority of the papers considered a different problems and to increase performance (See Tables 4–8).
simulation-based approach (26%; 35 out of 136), which does not ac
count for any specific sector and focuses on an application in a real case 5.4.2. AI techniques and SCOR processes
study. We analyzed the distribution of the different AI techniques used
across the various SCOR areas in order to gain insights into which AI
techniques or algorithms were used across different supply chain
5.4. Supply chain processes processes.
Genetic algorithms were the most used type of algorithm in the plan
Our analysis of the distribution of the selected papers based on the process (6 papers), followed by ANNs (4 papers). Many other techniques
different supply chain processes (SCOR areas) revealed that the enable were evenly distributed across the selected papers. Studies addressing
process received higher attention from researchers (41%; 55 out of 136 issues related to the source area mainly focused on reviewing different
papers); this was followed by the plan process (31%; 42 out of 136 pa AI techniques. The same trend can be noticed in the research dealing
pers) and the deliver process (15%; 21 out of 136 papers). The source with the make process. Genetic algorithms and particle swarm optimi
process (11%; 15 out of 136 papers), the make process (5%; 7 out of 136 zation were the predominant AI techniques used in papers related to the
papers), and the return process (2%; 3 out of 136 papers) received very
little research attention. No paper studied all six processes simulta
neously. The sum of the papers that studied each process was greater Table 3
than the total number of selected papers because some papers consid AI Techniques Across SC Processes.
ered more than one process in their research. AI Technique Plan Source Make Deliver Return Enable
None 17 11 5 11 2 32
5.4.1. Level of data analytics and SCOR processes Adaptive tabu search 1
To gain deeper insight into the type of analytics used to process the (ATS) algorithm
collected data, we analyzed the papers according to the level of data AI algorithm review 2 3 1 1 4
ANN 4 3
analytics that was used in the different supply chain processes, also Artificial immune 1
referred to as the SCOR areas. The results are presented in Table 2. system algorithm
Table 2 reveals that the plan process received significant focus in the Artificial WD 1
literature (42 papers). Twenty-one papers in this area used descriptive Association rule 1
mining
analytics through reviews and empirical studies. Of the remaining pa
Bioinspired algorithms 1
pers in the plan process, 18 papers used prescriptive analytics, and only C4.5 algorithm 1
three papers used predictive analytics. Of the 15 papers in the source Clustering 1
process, eight papers used prescriptive analytics, whereas seven papers Data mining algorithm 1
used descriptive analytics. In the make process, five papers used Decision tree 2 1 1 1 1 1
Dimensionality 1
descriptive analytics, and two papers used prescriptive analytics.
reduction
Descriptive analytics and prescriptive analytics were almost equally Genetic algorithm 6 3 5
used in the deliver process, with 11 and 10 papers, respectively. Of the Gradient descent 1
three papers focusing on the return process, two papers deployed algorithm
Intelligent agent 1
descriptive analytics, and one paper used prescriptive analytics. In the
Intelligent 1
enable process, which was the most studied supply chain area in the optimization
algorithms
Table 2 Kalman filter 1
Multi-objective 1
Level of Analytics Across SC Processes.
artificial bee colony
Supply Chain Processes Level of Analytics Nearest neighbor 1 1
Particle swarm 1 1 2
Descriptive Prescriptive Predictive
intelligence
Plan 21 18 3 Particle swarm 2
Source 7 8 0 optimization
Make 5 2 0 Regression trees 1 1
Deliver 11 10 0 SMART algorithm 1 1
Return 2 1 0 Support vector 1
Enable 31 21 3 machine
10
Table 4 genetic algorithms (five papers), followed by neural networks and par
AI for demand forecasting references. ticle swarm intelligence. Although many papers did not use any AI
Algorithm Key Contributions Author(s) techniques in their study, they formed part of our review because they
initiated the use of AI in the different listed sectors.
ANN Realized a daily demand predicting system in a Slimani et al.
supermarket using MLP by adding inputs that (2017) Genetic algorithms received significant attention in the literature
included previous demand, days’ classification, due to their potential to solve optimization problems; this is because
and average demand quantities they have few mathematical requirements and can handle any kind of
ANN Developed a forecasting model for retailers Bala (2012) objective function and constraint (Xu & Ding, 2011). This explains their
based on customer segmentation to improve
performance of inventory
wide use in the plan, deliver, and enable processes—these processes are
SVM Investigated the applicability and benefits ofML Carbonneau et al. best positioned to use optimization. The second most used technique
techniques in forecasting distorted demand (2007) was ANN, which was mostly used in the plan and enable processes.
signals with high noise in supply chains Neural networks are a family of statistical learning models inspired by
biological neural networks that are used to estimate or approximate
functions that might depend on a large number of inputs, such as de
Table 5 mand data in the plan process and the data generated by the partners in
AI for risk management and resilience references. the enable process. Most of the articles that used ANNs also used the
Algorithm Key Contributions Author(s) technique for planning.
ANN/SVM Used a decision support model incorporating an Rajesh
amalgamation of grey theory and layered analytic (2020) 6. Discussion
network process (ANP) to quantify various resilient
strategies for risk mitigation 6.1. Discussion of the findings
ANN Developed a forecasting model for retailers based on Bala (2012)
customer segmentation to improve inventory
performance
This study is the first attempt to explore the research work from 1996
Decision Proposed a two-stage decision support system (DSS) Mogre et al. to 2020 related to AI applications in SC from a descriptive point of
tree that would assist managers in selecting mitigation (2016) view—we used a systematic literature review combined with biblio
strategies for supply chain risks and mitigation metric analysis for this exploration. This work also presents a granular
tactics when risks occurred
classification of the papers. Combined, these elements were sufficient to
answer our research questions.
Table 6 6.1.1. RQ1: What are the general research trends of AI applications in
AI for transportation references. supply chains?
Algorithm Key Contributions Author(s) Based on the extant literature and the analysis of the output, we find
Genetic Proposed a novel technique for modeling Mokhtarinejad et al. that AI-related research has increased over the years. The present find
algorithm and solving location, routing, and (2015)
ings are restricted to the supply chain field due to the intention of this
scheduling problems via cross-docking in
the supply chain analysis to explore the correlation between AI and supply chain using
related keywords. Our analysis of authors’ keyword co-occurrence
network revealed the association of AI with different sub-domains of
supply chains, namely demand forecasting, supplier selection, supply
Table 7
chain network design, supply chain risk management, inventory man
AI for supplier selection references.
agement, sustainability, logistics, supply chain process management,
Algorithm Key Contributions Author(s)
and supply chain integration. The common element between these
ANN Proposed a new intelligent model to predict Vahdani et al. supply chain sub-fields, is their requirement of a decision-making pro
the performance rating of suppliers in the (2012) cess which justifies the association of these subfields to the different AI
cosmetics industry
Genetic Presented a new intelligent model using a Fallahpour et al.
algorithms. Artificial Intelligence techniques are growingly enhancing
algorithm genetic algorithm to solve the suppliers’ (2017) and enriching decision support through data analysis, data trends
performance evaluation and prioritization detection, forecasting and anticipation.
problems From a process perspective, the enable process gained more attention
ANN Described a framework using fuzzy logic and Lau et al. (2002)
by researchers when considering the use of AI, than the other processes.
neural networks for handling supplier
selection The enable process includes processes associated with supply chain
management such as business rules, data resources, contracts, compli
ance and risk management. This finding is in full agreement with the
previous one, and confirms that the management and decision making
Table 8
related activities in SC, consider the implementation of the different AI
AI for inventory management references.
techniques and algorithms and are trending in the research works due to
Algorithm Key Contributions Author(s)
the increasingly growing decision support potential of AI and the
ANN Explored the feasibility and relevance of ADP Çimen and awareness that the success of the supply chain requires more and more
algorithms to optimize inventory decisions Kirkbride (2016) attention to the management and consideration of the different parties
Genetic Provided empirical evidence that significant Borade and
algorithm economic benefits can be achieved by using a Sweeney (2014)
involved in a supply chain, but also to the benefits of an integrated,
genetic algorithm-based (GA-based) decision sustainable and adaptable supply chain. The diversity of the AI tech
support system (DSS) in a vendor-managed niques, provides the decision-makers with a wide selection to choose
inventory (VMI) supply chain from taking into consideration the SC process, the data required and the
aimed outputs.
deliver process. We deployed a decision tree to discuss topics pertaining
6.1.2. RQ2: What are the supply chain outcomes that AI achieves?
to the return process. In the enable area, authors relied on the review of
AI for demand forecasting. AI and machine learning can be applied to
different AI techniques (four papers); the most used technique was
improve demand forecasting. This is one of the most promising
11
applications of AI for supply chains. The exceptional performance of AI After determining the level of AI that drives such outcomes, we
for demand forecasting translates into tangible benefits across the sup further explored our results from an AI classification point of view
ply chain, starting with improved supplies which lead to fewer product (Fig. 11). Following the literature review, we conducted an in-depth
shortages, fewer overstocks and fewer waste (especially for food prod reading of the papers, finding that most of them concentrated on the
ucts). Planning is also improved, making it possible to optimize storage use of machine learning—specifically, reinforcement learning using
capacities or even reception / dispatch. AI methods offer a lot for time ANNs. For example, Ponte et al. (2015) proposed the application of AI
series forecasting, thanks to their potential of extracting models from techniques to reduce the bullwhip effect using a forecasting system
input data over long periods of time, thus they havec v perfect appli based on ANNs. Moraga et al. (2011) used ANNs to recognize behavioral
cability in demand forecasting. patterns and to predict future performance accordingly. Machine
AI for risk management and resilience. SCRM is a suitable field of learning is used for analyses because it can analyze a superior amount of
application for AI due to its resilient strategies that commonly rely on information and come up with structured results. Machine learning in
fast decision-making based on potentially large, multidimensional data creases the cost-effectiveness of the whole process while unlocking
sources. AI is a very promising technology for the optimization of the knowledge that was unobtainable using earlier technologies (see
supply chain and building its resilience. Indeed, it makes it possible to Fig. 12).
process and correlate a large number of data, to facilitate understanding
and to anticipate the impacts of external events. 7. AI-driven SC framework
AI for transportation. The structure of distribution systems can in
crease the transportation costs and reduce the ability to efficiently meet With reference to the review findings, we report that AI has signifi
customers’ demands. AI-based models provide near-optimal solutions cant potential for applications across different SC processes. The find
for a broad range of routing problems, thus ensuring on-time deliveries, ings indicate that the SC can be transformed when AI is adopted. Yet, it is
optimizes the transport of goods in the warehouse important to remind that a successful adoption of AI within the SC re
whether it is performed by robots, automated storage solutions, or quires a set of conditions which we refer to as success factors for AI
operators assisted by handling equipment. AI can make use of real-time adoption. This success is also dependent on the selection of the most
traffic data, robotics, computer-vision and autonomous vehicles, all of adapted AI algorithms (Bashir et al., 2020). Based on the findings, the
which can help build specific models to improve transportations. proposed framework has three key elements
AI for supplier selection. Supplier evaluation and selection constitute a Success factors for AI adoption, AI techniques, and AI-driven SC
critical and complex multi-criteria decision-making procedure that operations. Through this framework we aim to guide practitioners to
directly affects the supply chain. One of the methods for assisting a wards a successful adoption of AI techniques within the SC, and provide
company’s supplier selection process is AI, where machines can be researchers with a foundation for further research regarding the factors
trained by decision-makers or by using historical data to make pre that contribute to the success of AI adoption within the SC and the
dictions and recommendations. Artificial intelligence-based models like factors that should be considered for the selection of the AI algorithm to
ANN have been used extensively to solve supplier selection due to its use.
ability to deal with multiple constraints (quality, delivery, performance
history, production capability, service, engineering and technical 7.1. Success factors for AI adoption
capability, business structure, price, integrity, warranties, honesty,
reliability, reputation, commitment and financial position) which must In a world of data abundance, businesses are looking to gain a
be satisfied simultaneously, can predict new outcomes on past trends, competitive advantage through the use of this important asset. With the
and can process information at high speed. availability of large amounts of data, Artificial intelligence (AI) is being
AI for inventory management. Inventory is a primary source of cost and adopted in a wide range of operational settings, but many organizations
has a considerable influence on responsiveness. It increases the number are still at an early stage of adoption or fail to implement AI the right
of demands that can be met by ensuring product availability and read way (Mir et al., 2020). The correct implementation of AI can produce
iness at the right time. One of the most effective ways inventory man huge positive results and can transform the SC. Before considering the
agement can benefit from AI is by the automation of the process. For adoption of AI, SC managers should consider many factors, which we
instance, an artificial intelligent system would be able to track sales refer to as “Success factors for AI adoption”. Three broad factors should
automatically, thus stores data for monitoring the inventory in real- be considered, namely the data, knowledge of AI field, and SC capabil
time, avoiding overstocking or understocking. AI algorithms can also ities to adopt AI. The availability of data is very important in any AI
generate reports automatically regarding the change in demand, hence project. All AI algorithms require large dataset, especially for learning.
saving the time that could be spent on estimating the stock. Artificial Real-time data is also important for decision-making oriented AI projects
Neural Networks modeling is the most used technique in inventory as an AI algorithm requires real-time data to process in order to generate
management thanks to its capability of handling data with high vola real-time outputs. The external data is required when the organization is
tility more accurately. ANN thus overcomes the shortcomings of tradi using AI in a large scope and includes or invites the stakeholders in the
tional forecasting models. transformation of its SC. When it comes to knowledge of AI field, certain
AI techniques, such as scorecards and decision trees, are easy to un
6.1.3. RQ3: What level of AI drives such outcomes? derstand. But neural networks are more complex and mysterious. AI
Before determining the AI level that drives the previously reported should be transparent in terms of inputs, processes, and decisions. Or
outcomes, it is essential to define the three different AI levels. The first ganizations needs to understand at least briefly, the functioning of an
level is artificial narrow intelligence (ANI), also referred to as “weak AI” algorithm and how the decision was made. Thus, the users will be able to
or “narrow AI.” ANI is goal-oriented and is designed to perform singular manipulate the algorithm so they can view, approve, and override the
tasks. The second level is artificial general intelligence (AGI), also decisions. This is what we called “the black box effect” in our frame
referred to as “strong AI” or “deep AI.” This describes a machine’s ability work. Users should also acquire some deep knowledge in AI, to
to completely and successfully perform tasks like a human being. manipulate the algorithms to fit their needs. When it comes to the SC
Currently, AGI does not exist. The third level is artificial super intelli capabilities, and from the review of the literature, we conclude that one
gence (ASI). To be categorized as ASI, an AI needs to outperform human of the most important capabilities that a SC should develop is the
capabilities which does not exist yet. From this brief introduction to the adaptability, which is the ability to adjust a supply chain’s design to
three different levels of AI, the outcomes stated in response to RQ2 are meet structural shifts the markets and modify supply network in stra
driven by the first level of AI. tegies and technologies (Lee, 2004). Thus, the SC will be able to flow
12
Fig. 11. Artificial Intelligence classification.
with the AI adoption requirements and the changes it will bring along. resilience analysis or demand forecasting. The second category is
Also, the SC should develop its capability of learning, since the adoption modelling in which AI can be used to train a model in order to recognize
of AI without perspectives and ability to learn from it, the identification certain types of patterns. It can be used to solve routing problems for
of what works and what does not, will make no sense and won’t bring example and usually leads to an optimization. The third category is
any positive changes to the SC. control in which AI can exploit real-time data for decision-making, thus
controlling the different SC critical operations like risk control and in
7.2. AI techniques ventory control. Finally, there is learning. In the supply chain, AI aims
above all to improve operational performance, by automating certain
The second element in the framework consists of the AI techniques processes such as demand forecasting or production planning. The
that are used in the different SC processes. The literature review advantage is to be able, thanks to its analytical capacity, to correlate
revealed that AI is adopted for transforming the SC and improving its very quickly large volumes of data. AI adoption with the aim to learn
performance. Yet, of the large number of available AI techniques and from data or from past event will make it possible to carry out much
algorithms, only a few are explored and exploited in SC processes. In the finer projections and analyzes situations despite an increasingly unsta
framework, we adopted a classification of AI to present its richness and ble economic environment and more volatile demand. Machine learning
the unexplored algorithms for SC applications. Yet, the matching be in particular allows organizations to discover new patterns in the data,
tween the SC processes and the AI techniques is deduced from the per thus assist in decision-making, adaptability, more resilience and a more
formed systematic literature review. Therefore, studies should be developed SC maturity.
conducted to profoundly identify the factors that should be considered It is very important for organizations to position themselves in terms
when selecting the AI techniques to use in each SC process. of objectives or goals they aim to achieve when considering the adoption
of AI in their SCs as it will help structure the required data, build a
strategy for AI adoption and select the adequate AI techniques.
7.3. AI-driven SC operations
8. Implications of the study
The review findings reveal that AI adoption by SCs not only increases
performance, lowers costs, and reduces losses, but also brings about
8.1. Theoretical implications
complete changes to the SC, making it adaptable, agile, and resilient.
Each SC process can inevitably benefit from the potential of AI. In our
Based on the findings and the its discussions, we present the
framework we grouped these benefits onto four categories: analysis,
following broad areas as needing further investigations from
modelling, control, and learning.
researchers.
AI is very powerful at finding insights and patterns in large datasets
From a close reading of the papers reviewed in this study, we
that humans just can’t see, thus can generate deep and complete analysis
conclude that simulation data are the main data source for the studies,
regarding different SC operations, for instance performance analysis,
13
Fig. 12. AI-driven SC framework.
caused by the difficulty of acquiring historical data related to the supply 8.2. Practical implications
chain. Researchers should consider basing their studies on data gener
ated by the different supply chain structures. Future researchers should also examine data-processing approaches
With real supply chain data, future researchers could suggest which related to the application of AI for an enhanced operations management
AI methods can be applied best to certain supply chain tasks—currently, in different sectors and industries. Prediction and supplier selection are
many researchers try different AI algorithms for problem-solving and frequently addressed by using AI—this is mainly because manually
then compare the results of each algorithm to decide which one per made decisions might involve mistakes. Decision-making is not a process
forms better for the problem under consideration. In addition, the that is exclusive to supplier selection—this process is present in every
findings of this study shed light on a wider range of less-popular AI supply chai task; thus, other supply chain fields should be explored, and
techniques that might become more influential in future research. decision support tools must be improved. Future studies should evaluate
the impact of AI adoption on the human workforce. Identifying how AI
integration could impact the human workforce is crucial for maintaining
the right equilibrium among all the components in a supply chain.
14
Future studies may also compare the performance of different AI algo with an in-depth exploration of collaborations and theoretical bases of
rithms and techniques, to assist practitioners in the selection of the AI research. This study contributes to the body of knowledge through the
algorithms to opt for according to the need of their supply chain process. analysis of the evolution of AI applications in the supply chain and the
Finally, researchers should develop a framework which reflects the identification of research trends. The results of the analysis revealed that
transformational potential of AI in the field of supply chain. A frame the number of publications is increasing, and that more attention must
work could be very beneficial to practitioners, as it would help the be paid to AI applications in supply chains. Descriptive analysis showed
supply chain parties assess their state of SC maturity, identify the aspects a gradual increase in the number of publications related to AI since
requiring enhancement and decide the strategy to follow to achieve a 2010; this increase also applies to work in supply chain and supply chain
positive transformation of the SC. management. Thus, AI is used increasingly in supply chains as a new
solution to both traditional and novel problems. The main contribution
9. Limitations of this study is the classification of the research articles in order to
assimilate the current state of literature in this field and to understand
Our study has several limitations that we outline here. While we the role of AI in supply chains, in addition to the development of a AI-
conducted a literature review using the Scopus database to retrieve the driven SC framework, which can help academics and practitioners to
relevant material related to the topic, it is possible that we may have understand the factors that should be considered before the adoption of
missed papers during the selection process because of keywords choice AI and the importance of envisioning and setting the expected outputs of
for example. In addition, the analysis is based on our interpretation and AI adoption in a SC. The framework constitutes a foundation for further
thus dependent upon our perceptions and classifications of the collected research in this direction.
material.
CRediT authorship contribution statement
10. Conclusion
Youssra Riahi: Formal analysis, Software, Visualization. Tarik
This study was based on a systematic literature review combined Saikouk: Conceptualization, Supervision. Angappa Gunasekaran:
with bibliometric analysis to investigate the current state of research on Supervision. Ismail Badraoui: .
AI applications in the supply chain. We performed this review using 136
research articles from the Scopus database published between 1996 and Declaration of Competing Interest
2020. We found that among the three machine-learning algorithms
(supervised, unsupervised, and reinforcement learning), there was a The authors declare that they have no known competing financial
focus on reinforcement learning in the supply chain. We examined the interests or personal relationships that could have appeared to influence
“how” and “what” of the extant research on AI applied to supply chains the work reported in this paper.
Appendix 1
Journal-Wise Distribution of Selected Papers
Source Articles
International Journal of Production Research 32

Expert Systems with Applications 26
International Journal of Production Economics 21
Decision Support Systems 15
European Journal of Operational Research 13
Production Planning and Control 6
Computers in Industry 3
Computers and Operations Research 2
IEEE Transactions on Engineering Management 2
IEEE Transactions on Systems Man and Cybernetics: Systems 2
Journal of the Operational Research Society 2
Decision Sciences 1
European Journal of Information Systems 1
IEEE Transactions on Systems Man and Cybernetics Part A: Systems and Humans 1
Information Systems Research 1
International Journal of Forecasting 1
International Journal of Human Computer Studies 1
Journal of Computer Information Systems 1
Journal of Enterprise Information Management 1
Journal of Operations Management 1
Manufacturing and Service Operations Management 1
MIS Quarterly: Management Information Systems 1
Supply Chain Management: An International Journal 1
15
Appendix 2
Systematic Literature Review
Author(s) Level of Data AI Algorithm/Technique Sector/Industry of SC

Analytics Application Processes
Plan Source Make Deliver Return Enable
Cai and Lo (2020) PS AI algorithms review Retail √

Lezoche et al. (2020) DS AI algorithms review Agri-food √
Ren et al. (2020) PS – Reverse logistics
Flores and Villalobos PS Dimensionality Reduction Agriculture √
(2020)
Badakhshan et al. (2020) PS Genetic algorithm Distribution √
Brevik et al. (2020) PS Genetic algorithm Food √
Bag et al. (2020) DS AI algorithms review Green supply chain √
Govindan et al. (2019) DS Swarm intelligence Vehicle routing √
Fowler et al. (2019) PS – Semiconductor supply √
chain
Priore et al. (2019) DS C4.5 algorithm Wholesaling √
Puche et al. (2019) PS – Simulation-based √
Baryannis et al. (2019) DS SVM / ANN / Bayesian networks Review √
Schtter et al. (2019) PS – Resource allocation in √
food retail
Sharahi and Khalili- PS Genetic algorithm Natural gas industry √
Damghani (2019)
Zhao et al. (2019) DS – Retail √
Kellner et al. (2019) DS – Automotive √
Dubey et al. (2019) DS AI algorithms review Manufacturing √
Threr et al. (2019) DS AI algorithms review Infrastructure √
development
Azzamouri et al. (2019) DS – Fertilizer Production √
Uppari and Hasija (2019) DS – Laboratory √
Yu et al. (2019) PD Decision tree Oil industry √
Lechner and Reimann DS – Reprocessing √
(2019)
Papagiannidis et al. (2018) DS Clustering Urbanism √
Villegas and Pedregal DS Kalman filter Grocery retail √
(2018)
Hong et al. (2018) DS Particle swarm Optimization Liquor industry √
Gupta et al. (2018) DS – Process industry √
Hombach et al. (2018) DS – Biofuel √
Wang et al. (2018) DS – Vehicle routing √
Dellino et al. (2018) PD Genetic algorithm Food √
Kusiak (2018) DS Data mining algorithms Manufacturing √
Simchi and Wu (2018) DS Regression trees / random forest Online retail √
Ma et al. (2018) PS Logistic regression / support Manufacturing √ √ √ √ √
vector Machine / decision tree
Kannan et al. (2018) DS – Textile industry √
Yazdani et el. (2017) PS – Agriculture √
Wruck et al. (2017) PS – E-commerce √
Wanke et al. (2017) DS Genetic algorithms / ANN Retail √
Bogataj et al. (2017) PS – Food delivery √
Santos et al. (2017) PS – Pharmaceutical industry √
Guarnaschell et al. (2017) PS – Wood industry √
Nabelsi and Gagnon PS – Healthcare √ √
(2017)
Werthmann et al. (2017) DS – Automotive √
Imen et al. (2017) DS ANN Simulation-based √
Xu et al. (2017) DS AI algorithms review Simulation-based √
Chang et al. (2017) DS Genetic algorithm Food √
Zhang et al. (2017) PS Bioinspired algorithm Simulation-based √
Kumar et al. (2017) PS Artificial immune systems / Manufacturing √
particle swarm optimization
Kristianto et al. (2017) DS AI algorithms review Manufacturing √
Zhang et al. (2016) PS Multi-objective artificial bee Simulation-based √
colony
Moncayo and PS Artificial WD Simulation-based √
Mastrocinque (2016)
Dev et al. (2016) DS Decision tree Simulation-based √
Dweiri et al. (2016) PS – Automotive √ √
Charles et al. (2016) PS – Humanities √ √
Mogre et al. (2016) PS – Offshore wind √
Nikolopoulos et al. (2016) PD Nearest neighbor Automotive √
Syntetos et al. (2016) DS – Review √
Zimmer et al. (2016) DS AI algorithms review Review √
(continued on next page)
16
(continued )
Erskine et al. (2016) DS – Geospatial √

Kumar et al. (2015) PS – Green supply chain √
Borade and Sweeney PS Genetic algorithm Food manufacturing √
(2015)
Peiris et al. (2015) DS – Engineering, √
procurement, and
construction
Court et al. (2015) DS – Multi sector √
De Matta and Miller DS – Simulation-based √
(2015)
Scott et al. (2015) PS AI algorithms review Bioenergy √
Guo et al. (2015) PS Intelligent optimization Clothing manufacturing √
algorithms
Rodger (2014) PD Genetic algorithm Aerospace industry √
Mascle and Gosse (2014) PS AI algorithms review Food manufacturing √
Daaboul et al. (2014) PS – Shoemaking industry √
Kumar (2014) DS – Food √
Holimchayachotikul et al. DS Particle swarm intelligence – √
(2014)
Boza et al. (2014) PS – Ceramic industry √
Theißen and Spinler PS – Fast-moving consumer √
(2014) goods
Groves et al. (2014) DS – Simulation-based √
Miao et al. (2014) PS Adaptive tabu search (ATS) Simulation-based √
algorithm
Chong and Bai (2014) PD Neural network Multi-sector √
Ting et al. (2014) DS Association rule mining Red wine industry √
Kallestrup et al. (2014) PS – Oil industry √
Roozbeh et al. (2014) DS Genetic algorithm Simulation-based √
Yang and Fung (2014) PS – Simulation-based √
Manzini et al. (2014) DS Nearest neighbor Distribution √
Zhou and Piramuthu DS – Remanufacturing √ √
(2013)
Latha et al. (2013) PS Genetic algorithm Pump manufacturing √
Banerjee and Golhar DS – Fashion retail √
(2013)
Liu et al. (2013) DS – Electronics √
Dong and Srinivasan DS – retail / manufacturing √
(2013)
Lenny at el. (2013) PS – Malt √
Georgiadis and Athanasiou DS – Remanufacturing √
(2013)
Latha et al. (2013) DS Particle swarm intelligence Simulation-based √
Moon et al. (2012) PS Genetic algorithm Simulation-based √
Turrado et al. (2012) PS Support vector machine Retail √
Kadadevaramath et al. PS Particle swarm intelligence Simulation-based √
(2012)
Ferreira and Borenstein DS AI algorithms review Oil industry √
(2012)
Hahn and Kuhn (2012) DS – Retail- √
Erdem and Göen (2012) PS – White goods √
manufacturing
Kristianto et al. (2012) PS – Simulation-based √
Tako and Robinson (2012) DS – Review √
Okongwu et al. (2012) PS – Simulation-based √
Flisberg et al. (2012) PS – Petroleum industry √
Khataei et al. (2012) DS – Simulation-based √
Marchetta et al. (2012) PS Intelligent agent Automotive √
Lee et al. (2011) PS ANN Jewelry industry ✓
Kumar et al. (2011) DS Artificial immune system Simulation-based √
algorithm
Ferreira and Borenstein DS ANN Petroleum industry √
(2011)
Ghandforoush and Sen DS – Healthcare √
(2010)
Li et al. (2010) PD Neural networks Food √
Warre and Chang (2010) DS – Simulation-based √
Frhling et al. (2010) DS – Remanufacturing √
Zhou et al. (2010) PS – Food √
Di Giacomo and Patrizi DS – Review √
(2010)
Qu et al. (2010) DS Genetic algorithm Simulation-based √
Gorgiadis et al. (2010) DS – Remanufacturing √
(continued on next page)
17
(continued )
Boran and Goztepe (2010) DS – Simulation-based √

Collins et al. (2010) DS – Simulation-based
Kumar et al. (2010) DS Genetic algorithm / Particle Simulation-based √
swarm optimization / artificial
bee colony
Arora et al. (2010) PS – Simulation-based √
Quariguasi et al. (2010) PS – Electric and electronic √
industry
Lauras et al. (2010) PS – Aerospace √
Threr et al. (2010) DS – Simulation-based √
Hu and Bidanda (2009) DS – Electronics industry √
Chatfield et al. (2009) DS – Simulation-based √
Repoussis et al. (2009) DS – Oil industry √ √
O’Donnell et al. (2009) DS Genetic algorithm Simulation-based √
Effendigil et al. (2009) DS ANN Consumer goods industry √
Shang et al. (2008) DS – Pharmaceutical industry √
Jain et al. (2008) DS – Simulation-based √
Baptiste et al. (2008) DS – Simulation-based √
Thammakoranonta et al. PS – Simulation-based √ √
(2008)
Nissen and Sengupta DS – Review √
(2006)
Van et al. (2004) DS – Retail √
Salmi and Holmström PS – Retail √
(2004)
Schneeweiss (2003) DS – Review √
Giannoccaro and PS SMART algorithm Simulation-based √
Pontrandolfo (2002)
Pontrandolfo et al. (2002) PS SMART algorithm Simulation-based √
Ba et al. (2001) DS – Review √
Lin and Pai (2000) DS Gradient descent algorithm Simulation-based √
Swaminathan et al. (1998) DS – Simulation-based √
Authors DS AI algorithms review Review √ √ √ √ √ √
References Di Giacomo, L., & Patrizi, G. (2010). Methodological analysis of supply chains
management applications. European Journal of Operational Research, 207(1),
249–257.
Aria, M., & Cuccurullo, C. (2017). Bibliometrix: An R-tool for comprehensive science
Dweiri, F., Kumar, S., Khan, S. A., & Jain, V. (2016). Designing an integrated AHP based
mapping analysis. Journal of Informetrics, 11(4), 959–975.
decision support system for supplier selection in automotive industry. Expert Systems
Badakhshan, E., Humphreys, P., Maguire, L., & McIvor, R. (2020). Using simulation-
with Applications, 62, 273–283.
based system dynamics and genetic algorithms to reduce the cash flow bullwhip in
Efendigil, T., Önüt, S., & Kahraman, C. (2009). A decision support system for demand
the supply chain. International Journal of Production Research.
forecasting with artificial neural networks and neuro-fuzzy models: A comparative
Bala, P. K. (2012). Improving inventory performance with clustering-based demand
analysis. Expert Systems with Applications, 36(3 Part 2), 6697–6707.
forecasts. Journal of Modelling in Management, 7(1), 23–37.
Fahimnia, B., Sarkis, J., & Davarzani, H. (2015). Green supply chain management: A
Bashir, M. U., Swapnil, S., Kumar, K. A., & Prasad, G. M. (2020). Critical success factors
review and bibliometric analysis. International Journal of Production Economics, 162,
for integrating artificial intelligence and robotics. Digital Policy, Regulation and
101–114.
Governance, 22, 307–331.
Fallahpour, A., Wong, K. Y., Olugu, E. U., & Musa, S. N. (2017). A predictive integrated
Borade, A. B., & Sweeney, E. (2014). Decision support system for vendor managed
genetic-based model for supplier evaluation and selection. International Journal of
inventory supply chain: A case study. International Journal of Production Research, 53
Fuzzy Systems, 19(4), 1041–1057.
(16), 4789–4818.
Fröhling, M., Schwaderer, F., Bartusch, H., & Rentz, O. (2010). Integrated planning of
Borges, A. F. S., Laurindo, F. J. B., Spínola, M. M., Gonçalves, R. F., & Mattos, C. A.
transportation and recycling for multiple plants based on process simulation.
(2020). The strategic use of artificial intelligence in the digital era: Systematic
European Journal of Operational Research, 207, 958–970.
literature review and future research directions. International Journal of Information
Gao, D., Xu, Z., Ruan, Y. Z., & Lu, H. (2017). From a systematic literature review to
Management, 102–225.
integrated definition for sustainable supply chain innovation (SSCI). Journal of
Carbonneau, R., Vahidov, R., & Laframboise, K. (2007). Machine learning-based demand
Cleaner Production, 142, 1518–1538.
forecasting in supply chains. International Journal of Intelligent Information
Garfield, E. (2004). Historiographic mapping of knowledge domains literature. Journal of
Technologies, 3(4), 40–57.
Information Science, 30(2), 119–145.
Chang, Y., Erera, A. L., & White, C. (2017). Risk assessment of deliberate contamination
Georgiadis, P., & Athanasiou, E. (2013). Flexible long-term capacity planning in closed-
of food production facilities. IEEE Transactions on Systems, Man, and Cybernetics:
loop supply chains with remanufacturing. European Journal of Operational Research,
Systems, 47(3), 381–393.
225(1), 44–58.
Chong, A.-Y.-L., & Bai, R. (2014). Predicting open iOS adoption in SMEs: An integrated
Ghandforoush, P., & Sen, T. K. (2010). A DSS to manage platelet production supply chain
SEM-neural network approach. Expert Systems with Applications, 41(1), 221–229.
for regional blood centers. Decision Support Systems, 50(1), 32–42.
Çimen, M., & Kirkbride, C. (2016). Approximate dynamic programming algorithms for
Giannakis, M., & Papadopoulos, T. (2016). Supply chain sustainability: A risk
multidimensional flexible production-inventory problems. International Journal of
management approach. International Journal of Production Economics, 171, 455–470.
Production Research, 55(7), 2034–2050.
Giannoccaro, I., & Pontrandolfo, P. (2002). Inventory management in supply chains: A
Collin, J., Eloranta, E., & Holmström, J. (2009). How to design the right supply chains for
reinforcement learning approach. International Journal of Production Economics, 78,
your customers. Supply Chain Management, 14, 411–417.
153–161.
Čerka, P., Grigienė, J., & Sirbikytė, G. (2015). Liability for damages caused by artificial
Hashem, I., Yaqoob, I., Anuar, N., Mokhtar, S., Gani, A., & Khan, S. (2014). The rise of
intelligence. Computer Law & Security Review, 31, 376–389.
“big data” on cloud computing: Review and open research issues. Information
Corte, V., Del Gaudio, G., Sepe, F., & Sciarelli, F. (2019). Sustainable Tourism in the Open
Systems, 47, 98–115.
Innovation Realm: A Bibliometric Analysis. Sustainability, 11, 6114.
Jauch, L. R., Osborn, R. N., & Martin, T. N. (1980). Structured content analysis of cases: A
Dellino, G., Laudadio, T., Mari, R., Mastronardi, N., & Meloni, C. (2018). A reliable
complementary method for organizational research. The Academy of Management
decision support system for fresh food supply chain management. International
Review, 5(4), 517.
Journal of Production Research, 56(4), 1458–1485.
18
Kadadevaramath, R. S., Chen, J. C. H., Latha Shankar, B., & Rameshkumar, K. (2012). Ponte, B., Ruano, L., Pino, R., & De Fuentela, D. (2015). The bullwhip effect in water
Application of particle swarm intelligence algorithms in supply chain network demand management: Taming it through an artificial neural networks-based system.
architecture optimization. Expert Systems with Applications., 39, 10160–10176. Journal of Water Supply: Research and Technology. – AQUA, 64, 290–301.
Kallestrup, K. B., Lynge, L. H., Akkerman, R., & Oddsdottir, T. A. (2014). Decision Pozzi, R., & Strozzi, F. (2018). How assembly systems are adopting the technologies of
support in hierarchical planning systems: The case of procurement planning in oil I40: A preliminary landscape. Industrial Systems Engineering, 369–375.
refining industries. Decision Support Systems, 68, 49–63. Rajesh, R. (2020). A grey-layered ANP based decision support model for analyzing
Kellner, F., Lienland, B., & Utz, S. (2019). An a posteriori decision support methodology strategies of resilience in electronic supply chains. Engineering Applications of
for solving the multi-criteria supplier selection problem. European Journal of Artificial Intelligence, 87, Article 103338.
Operational Research, 272, 505–522. Repoussis, P. P., Paraskevopoulos, D. C., Zobolas, G., Tarantilis, C. D., & Ioannou, G.
Kristianto, Y., Gunasekaran, A., & Helo, P. (2017). Building the “Triple R” in global (2009). A web-based decision support system for waste lube oils collection and
manufacturing. International Journal of Production Economics, 183, 607–619. recycling. European Journal of Operational Research, 195(3), 676–700.
Kumar, V. N. S. A., Kumar, V., Brady, M., Garza-Reyes, J. A., & Simpson, M. (2017). Riahi, Y., & Riahi, S. (2018). Big data and big data analytics: Concepts, types and
Resolving forward-reverse logistics multi-period model using evolutionary technologies. International Journal of Research Engineering, 5(9), 524–528.
algorithms. International Journal of Production Economics, 183, 458–469. Rodger, J. A. (2014). Application of a fuzzy feasibility Bayesian probabilistic estimation
Lau, H. C. W., Hui, I. K., Chan, F. T. S., & Wong, C. W. Y. (2002). Monitoring the supply of of supply chain backorder aging, unfilled backorders, and customer wait time using
products in a supply chain environment: A fuzzy neural approach. Expert Systems, 19 stochastic simulation with Markov blankets. Expert Systems with Applications, 41(16),
(4), 235–243. 7005–7022.
Lee, H. L. (2004). The triple-A supply chain. Harvard Business Review, 82, 102–113. Salmi, L., & Holmström, J. (2004). Monitoring new product introductions with sell-
Lee, C. K. M., Ho, W., Ho, G. T. S., & Lau, H. C. W. (2011). Design and development of through data from channel partners. Supply Chain Management: An International
logistics workflow systems for demand management with RFID. Expert Systems with Journal, 9(3), 209–212.
Applications, 38(5), 5428–5437. Scott, J., Ho, W., Dey, P. K., & Talluri, S. (2015). A decision support system for supplier
Lezoche, M., Panetto, H., Kacprzyk, J., Hernandez, J. E., & Díaz, M. M. E. A. (2020). Agri- selection and order allocation in stochastic, multi-stakeholder and multi-criteria
food 4.0: A survey of the supply chains and technologies for the future agriculture. environments. International Journal of Production Economics, 166, 226–237.
Computers in Industry, 117. Seuring, S., & Gold, S. (2012). Conducting content-analysis based literature reviews in
Li, Y., Kramer, M. R., Beulens, A. J. M., & Van Der Vorst, J. G. A. J. (2010). A framework supply chain management. Supply Chain Management: An International Journal, 17(5),
for early warning and proactive control systems in food supply chain networks. 544–555.
Computers in Industry, 61(9), 852–862. Seuring, S., & Müller, M. (2008). From a literature review to a conceptual framework for
Liao, H., Tang, M., Luo, L., Li, C., Chiclana, F., & Zeng, X.-J. (2018). A bibliometric sustainable supply chain management. Journal of Cleaner Production, 16(15),
analysis and visualization of medical big data research. Sustainability, 10, 1–18. 1699–1710.
Mayring, P. (2014). Qualitative content analysis: Theoretical background and Skjølsvik, T., Pemer, F., & Løwendahl, B. (2017). Strategic management of professional
procedures. Approaches to Qualitative Research in Mathematics Education, 365–380. service firms: Reviewing ABS journals and identifying key research themes. Journal
Mentzer, J. T., DeWitt, W., Keebler, J. S., Min, S., Nix, N. W., Smith, C. D., & of Professions and Organization, 4, 203–225.
Zacharia, Z. G. (2001). Defining supply chain management. Journal of Business Swaminathan, J. M., Smith, S. F., & Sadeh, N. M. (1998). Modeling supply chain
Logistics, 22, 1–25. dynamics: A multiagent approach. Decision Sciences, 29(3), 607–632.
Mir, U., Sharma, S., Kar, A., & Gupta, M. P. (2020). Critical success factors for integrating Tawfik, G. M., Dila, K. A. S., Mohamed, M. Y. F., Tam, D. N. H., Kien, N. D.,
artificial intelligence and robotics. Digital Policy, Regulation and Governance, 22(4), Ahmed, A. M., & Huy, N. T. (2019). A step by step guide for conducting a systematic
307–331. review and meta-analysis with simulation data. Tropical Medicine and Health, 47, 46.
Mogre, R., Talluri, S. S., & Damico, F. (2016). A decision framework to mitigate supply Vahdani, B., Iranmanesh, S. H., Mousavi, S. M., & Abdollahzade, M. (2012). A locally
chain risks: An application in the offshore-wind industry. IEEE Transactions on linear neuro-fuzzy model for supplier selection in cosmetics industry. Applied
Engineering Management, 63(3), 316–325. Mathematical Modelling, 36(10), 4714–4727.
Mokhtarinejad, M., Ahmadi, A., Karimi, B., & Rahmati, S. H. A. (2015). A novel learning- Van Aart, C. J., Wielinga, B., & Schreiber, G. (2004). Organizational building blocks for
based approach for a new integrated location-routing and scheduling problem within design of distributed intelligent system. International Journal of Human Computer
cross-docking considering direct shipment. Applied Soft Computing, 34, 274–285. Studies, 61(5), 567–599.
Moraga, R., Rabelo, L., Jones, A., & Vila, J. (2011). Using neural networks to monitor Werthmann, D., Brandwein, D., Ruthenbeck, C., Scholz-Reiter, B., & Freitag, M. (2017).
supply chain behaviour. International Journal of Computer Applications in Technology, Towards a standardised information exchange within finished vehicle logistics based
40, 53–63. on RFID and EPCIS. International Journal of Production Research, 55(14), 4136–4152.
Min, H. (2010). Artificial intelligence in supply chain management: Theory and Wolfert, S., Ge, L., Verdouw, C., & Bogaardt, M.-J. (2017). Big data in smart farming – A
applications. International Journal of Logistics Research and Applications, 13, 13–39. review. Agricultural Systems, 153, 69–80.
Nabelsi, V., & Gagnon, S. (2017). Information technology strategy for a patient-oriented, Yong-Hak, J. (2013). Web of science. Thomson Reuters. http://wokinfo.com/media/pdf/
lean, and agile integration of hospital pharmacy and medical equipment supply WoSFS_08_7050.pdf.
chains. International Journal of Production Research, 55(14), 3929–3945. Yu, L., Zhao, Y., Tang, L., & Yang, Z. (2019). Online big data-driven oil consumption
Neto, J. Q. F., Walther, G., Bloemhof, J., van Nunen, J. A. E., & Spengler, T. (2010). From forecasting with Google Trends. International Journal of Forecasting, 35(1), 213–223.
closed-loop to sustainable supply chains: The WEEE case. International Journal of Xu, J., & Ding, C. (2011). A class of chance constrained multiobjective linear
Production, 48, 4463–4481. programming with birandom coefficients and its application to vendors selection.
Ngai, E. W. T., Peng, S., Alexander, P., & Moon, K. K. L. (2014). Decision support and International Journal of Production Economics, 131, 709–720.
intelligent systems in the textile and apparel supply chain: An academic review of Zhang, S., Lee, C. K. M., Wu, K., & Choy, K. L. (2016). Multi-objective optimization for
research articles. Expert Systems with Applications, 41, 81–91. sustainable supply chain network design considering multiple distribution channels.
Ni, D., Xiao, Z., & Lim, M. K. (2020). A systematic review of the research trends of Expert Systems with Applications, 65, 87–99.
machine learning in supply chain management. International Journal of Machine Zhao, D., & Strotmann, A. (2015). Analysis and visualization of citation networks: Synthesis
Learning and Cybernetics, 11, 1463–1482. lectures on information concepts, retrieval. and services.
Nikolopoulos, K. I., Babai, M. Z., & Bozos, K. (2016). Forecasting supply chain sporadic Zhou, W., & Piramuthu, S. (2013). Remanufacturing with RFID item-level information:
demand with nearest neighbor approaches. International Journal of Production Optimization, waste reduction and quality improvement. International Journal of
Economics, 177, 139–148. Production Economics, 145(2), 647–657.
Phulwani, P. R., Kumar, D., & Goyal, P. (2020). A Systematic Literature Review and Zupic, I., & Čater, T. (2014). Bibliometric methods in management and organization.
Bibliometric Analysis of Recycling Behavior. Journal of Global Marketing, 33, Organizational Research Methods, 18(3), 429–472.
354–376.
19

Expert Systems With Applications: Youssra Riahi, Tarik Saikouk, Angappa Gunasekaran, Ismail Badraoui

Uploaded by

Copyright:

Available Formats

Expert Systems With Applications: Youssra Riahi, Tarik Saikouk, Angappa Gunasekaran, Ismail Badraoui

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Expert Systems With Applications: Youssra Riahi, Tarik Saikouk, Angappa Gunasekaran, Ismail Badraoui

Uploaded by

Copyright:

Available Formats

Expert Systems With Applications 173 (2021) 114702

Contents lists available at ScienceDirect

Expert Systems With Applications

Artificial intelligence applications in supply chain: A descriptive

1. Introduction operate—and to create more agile and resilient supply chains—these

Records identiﬁed through Additional records identiﬁed

database searching through other sources

Records after duplicates removed

Records screened Records excluded

Full-text articles Full-text articles

the conference, book

Studies included for

Fig. 1. Material Collection Process.

Fig. 2. Bibliometric and Recommended Science Mapping Workflow.

Fig. 3. Year-Wise Distribution of Papers.

Fig. 4. Collaboration Network of Countries.

Fig. 5. Historical Direct Citation Network from the Historical Network.

Fig. 6. Co-Occurrence Network Based on Authors’ Keywords.

Fig. 7. Top Authors’ Productivity Over Time.

Fig. 8. Thematic Map.

Predictive processing techniques. To take advantage of this important asset, new

Fig. 11. Artificial Intelligence classification.

Fig. 12. AI-driven SC framework.

Journal-Wise Distribution of Selected Papers

International Journal of Production Research 32

Systematic Literature Review

Author(s) Level of Data AI Algorithm/Technique Sector/Industry of SC

Plan Source Make Deliver Return Enable

Cai and Lo (2020) PS AI algorithms review Retail √

Plan Source Make Deliver Return Enable

Erskine et al. (2016) DS – Geospatial √

Plan Source Make Deliver Return Enable

Boran and Goztepe (2010) DS – Simulation-based √

You might also like