Annals of Operations Research

https://doi.org/10.1007/s10479-020-03862-8

S.I.: ARTIFICIAL INTELLIGENCE IN OPERATIONS MANAGEMENT

Impact of artificial intelligence investment on firm value

Ariel K. H. Lui1 · Maggie C. M. Lee1 · Eric W. T. Ngai2

Accepted: 3 November 2020

© Springer Science+Business Media, LLC, part of Springer Nature 2021

Abstract
As artificial intelligence (AI) has recently gained momentum and attention, the interest and
investment in AI have also accelerated. However, the impact of AI on firm value is rarely
discussed. On the basis of the 119 announcements of 62 listed firms who have invested in AI,
this study finds that AI investment has a negative impact on the firms’ market value. The stock
prices of the firms decrease by 1.77% on the day of the announcement. Nonmanufacturing
firms and firms with weak information technology capabilities or low credit ratings suffer a
more negative impact compared with other firms. The findings suggest that investors perceive
AI investment announcements to be unwelcome news for the majority of firms. Subsequently,
the characteristics affecting the shareholders’ reaction towards AI adoption are presented.
This research offers one of the first empirical evidence about the market value of AI and
provides a reference for firms interested in investing in AI.

Keywords Artificial intelligence · AI · Event study · Market value · Contextual factors

1 Introduction

Artificial intelligence (AI) is a disruptive technology. Companies across sectors have increas-
ingly harnessed AI’s power in firm operations, and managers have made major investments
in adopting AI to improve their operations. For instance, in the banking industry, Bank of
America has used a virtual assistant called Erica to help customers handle basic banking

B Ariel K. H. Lui
[email protected]
Maggie C. M. Lee
[email protected]
Eric W. T. Ngai
[email protected]
1 Department of Business Technology and Entrepreneurship, Swinburne Business School, Swinburne
University of Technology, Hawthorn, Australia
2 Department of Management and Marketing, The Hong Kong Polytechnic University, Hung Hum,
Kowloon, Hong Kong

123
 Annals of Operations Research

tasks while the Bank of New York Mellon Corp has implemented more than 220 “bots” to
handle repetitive tasks (Castellanos 2018). Analysts suggest that AI can help the banking
industry save more than US$1 trillion by 2030 by reducing 22% of their firms’ operating
expenses (Joyce 2018). McKinsey Global Institute also has reported that AI can provide an
additional $13 trillion per year to the global economic output by 2030 (Bughin et al. 2018).
According to a survey conducted in 2016 by Statista (2016), a provider of market research
and research and analysis services, revenues from AI for enterprise applications can grow
from $1.62 billion in 2018 to $31.2 billion in 2025. Moreover, 84% of enterprises believe
that AI implementation will spur more competitive edges, and 75% of them believe that AI
implementation will lead to new business opportunities. In addition, according to a study
published by Fortune (Murry 2017), more than 80% of Fortune 500 Chief Executive Officers
(CEOs) believe that AI is extremely critical to their firms’ futures.
Despite all the hype, some companies remain concerned about the technologies and hes-
itate to use them. Firms worry that machines are not armed with functions that mirror the
same goals as those of the management, and this gap may lead to incorrect decisions or cause
fatal mistakes in operations. First, AI adoption will likely be a substantial challenge for orga-
nizations since they may experience reputational damage, revenue losses, and diminished
public trust if the AI machine falters. Second, costs are another issue that may prevent firms
from using AI technology. Apart from the costs associated with the implementation (i.e.,
system costs, project and consulting fees, system integration costs, and system upgrading
costs), various other costs will likely be incurred when organizations transit to AI (Bughin
et al. 2018). The transition costs may include those costs incurred while restructuring the
organization, severance for workers who may be displaced by the new technology, and the
costs to upgrade the skills of existing employees (Bughin et al. 2018). Therefore, the impact
of AI on firms may not be always positive as people expect them to be.
The objective of this research is to provide timely and objective evidence of the value of
AI investment. The investor perspective is considered to examine the value of AI. The event
study methodology is employed to examine the effect of AI investment on market value.
Since a stock market response is instantaneous and can reflect investors’ expectations of
future firm performance (Bose and Pal 2012), an event study is an appropriate approach for
this research. The event study is also widely cited in the literature as a means of investigating
how information technology (IT) investments affect firm value. Some previous studies have
shown that IT investments are positively correlated with the market value of firms (Dehning
et al. 2003; Subramani and Walden 2001), whereas other studies have reported a negative
relationship (Bose et al. 2011). Thus, we have searched and analyzed relevant articles from
2015 to 2019 to identify events that can reveal information about a firm’s initiative in adopting
AI. The rest of the study is organized as follows. Section 2 provides the theoretical back-
ground, and Sect. 3 enumerates the hypotheses. Section 4 presents the research method, and
Sect. 5 presents the result. The result of our investigation and the implications are discussed
in Sect. 6.

2 Theoretical background

2.1 AI as a disruptive technology in operations management

As a general term, AI refers to computer systems that perform tasks that usually require
human intelligence (Mueller and Massaron 2018). AI is specifically referred but not limited

123
Annals of Operations Research

to different types of technologies, such as expert systems, natural language processing, vir-
tual/intelligent agents, robotic process automation, speech recognition and machine vision,
artificial natural networks, etc. Moreover, AI is commonly regarded a disruptive technol-
ogy since it can disrupt operations in different industries, including agriculture (Bannerjee
et al. 2018), manufacturing (Benotsmane et al. 2019), banking (Fethi and Pasiouras 2010),
and healthcare (Fan et al. 2020; Jiang et al. 2017). Since the 1980s, numerous studies have
attempted to investigate and solve problems in operations management (OM) by using AI
(Kobbacy et al. 2007).
According to two detailed literature reviews conducted by Kobbacy and Vadera (2011)
and Kobbacy et al. (2007), over 1200 papers from 1995 to 2004 and over 1400 papers from
2005 to 2009 on the application of AI technologies in the areas of OM were published. AI
is employed in the following four areas of OM: design; scheduling; planning and control;
and quality, maintenance, and fault diagnosis (Kobbacy and Vadera 2011; Kobbacy et al.
2007). The commonly used AI technologies include case-based reasoning (CBR), genetic
algorithms (GAs), neural networks (NNs), knowledge-based systems, fuzzy logic, and hybrid
approaches (Kobbacy and Vadera 2011). Before 2005, fuzzy logic was heavily used in design
and scheduling, whereas NNs dominated process planning and control (Kobbacy et al. 2007).
From 2003 to 2004, a dramatic increase in the use of GAs was found for design and scheduling.
From 2001 to 2002, the use of NNs dominated and increased significantly for the other two
aspects (Kobbacy and Vadera 2011).
With the rapid development of computing and networking technologies in recent years, the
number of studies on AI in the OM domain has increased. Chou et al. (2010) presented and
compared multiple regression analysis, artificial NNs, CBR, and hybrid AI models to predict
the costs of the equipment used to manufacture thin-film transistor liquid–crystal displays.
Munguia et al. (2010) proposed a system that uses several AI technologies, such as fuzzy
logic, to guide the design of the optimum selection of production parameters according to the
product requirements in the early design stages. Lawrynowicz (2011) used GAs for job shop
scheduling in supply networks and collective scheduling in an industrial cluster. McNaught
and Chan (2011) described a Bayesian network-based intelligent decision support system
that helped operators in Motorola diagnose and correct faults during product system testing.
With the aim of reducing emissions from waiting trucks and improving crane operations at
container terminals, Do et al. (2016) developed a method that used a discrete event simulation
to measure the total truck waiting time and crane moving distance and then applied a GA
to reduce the emissions produced from these trucks and cranes. Kalayci et al. (2016) sug-
gested a hybrid algorithm that combines a GA with a variable neighborhood search method to
solve the sequence-dependent disassembly line balancing problem in reverse supply chains.
Aouadni and Rebai (2017) proposed a decision support system based on the multi-criteria
satisfaction analysis method and the continuous GA and multi-objective GA to help organi-
zations evaluate and measure their employees’ job satisfaction. Chen et al. (2020) presented
a multi-objective GA for energy-efficient hybrid flow shop scheduling with lot streaming to
reduce the completion time and energy costs at the manufacturing-system level. Yang et al.
(2019) proposed a grey-box model based on the GA to help shipping companies predict
fuel consumption and showed that the proposed model is remarkably better than the latest
grey-box model. More recently, Dosdoğru et al. (2020) developed a new hybrid AI-based
decision support framework to help resolve the inventory routing problem.

123
 Annals of Operations Research

2.2 Impact of AI investment on a firm

With the adoption of AI technologies becoming a leading trend in most companies to manage
their operations in recent decades, the interest of researchers has focused on investigating
whether AI technologies can ultimately bring benefits and value to companies or impact
them altogether. To the best of our knowledge, the number of studies that utilized qualita-
tive and quantitative approaches to assess the impact of AI on firms in practice, generated
through questionnaires and interviews with leaders and practitioners in specific industries,
is extremely limited. Jain (2019) collected data from 50 selected business decision-makers
and regular employees engaged in Indian firms through an online survey, and the results
of the analysis showed a significant impact of AI on the economic growth of their busi-
nesses. In a recent study about the impact of AI on pharmaceutical businesses, the results
of a questionnaire indicate that AI has a positive effect on the performance of such business
in Thailand (Chetthamrongchai and Jermsittiparsert 2020). In a survey conducted by Basri
(2020), AI-assisted social media marketing helped increase the number of customers and cus-
tomer bases and improved firm profitability. In a recently published work, Minevich (2020)
reported the process of how major players in some industries used and benefitted from their
AI projects. For example, Domino’s Pizza has reduced delivery time and improved delivery
time prediction accuracy from 75 to 95%, and mining companies in Australia have cut mining
costs, improved worker safety, and boosted productivity by 20% using autonomous trucks
and drilling technology. However, no academic research has quantified the economic values
created for firms through the adoption of AI. Obviously, we have identified a huge gap in the
AI and OM research literature that has not yet been explored.

2.3 IT’s influence on firms’ market value

Previous studies have used the event study methodology to examine the impact of various
types of IT investments on firm value (e.g., Dehning et al. 2003; Teo et al. 2016). Previous
studies show inconclusive findings, and no insignificant, negative, and positive relationships
were shown between IT investment and the market value of a firm. Studies in this field suggest
that the effects of IT investment on market value are contingent upon various contextual
factors, including the characteristics of the adopting firms, the IT investment type, and the
industry type. Table 1 summarizes the IT investment-related event studies.

3 Hypotheses

The prediction of the costs and benefits of applying AI determines the effect of the AI
investment on the market value of a firm. The market will weigh the discounted value of
the long-term benefits brought by the AI investment and the transitory costs. The market
value will grow if the expected long-term benefits are higher than the costs foreseen for the
investment and vice versa.
Investing in an emerging technology, such as AI, will generally increase the firm’s expen-
ditures and assets but reduce its cash flow. Moreover, AI usually involves huge investments
and high risk, and its benefits and payback time are difficult to measure. Other than the costs
associated with the implementation, such as the system costs, project and consulting fees,
and system integration and system upgrading expenses, other various costs are likely to be
incurred when organizations transit to AI (Bughin et al. 2018). Transition costs may include

123
Annals of Operations Research

Table 1 Prior event studies examining stock market reaction to IT investments announcements
References Investment Key findings

Dos Santos et al. (1993) IT Investment No abnormal returns on full sample

1.03% returns on innovative IT investment announcements
Chatterjee et al. (2002) IT Investment 1.99% returns on full sample
2.01% returns on Infrastructure announcements
0.84% returns on application announcements
Hunter (2003) IT investment − 0.85% returns for retail firms
− 0.68% returns on exploitative IT investment
− 1.11% returns on exploratory IT investment
No support found for size, slack, and time
Bose et al. (2011) RFID − 0.29% on full sample
US-based, low growth potential, and less diversified firms
experience more negative impact
Teo et al. (2016) Business analytics Positive return on full sample
More positive to firms with high sales growth and high
return on assets

the costs incurred while restructuring the organization, severance if some workers may be
displaced by new technologies, and the costs to upgrade the skills of existing employees
(Bughin et al. 2018). In addition, in a recent survey involving 29 company leaders from AI-
related companies, only 48% of them felt positive about their AI startup. Those AI-related
companies cannot easily prove that AI is beneficial to their company (Faggella 2019). There-
fore, investors may observe that the firms investing in disruptive technology, including AI,
are financially unattractive in the short term and thus may not choose the stocks of those
firms (Staw et al. 1981).
Nonetheless, information suggests that technology will bring long-term advantages, and
the costs are temporary in nature. Thus, the market may consider AI adoption as positive news
for a firm. In the last decade, the research and surveys conducted by different organizations
have suggested that AI is important and good for companies. International companies, such
as Amazon, Apple, and Facebook, are using AI as part of the business to automate their
operations successfully (Marr 2019). In addition, major players in different industries reported
their use of and benefits from AI. Domino’s Pizza has used AI to reduce delivery times
and predict delivery times more accurately while Barclay’s has used AI to detect fraud
and improve the customer experience through chatbots (Minevich 2020). An analyst also
predicted that the cost reduction in the operating expenses of financial institutions can reach
22% after using AI (Joyce 2018). This kind of good news is notably impressive to investors
and enables them to have great expectations of AI. Hence, investors may observe the firms
that claim they are implementing AI attractive and may consider investing in them.
Since AI implementation will have great impacts on the costs, risks, and revenues of com-
panies, investors will pay much attention on this technology and consider AI implementation
as one of the factors prior to making investment decisions. Some investors may regard AI
adoption as good news, but other investors may stay away from companies implementing AI.
Previous research has also illustrated the significant effects, both positive (Chatterjee et al.
2002; Teo et al. 2016) and negative (Hunter 2003; Tanriverdi and Ruefli 2004), of innovative
IT on the market value of a firm. Therefore, we hypothesize the following:

H1 The announcement of AI investment has a significant impact on a firm’s market value.

123
 Annals of Operations Research

IT capability is defined as an organization’s ability to identify information systems to

meet business needs, deploy these systems in a cost-effective manner, and provide long-term
maintenance and support for the systems (Karimi et al. 2007; Ross et al. 1996). IT capa-
bilities, including IT infrastructure capability and IT business spanning capability, enhance
organizational agility, such as operational adjustment agility (Lu and Ramamurthy 2011).
Prior research has found the association of IT capability with firm performance (Lu and
Ramamurthy 2011; Stoel and Muhanna 2009). Firms with a strong IT capability have better
IT expertise and IT infrastructure and thus are more likely to successfully implement new IT
technology than firms with weak IT capability. However, firms with a weak IT capability give
less effort to IT innovations compared with their counterparts with a strong IT capability.
Therefore, the firms with a weak IT capability are less capable of making the best use of IT
innovation or less likely to successfully implement new technology, such as AI. Thus, we
hypothesize the following:

H2 Firms with a weak IT capability experience a more negative impact on their market value
than firms with a strong IT capability.

Financial healthiness is an important factor in determining the success of IT adoption

(Hayes et al. 2001). Firms may need to commit large amounts of money to acquire physical
and human resources for the implementation of AI (Bughin et al. 2018). Hence, sufficient
resources are needed for successful AI implementation. However, the expenses required in
adopting AI may increase the debt or reduce the cash flow of the company. Financially
unhealthy companies are expected to suffer more negative effect of AI investment than
financially healthy firms. If the return of the AI investment cannot cover the costs of investing
in the technology, then the financially unhealthy firm is likely to experience a poor cash flow.
A third-party credit rating can be used to measure a firm’s financial healthiness (White
2010). A low credit rating suggests the difficulty of a firm to pay its debts, whereas a high
credit rating indicates the financial strength of a firm and its ability to pay its debts. We can
consider low-credit rating firms as financially unhealthy companies and high-credit rating
firms as financially healthy companies. Previous work has also studied investors’ responses
to credit rating announcements (Castellano and D’Ecclesia 2013). Therefore, the market
may observe low-credit rating (financially unhealthy) firms as more likely to experience
the negative impacts of AI adoption than high-credit rating (financially healthy) companies.
Hence, we suggest the following hypothesis:

H3 Firms with low credit ratings experience a more negative impact on their market value
than firms with high credit ratings.

The link between AI investment and industry type is unclear and has not yet been
extensively studied. Previous studies suggest that industry type affects the influence of IT
investment on firm performance (Dos Santos et al. 1993; Kohli and Devaraj 2003; Sohal et al.
2001). Ranganathan and Brown (2006) determined that ERP platforms provide more benefits
to firms in manufacturing industries. In the context of AI adoption, investors may perceive
AI adoption as more potentially beneficial for firms in the manufacturing industry because
operational performance, such as labor productivity and quality improvement, seems to be
more potentially enhanced by AI than other companies (Bughin et al. 2018; Chui et al. 2018).
In particular, investors may perceive AI adoption to be less beneficial to firms in nonman-
ufacturing industries (Bughin et al. 2018). Nonmanufacturers with AI adoption may not be
able to cover the costs of the AI adoption, and enhancements in productivity and profitability
are difficult to observe. Therefore, we make the following hypothesis:

123
Annals of Operations Research

H4 Firms in nonmanufacturing industries experience a more negative impact on their market

value than firms in manufacturing industries.

4 Research methodology

We initially adopted the event study methodology, which is popular in the field of finance
and accounting, to estimate the stock market reaction to AI investment announcements, a
phenomenon that has become increasing common in the context of OM and IT investments
(Bose and Leung 2014; Bose and Pal 2012; Lam et al. 2019). The event study produces reliable
results even under less perfect conditions, and it allows investors to quantify the instantaneous
effect of an activity implemented by a firm instead of waiting for the effect estimation of its
accounting measures, which may only be made available after several months (Bose and Pal
2012). First, we collected news announcements related to the AI investments of firms. The
collected data were filtered and processed before the analysis. Subsequently, we estimated
the expected return by using a regression model and calculated the abnormal returns (ARs)
and cumulative abnormal returns (CARs) in the event window. The significance level of the
CARs was evaluated using different parametric and nonparametric tests. Finally, we split the
data into subsamples for further analysis. The ARs of the subsamples were computed, and
the significance of the various factors in the subsampling was analyzed.

4.1 Data collection and processing

We used Factiva to search for announcements containing pertinent keywords, such as “AI”
and “Artificial Intelligence,” from January 2015 to February 2019 and subsequently collected
the news announcements related to AI investments by the firms. Each announcement was
carefully examined to determine its relatedness to the AI investments of firms. We eliminated
announcements that were irrelevant or related to nonlisted firms. Only the earliest announce-
ment was selected when more than one announcement of the same event was found. We also
removed announcements related to stocks if they have insufficient historical data or thinly
traded stocks. The announcement of firms jointly engaged in AI initiatives was treated as
a separate event for each firm. The announcements related to confounding news, including
earnings announcements/dividends, acquisitions/mergers, top management changes, and rat-
ing changes (e.g., credit rating), within a five-day window were excluded. After a careful
processing of the data, a total of 119 valid announcements representing 62 sample firms
from 2015 to 2019 were maintained. Some of the announcements included in our sample are
shown in Table 2.
We collected stock price information from the Center for Research in Security Prices
(CSRP) and financial data from COMPUSTAT. Following previous research (Bharadwaj
2000; Lim et al. 2013; Muhanna and Stoel 2010), we determined if a firm with a high IT
capability was included in the InformationWeek 500 in the announcement year or in any other
year within our five-year study period. We adopted domestic long-term credit rating which
is produced by Standard and Poor to indicate a firm’s credit rating (Bose and Leung 2013).
The ratings equal to or above A- were classified as high credit ratings. We used SIC codes
to define whether the firms were part of the manufacturing sector. A firm was considered
to be a manufacturing firm if its SIC code is from 20 to 39; otherwise, it was regarded as a
nonmanufacturing firm.

123
 Annals of Operations Research

Table 2 Sample of announcements

Firm Date News

Citigroup Feb 1 2019 Citigroup is deploying AI to monitor corporate customers’ payments

Kraft Heinz Oct 3 2017 Kraft Heinz is working with AI across all areas to identify inefficiencies and
optimize performance
Mastercard Nov 30 2016 “Mastercard today introduced Decision Intelligence which uses AI
technology to help financial institutions increase the accuracy of real-time
approvals of genuine transactions and reduce false declines.” From Dow
Jones Institutional News
Pfizer May 3 2017 “Pfizer Australia is rolling out AI-based digital analyst tools from
Complexica in a bid to improve the pharmaceutical company’s data-driven
sales and marketing decision making.” From Complexica

Table 3 presents the description of the sample firms in our dataset. Panels A and B present
how the sample firms are distributed across years and industries, respectively. Panel C shows
the sample firms’ financial statistics in terms of the number of employees, total assets, return
on assets, cash, and long-term debt.

4.2 Abnormal stock return

We employed the capital asset pricing model (CAPM), which is the most common compu-
tation market model, to estimate abnormal stock returns. CAPM is represented by

E(Rit )  αi + βi Rmt + εit , (1)

where Rit is the rate of return for announcement i on day t, Rmt is the rate of return of market
index m on day t, α is the y-intercept and β i is the slope for announcement i, and ε it is the error
term. We considered multiple stock indices for each stock and selected the market index with
the highest R2 for the regression model. To obtain the main market reaction corresponding
to the events, the AR was estimated one day before the announcement, on the announcement
day, and one day after the announcement. We used stock price data for the 200-day estimation
period occurring one month before the AI investment announcement to build the regression
model (Eq. 1) and calculate the expected return for each sample firm.
To calculate the abnormal return (AR) in the event window, i.e., ARit , as the difference
between the actual return, i.e., Rit , and the expected return, the formula used is shown in
Eq. (2).

A Rit  Rit − (σi + bi Rmt ) (2)

The cumulative abnormal return (CAR), which is an aggregation of the ARs for the event
window, is computed using Eq. (3):

+1
C A Ri  A Rit (3)
t−1

To assess the significance level of CAR, as illustrated by Dos Santos et al. (1993), we used
the parametric Z-test and two nonparametric tests, namely, the sign test and the Corrado’s
rank test, to improve the robustness of the results.

123
Annals of Operations Research

Table 3 The description of the sample firms

Panel A: Distribution of the AI investment announcement by year

Years Number Percentage

2015 7 5.88
2016 20 16.81
2017 37 31.09
2018 36 30.25
2019 19 15.97
Total 119 100

Panel B: Distribution of AI investment announcements by industry

SIC Industry Frequency Percentage

16 Heamy Construction, Except Building Construction, Contractor 1 0.84

20 Food & Kindred Products 4 3.36
28 Chemicals & Allied Products 11 9.24
29 Petroleum Refining & Related Industries 1 0.84
30 Rubber & Miscellaneous Plastic Products 1 0.84
35 Industrial & Commercial Machinery & Computer Equipment 2 1.68
36 Electronic & Other Electrical Equipment & Components 5 4.20
37 Transportation Equipment 4 3.36
42 Motor Freight Transportation 1 0.84
48 Communications 2 1.68
52 Building Materials, Hardware, Garden Supplies & Mobile Homes 1 0.84
53 General Merchandise Stores 1 0.84
58 Eating & Drinking Places 2 1.68
59 Miscellaneous Retail 7 5.88
60 Depository Institutions 14 11.76
61 Nondepository Credit Institutions 13 10.92
62 Security & Commodity Brokers, Dealers, Exchanges & Services 10 8.40
63 Insurance Carriers 4 3.36
67 Holding & Other Investment Offices 2 1.68
73 Business Services 29 24.37
80 Health Services 1 0.84
87 Nonclassifiable Establishments 1 0.84
99 Engineering, Accounting, Research, & Management Services 2 1.68
Total 119 100

Panel C: Financial details of the firms making AI investment announcements

Variable Employee no.a Total assetsb ROA Long term debtb Cashb

Mean 134 396,139 0.14 44,906 11,429

Median 72 150,208 0.12 18,482 8079
Std. Deviation 681,973 232 0.12 16,012 64,710
Minimum 2 1694 − 0.25 0 165
Maximum 2200 2622,532 0.39 249,637 51,840
a Thousands
b USD in Millions

123
 Annals of Operations Research

Table 4 Overall impact of announcement of AI investment on stock price in varies event windows
[− 1] [0] [1] [− 1, 0] [0, 1] [− 1, 1]

Mean CAR − 1.66% − 1.77% − 1.51% − 3.44% − 3.28% − 4.94%

Z test − 0.01 − 1.51 0.65 − 1.08 − 0.60 − 0.50
Z test (p Value) 0.49 0.07 0.26 0.14 0.27 0.31
Median CAR 0.11% − 0.10% − 0.04% 0.01% − 0.12% 0.13%
No. of +ve:−ve returns 63:56 50:69 57:62 60:69 57:62 63:56
Sign test 0.70 − 1.68 − 0.40 0.15 − 0.40 0.70
Signed test (p value) 0.24 0.05 0.34 0.44 0.34 0.24
Corrado test 0.90 − 1.31 0.49 − 0.29 − 0.58 0.04
Corrado’s Rank test (p value) 0.18 0.09 0.31 0.39 0.28 0.48

5 Results

Table 4 shows the CARs of all firms in the dataset that have released announcements of
their AI activities. We estimate the CARs of different event periods, including specific event
windows ([− 1], [0], and [1]) and multiple day event windows ([− 1, 0], [0, 1], and [−
1, 1]). Here, [− 1], [0], and [1] represent the day before, the day of, and the day after the
announcement, respectively. The results show that the most significant impact appears on
day [0] that the announcement is made.
Using both parametric and nonparametric tests, on day [0], the mean (− 1.77%) and
median (− 0.10%) ARs are statistically significant and negative. This finding is reasonable
since investors observe AI investment announcements negatively and reacts too soon after
the release of the announcements. The result was not significant for the day preceding the
announcement [− 1], suggesting that no information leakage occurs before the event day.
Therefore, H1 is supported.
Table 5 demonstrates the findings of the subsample analysis on the event day [0], based on
which H2–H4 are tested. The firm subsamples with weak IT capabilities in Panel A undergo a
negative change, which is statistically significant (10%) according to both the parametric and
nonparametric tests. However, firms with strong IT capabilities yield insignificant results in
both tests. The result suggests that firms with weak IT capabilities produce a strong negative
impact when they adopt AI. Therefore, H2 is supported. The firms with low credit ratings in
Panel B have statistically significant negative mean and median CARs. Both parametric and
nonparametric tests are significant at the 10% level. However, the results are insignificant for
firms with high credit ratings. H3, which predicts that firms with low credit ratings produce
a strong negative impact, is therefore supported by the result. Finally, a significant negative
impact upon an AI investment announcement is observed among nonmanufacturing firms on
the event day [0] in Panel C. The AR results are significant (10%) and negative according to
both parametric and nonparametric tests for these nonmanufacturing firms. By contrast, the
results are insignificant for the manufacturing firms. This finding suggests that AI investments
strongly and negatively influence nonmanufacturing firms. Therefore, H4 is supported.

123
Annals of Operations Research

Table 5 Impact of AI investment on stock price across various factors on the event day [0]
Mean t test (p value) Median Mann–Whitney
test (p value)
Panel A: Impact of IT capability

Low High Difference in subgroup (Low–High)

Sample size 71 48
Mean CAR − 2.98% 0.01% − 2.99 0.09 − 0.18 0.07
Z test (p value) 0.02 0.44
Median CAR − 0.20% − 0.02%
Sign test (p value) 0.03 0.45
Corrado test (p 0.10 0.44
value)
Panel B: Impact of credit rating

Low High Difference in subgroup (Low–High)

Sample size 36 83
Mean CAR − 0.25% − 0.44% 0.19% 0.08 − 0.07 0.09
Z test (p value) 0.09 0.12
Median CAR − 0.15% − 0.08%
Sign test (p value) 0.06 0.20
Corrado test (p 0.04 0.34
value)
Panel C: Impact of type of industry

Non- Manufactory Difference in subgroups

manufactory (Non-manufactory–manufactory)

Sample size 90 29
Mean CAR − 2.38% 0.11%
Z test (p value) 0.02 0.27 − 2.49% 0.10 − 0.24 0.05
Median CAR − 0.14% 0.10%
Sign test (p value) 0.03 0.39
Corrado test (p 0.06 0.30
value)

6 Discussion

The overall purpose of this research is to evaluate the impact of AI adoption on the market
value of a firm. We suggest a hypothesis in which the announcement of AI adoption will
result in significant abnormal market returns. In addition, we hypothesize that the AI adoption
announcement will have a significant negative impact on companies with a weak IT capability,
with a low credit rating, or that are part of nonmanufacturing industries. By using the event
study methodology, we test our hypotheses by analyzing the stock market reactions to an
AI adoption announcement from 2015 to 2019. The full sample analytical results prove the
significant ARs associated with AI adoption announcements. Furthermore, our study finds

123
 Annals of Operations Research

that negative ARs are strongly associated with AI adoption for firms with a weak IT capability
and weak financial healthiness or in nonmanufacturing industries.
The full sample analysis also provides empirical evidence that AI investment announce-
ments lead to negative abnormal stock returns. The result shows that AI investment has a
negative impact (− 1.77%) on the day of the announcement. The findings of this research align
with those of previous IT-related event studies that find negative effects on the announcement
day (Bose et al. 2011; Tanriverdi and Ruefli 2004). However, the results contradict most of
the previous IT investment-related event studies that find positive returns (ARs from 0.18 to
8.2%) (Oh et al. 2006; Subramani and Walden 2001; Teo et al. 2016) or insignificant returns
(ARs from 0.02% to 0.09%) (Dos Santos et al. 1993; Im et al. 2001).
We also investigate how contextual factors influence market response towards AI invest-
ments. We found that IT capability, credit rating, and type of industry are strong moderating
factors. Firms with low IT capabilities and low credit ratings create more negative abnormal
market value (− 2.98% and − 0.25%, respectively) than their counterparts on the announce-
ment day. The results are reasonable since investors may not be confident in their low IT
capability or low credit rating, which may affect the successful implementation of the AI
technology. In addition, firms in nonmanufacturing industries have a more negative abnormal
market value (− 2.38%) than firms in manufacturing industries on the event day [0]. This
result agrees with the view of previous reports that expect the manufacturing industry to
benefit the most from AI technologies (Bughin et al. 2018; Chui et al. 2018). On the basis
of the results, this study offers a number of theoretical and managerial contributions and
implications, as discussed in the next subsections.

6.1 Contributions and implications

From a theoretical perspective, this study offers a number of contributions to theory and thus
enriches the AI, IT, and OM literature. First, this study provides empirical evidence that AI
announcement results in negative ARs. AI is believed to be extremely important to firms’
futures in different industries (Murry 2017); however, does AI indeed always have positive
impacts on firms? A number of papers and reports discuss the advantages and importance
of AI in operations, such as the works published by Makridakis (2017), Joyce (2018) and
Fragapane et al. (2020). Nonetheless, the empirical papers that support the financial benefits
are lacking. Our results demonstrate the relationship between AI investment and the firm’s
market value. To the best of our knowledge, this study is the first one to objectively measure
how AI investment influences the market value of firm adopters. In particular, this research
contributes to an increasingly important part of the literature covering the impacts of AI
investment on business value and the business model. Another theoretical contribution is the
examination of the effects of certain contextual factors, namely, the IT capacity, the credit
rating, and the type of industry on the returns of AI investment announcements. The results
illustrate the importance of considering these factors in AI investment, thus providing a
much broader picture of the OM and IT research areas. Third, this study allows researchers
to comprehensively understand the impacts of adopting disruptive innovation on market
value, thus providing theoretical guidance for future studies. Researchers can also conduct
further studies to gain more insights into the influence of other disruptive innovations and
other contextual factors on firm values.
This research holds important practical and managerial implications to practitioners and
managers. First, our study concludes that AI investment can reduce a firm’s market value.
AI adopters need to pay particular attention to the release of new investments since they are

123
Annals of Operations Research

vital for investors and the market value of firms. Managers of AI adoption need to ensure
that their senior management is aware of the potential negative market response towards
an AI investment announcement. The awareness may help obtain a strong commitment and
support from the senior management, even when short-term market losses occur. Second, on
the basis of the examined contextual factors of AI investment, the subsample analysis provides
a much deeper understanding of how an AI investment is likely to create market value for the
adopting firm. We find that the market reacts more favorably towards manufacturing firms,
but it responds more negatively to nonmanufacturing firms. Moreover, AI investments will
lead to significant negative impacts on firms with a weak IT capability and low credit rating.
To avoid suffering, nonmanufacturing firms and firms with a weak IT capability or low credit
rating should apply appropriate risk management strategies for AI adoption. These firms may
consider investing in AI only after they improve their IT capability or financial healthiness.
Third, the findings of our study are particular useful and timely for operations managers
in the case of AI initiatives that have recently blossomed. Without a doubt, managers will
make investment decisions that maximize the value of firms. However, although managers
make operational decisions by relying on accounting measures, including profitability and
productivity, these statistics are normally unavailable immediately for the firm when AI
initiative announcements are released. In fact, the data may not be available for many months.
Additionally, while it may take years of observations to estimate the direct effect of an event
on firm value, such measurements may not always reflect the actual financial effect of the
event because accounting statistics can be manipulated by managers. Although the results of
event analysis cannot fully measure the actual performance of AI implementation, they can
reflect the concerns of investors about the adoption of AI and the risks associated with the
adoption. Therefore, by using the event study methodology to measure the financial effect
of an AI initiative, this research provides managers with a timely understanding of how the
market reacts to AI initiatives, further supporting their needs to make decisions on these
disruptive innovation investments.

6.2 Limitation and future research

Similar to previous research that used the event study methodology, the current research also
has several limitations. First, the sample is limited to only U.S. publicly listed firms. Therefore,
the results may not be generalizable to small and medium firms. Future studies may consider
announcements from other countries, such as China, and compare how the market reaction
differs across countries. Moreover, firms may have adopted AI but did not announce the
initiative in the press; thus, such firms are not included in this study. Moreover, although we
have carefully filtered out the confounding events within a five-day window, we are unable to
consider all of the confounding factors in the research. For instance, we have not considered
the confounding events among companies (Dardan et al. 2005). Finally, the research only
examines the short-term impacts of AI investments. Examining the long-term effects of AI
on operational performance will be worthwhile. The use of other methodologies, including
surveys and case studies, to explore the effects of AI initiatives may also be considered.

Acknowledgements The authors are grateful for the constructive comments of the guest editor and referees
on an earlier version of this paper.

123
 Annals of Operations Research

References
Aouadni, I., & Rebai, A. (2017). Decision support system based on genetic algorithm and multi-criteria
satisfaction analysis (MUSA) method for measuring job satisfaction. Annals of Operations Research,
256(1), 3–20.
Bannerjee, G., Sarkar, U., Das, S., & Ghosh, I. (2018). Artificial intelligence in agriculture: A literature survey.
International Journal of Scientific Research in Computer Science Applications and Management Studies,
7(3), 1–6.
Basri, W. (2020). Examining the impact of artificial intelligence (AI)-assisted social media marketing on
the performance of small and medium enterprises: toward effective business management in the Saudi
Arabian context. International Journal of Computational Intelligence Systems, 13(1), 142–152.
Benotsmane, R., Kovács, G., & Dudás, L. (2019). Economic, social impacts and operation of smart factories
in industry 4.0 focusing on simulation and artificial intelligence of collaborating robots. Social Sciences,
8(5), 143.
Bharadwaj, A. S. (2000). A resource-based perspective on information technology capability and firm perfor-
mance: an empirical investigation. MIS Quarterly, 24(1), 169–196.
Bose, I., & Leung, A. C. M. (2013). The impact of adoption of identity theft countermeasures on firm value.
Decision Support Systems, 55(3), 753–763.
Bose, I., & Leung, A. C. M. (2014). Do phishing alerts impact global corporations? A firm value analysis.
Decision Support Systems, 64, 67–78.
Bose, I., Lui, A. K. H., & Ngai, E. W. T. (2011). The impact of RFID adoption on the market value of firms: An
empirical analysis. Journal of Organizational Computing and Electronic Commerce, 21(4), 268–294.
Bose, I., & Pal, R. (2012). Do green supply chain management initiatives impact stock prices of firms? Decision
Support Systems, 52(3), 624–634.
Bughin, J., Seong, J., Manyika, J., Chui, M., & Joshi, R. (2018). Notes from the AI frontier: Modeling the
impact of AI on the world economy. McKinsey Global Institute.
Castellano, R., & D’Ecclesia, R. L. (2013). CDS volatility: The key signal of credit quality. Annals of Opera-
tions Research, 205(1), 89–107.
Castellanos, S. (2018). Bank of America tech chief defines responsible AI projects. The Wall Street Journal
Retrieved July 3, 2020 from https://blogs.wsj.com/cio/2018/12/05/bank-of-america-tech-chief-defines-
responsible-ai-projects/.
Chatterjee, D., Pacini, C., & Sambamurthy, V. (2002). The shareholder-wealth and trading-volume effects of
information-technology infrastructure investments. Journal of Management Information Systems, 19(2),
7–42.
Chen, T.-L., Cheng, C.-Y., & Chou, Y.-H. (2020). Multi-objective genetic algorithm for energy-efficient hybrid
flow shop scheduling with lot streaming. Annals of Operations Research, 290, 813–836.
Chetthamrongchai, P., & Jermsittiparsert, K. (2020). The impact of artificial intelligence outcomes on the
performance of pharmacy business in Thailand. Systematic Reviews in Pharmacy, 11(1), 139–148.
Chou, J.-S., Tai, Y., & Chang, L.-J. (2010). Predicting the development cost of TFT-LCD manufacturing
equipment with artificial intelligence models. International Journal of Production Economics, 128(1),
339–350.
Chui, M., Henke, N., & Miremadi, M. (2018). Most of AI’s business uses will be in two areas. Retrieved July
3, 2020 from https://hbr.org/2018/07/most-of-ais-business-uses-will-be-in-two-areas.
Dardan, M., Stylianou, A., & Dardan, S. (2005). The valuation of eCommerce announcements during fluctu-
ating financial markets. Journal of Electronic Commerce Research, 6(4), 312–326.
Dehning, B., Richardson, V. J., & Zmud, R. W. (2003). The value relevance of announcements of transforma-
tional information technology investments. MIS Quarterly, 27(4), 637–656.
Do, N. A. D., Nielsen, I. E., Chen, G., & Nielsen, P. (2016). A simulation-based genetic algorithm approach for
reducing emissions from import container pick-up operation at container terminal. Annals of Operations
Research, 242(2), 285–301.
Dos Santos, B. L., Peffers, K., & Mauer, D. C. (1993). The impact of information technology investment
announcements on the market value of the firm. Information Systems Research, 4(1), 1–23.
Dosdoğru, A. T., Boru İpek, A., & Göçken, M. (2020). A novel hybrid artificial intelligence-based decision
support framework to predict lead time. International Journal of Logistics Research and Applications.
https://doi.org/10.1080/13675567.2020.1749249.
Faggella, D. (2019). How investors feel about artificial intelligence—From 29 AI founders and executives.
Retrieved May 14, 2020 from https://emerj.com/ai-market-research/how-investors-feel-about-artificial-
intelligence-from-29-ai-founders-and-executives/.

123
Annals of Operations Research

Fan, W., Liu, J., Zhu, S., & Pardalos, P. M. (2020). Investigating the impacting factors for the healthcare
professionals to adopt artificial intelligence-based medical diagnosis support system (AIMDSS). Annals
of Operations Research, 294, 567–592.
Fethi, M. D., & Pasiouras, F. (2010). Assessing bank efficiency and performance with operational research and
artificial intelligence techniques: A survey. European Journal of Operational Research, 204(2), 189–198.
Fragapane, G., Ivanov, D., Peron, M., Sgarbossa, F., & Strandhagen, J. O. (2020). Increasing flexibility and
productivity in Industry 4.0 production networks with autonomous mobile robots and smart intralogistics.
Annals of Operations Research. https://doi.org/10.1007/s10479-020-03526-7.
Hayes, D. C., Hunton, J. E., & Reck, J. L. (2001). Market reactions to ERP implementation announcements.
Journal of Information Systems, 15(1), 3–18.
Hunter, S. D. (2003). Information technology, organizational learning, and the market value of the firm. Journal
of Information Technology Theory and Application, 5(1), 1–28.
Im, K. S., Dow, K. E., & Grover, V. (2001). A reexamination of IT investment and the market value of the
firm—An event study methodology. Information Systems Research, 12(1), 103–117.
Jain, V. (2019). An impact of artificial intelligence on business. International Journal of Research and Ana-
lytical Reviews, 6(2), 302–308.
Jiang, F., Jiang, Y., Zhi, H., Dong, Y., Li, H., Ma, S., et al. (2017). Artificial intelligence in healthcare: Past,
present and future. Stroke and Vascular Neurology, 2(4), 230–243.
Joyce, L. (2018). Artificial intelligence and the banking industry’s $1 trillion opportunity. Retrieved Jun 4,
2020 from https://thefinancialbrand.com/72653/artificial-intelligence-trends-banking-industry/.
Kalayci, C. B., Polat, O., & Gupta, S. M. (2016). A hybrid genetic algorithm for sequence-dependent disas-
sembly line balancing problem. Annals of Operations Research, 242(2), 321–354.
Karimi, J., Somers, T. M., & Bhattacherjee, A. (2007). The role of information systems resources in ERP
capability building and business process outcomes. Journal of Management Information Systems, 24(2),
221–260.
Kobbacy, K. A., & Vadera, S. (2011). A survey of AI in operations management from 2005 to 2009. Journal
of Manufacturing Technology Management, 22(6), 706–733.
Kobbacy, K. A., Vadera, S., & Rasmy, M. H. (2007). AI and OR in management of operations: history and
trends. Journal of the Operational Research Society, 58(1), 10–28.
Kohli, R., & Devaraj, S. (2003). Measuring information technology payoff: A meta-analysis of structural
variables in firm-level empirical research. Information Systems Research, 14(2), 127–145.
Lam, H. K., Zhan, Y., Zhang, M., Wang, Y., & Lyons, A. (2019). The effect of supply chain finance initiatives
on the market value of service providers. International Journal of Production Economics, 216, 227–238.
Lawrynowicz, A. (2011). Advanced scheduling with genetic algorithms in supply networks. Journal of Man-
ufacturing Technology Management, 22(6), 748–769.
Lim, J.-H., Stratopoulos, T. C., & Wirjanto, T. S. (2013). Sustainability of a firm’s reputation for information
technology capability: The role of senior IT executives. Journal of Management Information Systems,
30(1), 57–96.
Lu, Y., & Ramamurthy, K. (2011). Understanding the link between information technology capability and
organizational agility: An empirical examination. MIS Quarterly, 35(4), 931–954.
Makridakis, S. (2017). The forthcoming artificial intelligence (AI) revolution: Its impact on society and firms.
Futures, 90, 46–60.
Marr, B. (2019). The 10 best examples of how companies use artificial intelligence in practice. Retrieved
Jun 3, 2020 from https://www.forbes.com/sites/bernardmarr/2019/12/09/the-10-best-examples-of-how-
companies-use-artificial-intelligence-in-practice/#5591731a7978.
McNaught, K., & Chan, A. (2011). Bayesian networks in manufacturing. Journal of Manufacturing Technology
Management, 22(6), 734–747.
Minevich, M. (2020). 4 Ways that you can prove ROI from AI. Retrieved May 15, 2020 from https://www.
forbes.com/sites/markminevich/2020/03/03/4-ways-that-you-can-prove-roi-from-ai/#208b5fa784a7.
Mueller, J. P., & Massaron, L. (2018). Artificial intelligence for dummies. Hoboken: Wiley.
Muhanna, W. A., & Stoel, M. D. (2010). How do investors value IT? An empirical investigation of the value
relevance of IT capability and IT spending across industries. Journal of Information Systems, 24(1),
43–66.
Munguia, J., Lloveras, J., Llorens, S., & Laoui, T. (2010). Development of an AI-based rapid manufacturing
advice system. International Journal of Production Research, 48(8), 2261–2278.
Murry, A. (2017). Fortune 500 CEOs see A.I. As a big challenge. Retrieved May 18, 2020 from https://fortune.
com/2017/06/08/fortune-500-ceos-survey-ai/.
Oh, W., Kim, J. W., & Richardson, V. J. (2006). The moderating effect of context on the market reaction to IT
investments. Journal of Information Systems, 20(1), 19–44.

123
 Annals of Operations Research

Ranganathan, C., & Brown, C. V. (2006). ERP investments and the market value of firms: Toward an under-
standing of influential ERP project variables. Information Systems Research, 17(2), 145–161.
Ross, J. W., Beath, C. M., & Goodhue, D. L. (1996). Develop long-term competitiveness through IT assets.
Sloan Management Review, 38(1), 31–42.
Sohal, A. S., Moss, S., & Ng, L. (2001). Comparing IT success in manufacturing and service industries.
International Journal of Operations & Production Management, 21(1/2), 30–45.
Statista. (2016). Enterprise artificial intelligence market revenue worldwide 2016–2025. Retrieved May 18,
2020 from https://www.statista.com/statistics/607612/worldwide-artificial-intelligence-for-enterprise-
applications/.
Staw, B. M., Sandelands, L. E., & Dutton, J. E. (1981). Threat rigidity effects in organizational behavior: A
multilevel analysis. Administrative Science Quarterly, 26(4), 501–524.
Stoel, M. D., & Muhanna, W. A. (2009). IT capabilities and firm performance: A contingency analysis of the
role of industry and IT capability type. Information & Management, 46(3), 181–189.
Subramani, M., & Walden, E. (2001). The Impact of e-Commerce announcements on the market value of
firms. Information Systems Research, 12(2), 135–154.
Tanriverdi, H., & Ruefli, T. W. (2004). The role of information technology in risk/return relations of firms.
Journal of the Association for Information Systems, 5(11–12), 421–447.
Teo, T. S., Nishant, R., & Koh, P. B. (2016). Do shareholders favor business analytics announcements? The
Journal of Strategic Information Systems, 25(4), 259–276.
White, L. J. (2010). Markets: The credit rating agencies. Journal of Economic Perspectives, 24(2), 211–226.
Yang, L., Chen, G., Rytter, N. G. M., Zhao, J., & Yang, D. (2019). A genetic algorithm-based grey-box model for
ship fuel consumption prediction towards sustainable shipping. Annals of Operations Research. https://
doi.org/10.1007/s10479-019-03183-5.

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.

123