Adoption of AI in response to COVID-19—a configurational perspective

Abstract

Although the importance of artificial intelligence (AI) has often been highlighted in strategic agility and decision outcomes, whether it helps firms strengthen their competitiveness and the means firms use to achieve such competitiveness are still under-researched. Our research thus joins the recent discussion on digitalization trends and strategic responses to COVID-19 to better understand how firms strengthen their competitiveness during such challenging times. Namely, this study incorporates the strategic responses to COVID-19 into the technology–organization–environment (TOE) framework by investigating the impacts of different configurations of TOE contexts and strategic responses on a firm’s competitive advantage. We used fuzzy-set qualitative comparative analysis to investigate how TOE contexts and strategic responses integrate into configurations and impact a firm’s competiveness. By applying a configurational approach with data from 514 exporting firms in China, we find a strong indication of the equifinality of different strategies, indicating that multiple strategic paths can be used to respond to crises. The adoption of AI, while important, is not sufficient to enhance a firm’s competitiveness. Our results stress the significance of data quality, organizational resources and capabilities, and digital business model innovation for AI adoption. We also identify successful strategic paths of AI adoption aversion and ambidextrous strategies. The findings have practical implications for firms seeking effective strategies to respond to future crises and sustain their competitive advantages.

Introduction

Analyzing the relationship between the adoption of artificial intelligence (AI) and firm competitiveness is paramount for business researchers and practitioners, especially in the context of the COVID-19 pandemic. UNCTAD [1] revealed a surge in online shopping due to COVID-19, which is likely to have lasting effects as the world economy recovers. Although surge in digitization has boosted the availability of global open resources, it has raised the bar and may create new divides. There are significant discrepancies between firms in terms of value creation and value capture in the digital economy [2, 3]. How firms transform data and digital connectivity into sustained competitive advantage is also unclear [4]. Therefore, how businesses seize the opportunities in the age of big data and digitalization remains an unresolved question. Sustained organizational performance is rooted in competitive advantage. However, attaining competitive advantage is challenging for any firm [5]. Given this background, one key question is how to sustain firms’ competitive advantages while responding to COVID-19 effectively in the age of big data.

Although the importance of AI adoption has often been highlighted in strategic agility and decision outcomes [6], whether it helps firms strengthen their competitiveness and the means through which firms achieve such competitiveness are still under-researched. Our study builds on the technology–organization–environment (TOE) framework [4], coupled with the strategic responses to COVID-19, to shed light on the combined influences of technological, organizational, and environmental factors along with strategic choices on firms’ competitive advantages. These factors are interrelated and act as a complete entity rather than existing as isolated parts. The configurational approach enables systemic and holistic views of AI adoption and its context [7]. A configuration is a specific set of causal variables with a synergetic nature allowing investigation of interrelated structures instead of entities that are examined in isolation [8]. Thus, we propose a configurational approach to investigate AI adoption and the contextual conditions that impact a firm’s competitive advantage.

We contribute to this emergent dialogue in two ways, as follows. First, there are few systematic investigations of how technology integrates with organizational, environmental, and strategic factors and has a combined impact on firms’ competitive advantages. Although studies on how AI need to be supported by decision-makers’ competencies are more frequent, they tend to focus on the technical side and the capabilities related to utilizing technology [9]. Insufficient attention has thus been paid to the configurational effects of organizational strategies and the external environment, thereby ignoring the complex nature and underestimating the interconnectivity of the factors that influence firm competitiveness. By addressing these gaps, this study provides new theoretical and empirical insights that link AI adoption to a firm’s competitive advantage through internal and external factors.

Second, we consider the configurational effects of the factors on firms’ competitive advantage. In international business, the configuration of the explanatory variables simultaneously determines outcomes in many cases. However, several empirical studies assume that the relationships between focal variables are linear, unifinal, and symmetrical, which implies a mismatch between the nature of the phenomena and the empirical methods employed [10]. Owing to the complexity of the strategic responses to COVID-19 and the multifaceted nature of AI adoption, Witt et al. [11] suggested that future studies should consider configurational research methodologies. Therefore, this study employs fuzzy-set qualitative comparative analysis (fsQCA), one of the emerging yet underused methods in the international business field [10], to analyze the complex patterns of causality among the adoption of AI, decision styles, business strategy, external environment, and a firm’s competitive advantage.

Drawing from the TOE framework, this study finds that AI adoption is insufficient for firms’ competitive advantage. High-quality data, technological competencies, and business model innovations pertaining to digital transformation are also required. Further, multiple success formulas exist, meaning it is possible to sustain a competitive advantage without adopting AI. Through this rich explanation and empirical assessment, we contribute to clarifying how AI adoption impacts firms’ competitive advantage during the COVID-19 pandemic.

Theoretical framework

The TOE theory explains the adoption and implementation of information systems in an organization [12]. This framework considers the context of technological adoption and implementation under three dimensions: technological, organizational, and environmental [12]. The technological context denotes the features of technological innovation, the organizational context represents organizational resources and capabilities, and the environmental context specifies the environment in which technological adoption operates [13]. We also include the strategic responses to COVID-19 in our framework, as firms tend to respond to exogenous shocks in different ways, which are critical to the development of competitive advantage [14, 15]. Extant research has found that the assimilation of an information system in an organization, including the adoption, implementation, and post-implementation of the system, is affected by the factors related to technology, organization, and environment [13, 16, 17]. However, these studies examined the TOE dimensions as individual contributing factors rather than configurations, missing the important mechanism by which they are successfully integrated. To fill this gap, we seek to investigate how technological applications are integrated into the organizational context and the external environment together with business strategies in response to COVID-19. Fig. 1 presents the theoretical framework.

Fig. 1.

Theoretical framework

Technological context

The technological context mainly includes the characteristics of an enterprise system, such as the quality of the system [12]. Here, we look at the adoption of AI for which data quality is a critical input factor. AI is an intelligent system that can be used to interpret external data, learn from the data, and learn to achieve specific goals and tasks [6]. The adoption of AI is of great value, especially for non-routine decisions [18], such as strategic decisions on the responses to COVID-19, a problem far from being well-structured. It thus helps reduce the degree of uncertainty in strategic decision-making by accomplishing quantitative tasks [19] and accumulating more knowledge from data [6]. In an environment with greater complexities and dynamism, AI can help anticipate the changes in the organizational environment and offers a more proactive approach to problem-solving [20, 21].

The ultimate values and results of data analytics are significantly affected by the quality of the data (e.g., timely and relevant) used [22, 23]. Data quality is defined as the extent to which data reflect the facts and characteristics of an entity or event [24]. High-quality data are conceptualized as intrinsic, contextual, representational, and accessible [25]. Intrinsic data refer to data correctness and accuracy. High-quality contextual data represent the completeness, timeliness, and relevance of the data. Representational data indicate the clarity and consistency of the data, while accessibility refers to the ease of obtaining the data. High-quality data are considered an intangible IT-enabled resource on which the ultimate results and business insights depend [26]. Data quality has also been reported as the main obstacle to data analytics competency [27].

Organizational context

The organizational context relates to overt characteristics such as organization size and structure as well as organizational readiness such as organizational resources and capabilities and human components [16]. Despite the significant role that high-quality data plays in strategic decision-making, firm resources and capabilities are also critical for successfully employing this tool [28]. The failure of data analytics in decision-making could be due to the focus on the data aspect while neglecting the challenges from the human components [29]. Successful decision-making within firms can be hampered when domain knowledge and analytical capabilities are lacking [30]. Domain knowledge and analytical capabilities are a combination of resources required for effective data analysis [31]. Individuals with sufficient domain knowledge have a deep understanding of the practices and procedures in the industry and are thus more capable of identifying the key attributes from data insights [32] and solving problems more effectively [23]. Similarly, analytical capabilities facilitate data analysis and interpretation, making the processes error-free and effective [32].

Additionally, the human idiosyncrasies in decision-making styles can impact the decision-making process. Data analytics use historical data; thus, they can amplify existing biases [33]. Disregarding human reasoning has long been a criticism of AI adoption [34]. More recently, critiques have challenged the applicability of AI in strategic decision-making, especially the conceptual work on a firm’s strategy [21] and ignoring morals, values, and ethical standards in algorithm-driven decision-making [35]. Therefore, the interplay between humans and AI is crucial for maximizing the benefits of AI and big data analytics. Dual process theory suggests that human information processing is accomplished by two dissimilar but complementary styles: intuition and rationality [36]. It is believed that intuition and rationality can be operated in parallel, and the relative contribution of either style depends on the decision-maker and the situation [37]. Intuition is considered as “affectively charged judgments that arise through rapid, non-conscious, and holistic associations” [38] (p. 40). Under most circumstances, intuition and rationality are synchronized through seamless, harmonious, and synergistic operations [39]. Such synchronized integration helps generate holistic components in information processing and is thus more adequate for more complex situations [37].

Role of business strategies

A strategy emerges from responding to changing situations [40] and formulates well-structured plans with the purpose of value creation and capture, thus strengthening competitive advantage for businesses [41]. The adoption of AI will not be achieved if business strategies are not aligned, because the adoption is related to the level of IT investment, efficiency, and innovation [42]. Two opposing strategies have been identified in response to COVID-19: retrenchment and innovation [14, 15]. Retrenchment is the most commonly used strategy to reduce costs and the scope of business activities. It thus helps firms focus on the core of their business [43]. Innovation refers to strategic transformation. To remain competitive in the market, managers constantly reflect on current customer needs and explore opportunities and viabilities under different business models [44]. For example, firms are undergoing business model and digital transformations during the pandemic. As such, it is believed that actively seeking new ways of doing business can be an effective strategic response to crises [45]. Empowered by AI and big data analytics, and given the upsurge in online shopping, a digital business model can generate revenue from more personalized products and services [46]. In light of the configurational approach, firms can use equally valuable hybrid strategies, where synergies can be created from digital business model innovation and retrenchment because digital technologies can function as a key driver of scaling the advantages that feed into the basis of low-cost strategies [47]. Zott and Amit [48] also suggested the potential of innovation and retrenchment strategies to act as complements rather than substitutes.

Environmental context

Firm strategies are not implemented or executed in vacuum [47]. Instead, the external environment, especially at the industry level, plays a significant role in developing a firm’s competitive advantage [49, 50]. Competitive intensity and dynamism are industry-level factors [13, 51]. A competitive environment may reduce the value of the existing business model and force the firm to develop new business models for higher value creation or lower the cost of higher value capture [47, 52]. However, when competitive intensity is high, retrenchment alone may be insufficient to sustain a firm’s competitive advantage.

Environmental dynamism is characterized by technological changes, variations in customer preferences, changes in product demand, supply of materials, and the unpredictability of change [53]. Firms must innovate to avoid the risk of obsolesce [13]. Simultaneously, firms can develop exploitative strategies to cope with the threat and capitalize on previous innovative efforts [54]. Therefore, dynamic business environments may push firms to engage in both exploratory innovation and exploitative retrenchment simultaneously [55]. Consistent with the TOE framework, the benefits of AI in successfully formulating and implementing innovation and/or retrenchment strategies depend not only on the development of internal capabilities (i.e., domain knowledge and analytical capabilities), but also on the external environment.

Research methods

Research design

This study used a survey methodology to collect primary data on how Chinese exporting firms responded to COVID-19. We surveyed 1561 firms in 2021 and received 533 responses. After deleting 19 disengaged responses, the 514 usable responses were used for analysis. Of the firms, 45.3% were privately owned, 44.4% were listed, and 10.3% were state-owned. The positions that our participants held were directors (4.9%), executive management (40.5%), senior management (51.9%), and supervisory to middle-level management (2.7%). We conducted non-response bias tests by equally dividing the respondents into early and later waves based on survey completion date and time. We compared the two waves on a range of firm characteristics, including the number of employees, firm age, industry, and individual characteristics such as qualifications. No systematic difference was found between responding and non-responding firms, indicating that non-responding bias was not a major issue.

Measures

The measures used in this study are well established in existing studies. Competitive advantage was measured using a four-item seven-point Likert scale adopted from Clauss et al. [56]. AI-based information system adoption was measured using a six-item seven-point Likert scale based on Park et al. [57]. Data quality, analytical skills, and domain knowledge were adopted from Ghasemaghaei et al. [32]. Intuition was measured on a five-item scale from existing research [58–60]. Rationality was measured using a four-item scale [61]. Environmental competitiveness was measured by the competitive intensity in an industry [53, 56, 62]. Environment dynamism was measured by the degree of change in an industry [53, 63]. Digital business model innovation was adapted from existing measures and measured on an eight-item scale [64–66]. Retrenchment was developed according to Wenzel et al. [14] and was measured by a binary variable, where “0” represented hiring new staff during COVID-19 and “1” represented layoffs. The detailed survey questions are presented in the Appendix.

We also validated the results using confirmatory factor analysis (CFA) [67]. The results suggest that both convergent and divergent validity were achieved. The measurement model results showed an acceptable fit (CMIN/DF=1.775, CFI=0.953, IFI=0.953, TLI=0.947, RMSEA=0.039). As per Table 1, composite reliability and average variance extracted (AVE) reach the common cutoff point of 0.6 and 0.5 for all the constructs, ascertaining convergent validity. The reliability of the measurement items was assessed using Cronbach’s alpha, and the results of all constructs were higher than the recommended level of 0.70 [68], indicating satisfactory reliability. The squared root of the AVEs of the constructs was greater than the correlations of each construct, demonstrating strong discriminant validity [69]. In summary, the measurement model fits the data reasonably well, and the constructs demonstrated measurement properties sufficient for further analysis.

Table 1.

Correlation matrix and assessment of measurement validity

*p<0.05, **p<0.005, and ***p<0.001; the diagonal elements in bold are the square roots of AVE

FsQCA

FsQCA is a case-based technique that focuses on configurational effects. It follows a set-theoretic approach using Boolean algebra to identify causal relationships between combinations of attributes and outcomes [70]. It is driven by complexity theory, which allows testing causality with three features: (1) conjunction, meaning the outcomes are driven by the interdependence of multiple conditions rather than a single cause; (2) equifinality, indicating more than one pathway to a certain outcome; and (3) asymmetry, implying that one particular attribute in a configuration could be causally related to a good outcome but could be unrelated or inversely related to a good outcome in another configuration [71]. Conventional correlation-based approaches are not designed to address conjunctural, equifinal, or asymmetrical causal relations [72]. Given the theoretical novelty of examining AI-based information systems and its complex relationships with firm capabilities, decision styles, strategy, and the external environment, fsQCA can reveal synergistic effects and causal complexity by focusing on the effects of combinations of attributes rather than net effects.

For fuzzy sets, we must calibrate the original data into scores between 0 and 1, indicating their degree of membership by values that indicate three key breakpoints: (1) full membership, (2) full non-membership, and (3) crossover point. We adopted the direct approach to calibrate outcomes and causal conditions [70, 73]. To complete the transformation of our variables into a fuzzy set, we followed the anchor points suggested by Pappas and Woodside [7] and Fiss [70], using the 95th percentile as the anchor point for the full membership, the 5th percentile for the non-membership, and the 50th percentile for the crossover points. Table 2 presents the results.

Table 2.

Calibration of variables

Variable	Threshold
	Full membership	Crossover	Full non-membership
Competitive advantage	9.571	8.000	5.714
Adoption of AI	6.667	5.500	3.500
Data quality	6.667	5.500	3.958
Domain knowledge	7.000	6.000	4.500
Analytic capability	7.000	6.000	4.333
Rationality	6.600	5.200	2.800
Intuition	6.750	5.750	4.500
Environmental dynamism	6.750	5.500	4.250
Environment competition	6.667	5.500	4.292
Digital business model innovation	6.667	5.778	4.222
Retrenchment	0.950	0.501	0.050

Following the calibration of variables, we conducted necessity analyses for all attributes and their negation, applying a recommended consistency benchmark of 0.9 [74] and taking coverage as a measure of the relevance of a necessary condition. As shown in Table 3, the individual conditions reveal that all conditions have less than 0.9 consistency for competitive advantage and negation. This indicates that no individual factors contributed to the outcomes. It is likely that a combination of several factors has an influential impact. We then conducted sufficiency analyses using Ragin’s [72] truth table algorithm to identify attribute combinations consistently linked to an outcome. We set the frequency cutoff as three cases per configuration, as suggested for samples over 150 [70], and applied a consistency benchmark of ≥0.8 [72] complemented by a proportional reduction in inconsistency score of ≥0.7 [75, 76] to avoid simultaneous subset relationships of attribute combinations of both the outcome and its absence. All analyses were conducted using fsQCA 3.0.

Table 3.

Analysis of necessary conditions

Variable	Competitive advantage		Negation of competitive advantage
	Consistency	Coverage	Consistency	Coverage
Adoption of AI	0.795	0.772	0.583	0.552
Data quality	0.763	0.758	0.432	0.419
Domain knowledge	0.792	0.807	0.539	0.536
Analytic capability	0.776	0.799	0.539	0.542
Rationality	0.726	0.701	0.620	0.584
Intuition	0.727	0.692	0.465	0.431
Environmental dynamism	0.790	0.763	0.577	0.543
Environment competition	0.750	0.763	0.549	0.545
Digital business model innovation	0.767	0.783	0.395	0.393
Retrenchment	0.609	0.526	0.659	0.556

Results

Table 4 presents the results. Ragin’s [72] truth table analysis procedure produced different solutions. FsQCA generates three different types of solutions: a complex solution that only uses configurations with existing data, an intermediate solution that further uses “easy” counterfactuals [72], and a parsimonious one that provides a simpler solution. By leveraging these different forms of evaluation, we can compare the intermediate and parsimonious solutions to identify “core” and “peripheral” elements [70] based on the strength of the evidence. Following current best practices, we report a combination of intermediate and parsimonious solutions (e.g., [70]). Core conditions are those contained in the intermediate and parsimonious solutions, whereas the peripheral conditions are those that are contained in the intermediate solutions but not the parsimonious ones. In Table 4, the black circles (●) indicate the presence of a causal condition, whereas the crossed circles (⊗) indicate its absence [77]. Large circles indicate core conditions, whereas small circles indicate peripheral conditions. The table also reports consistency and coverage measures [72]. We obtained a coverage of 0.516 for a greater competitive advantage, indicating the empirical importance of the solutions as a whole. The solution consistency was 0.917, significantly higher than the required consistency of 0.80 [70], which means the overall solutions corresponded well with the data.

Table 4.

Configurational solutions for a high competitive advantage

The black circles (●) indicate the presence of a condition and the crossed circles; ( ) indicates its absence. Large circles indicate core conditions, and small ones indicate peripheral conditions. Blank spaces mean “do not know”

Table 4 reveals four first-order configurational solutions that lead to a stronger competitive advantage. These solutions show the presence or absence of core and peripheral conditions, referring to the equifinality of solutions across types [70]. In solution 1 (1a, 1b, 1c, 1d) and solution 2 (2a, 2b), second-order solutions were found to be present, demonstrating the existence of within-type equifinality.

Solution 1 indicates the presence of high-quality data, digital business model innovation within a dynamic and competitive environment combined with AI adoption and competencies (solution 1a, solution 1b), or the use of rationality and intuition (solutions 1c, 1d). Data quality, environmental competition, and digital business model innovation are core conditions, while AI adoption, competencies, decision styles, and environment dynamism are peripheral ones. Solutions 2a and 2b report a hybrid strategy, as both digital business model innovation and retrenchment are core conditions. Rational analysis, rather than intuition, is a peripheral condition, in addition to data quality and competencies. The adoption of AI may be present as a peripheral condition or may not matter in solution 2. Solution 3 represents another configuration associated with a hybrid strategy and the use of rationality in decision-making. Although it has high data quality, AI is not adopted in this solution. Solution 4 indicates the presence of digital business model innovation and the absence of a dynamic environment as core conditions, whereas AI adoption, data quality, competencies, rationality, and the absence of a competitive environment and retrenchment strategy are peripheral conditions.

It is worth noting that high data quality and digital business model innovation are necessary and sufficient conditions for a stronger competitive advantage. However, all other conditions are necessary but not sufficient, indicating the conjunctural nature and asymmetric impact that the adoption of AI and the organizational and environmental contexts have on firm competitiveness.

Discussion

We employed a configurational approach to the strategic responses to COVID-19 to understand how the adoption of AI interacts with other contributors—technical, organizational, environmental context, and strategic responses. We highlight how AI adoption and related data may be critical to adopting a digital business model and sustaining a firm’s competitiveness. Consistent with the TOE framework in that technological adoption must be used in conjunction with organizational capabilities and forces from the external environment, we can understand that AI adoption alone falls short of being a sufficient contributor to sustain competitive advantage. Simultaneously, we demonstrated how AI adoption might be successfully combined with other necessary items, such as high digital competencies and a digital business model.

Role of AI adoption in sustaining competitive advantage

Consistent with the TOE framework, AI adoption is related to technological, organizational, and environmental factors. However, we found several successful paths for AI adoption to sustain a firm’s competitive advantage. To be specific, the findings revealed configurations with a stronger competitive advantage, in which the adoption of AI was either present (solutions 1a, 1b, 1d, 2a, and 4), absent (solution 3), or did not matter (solutions 1c and 2b). In configurations where the AI adoption was a necessary condition, we found no evidence that it was itself sufficient for a stronger competitive advantage. As demonstrated by solutions 1a, 1b, 2a, and 4, firms must combine it with other contributors and strategies to strengthen competitive advantage. In particular, high-quality data ensures the accuracy of the input side of AI adoption, whereas domain knowledge and analytical capability ensure the analysis and interpretation of data, that is, the output side of AI adoption. Proactively integrating AI adoption is thus beneficial, even in industries that are currently more stable and less competitive (as shown by solutions 2a and 4).

Additionally, as shown by solution 1d, the use of both intuition and rationality seems to substitute the shortages of domain knowledge and analytical capability. Human factors in decision-making become especially crucial when domain knowledge and analytical capability are not strong enough to interpret the data produced by AI. This provides strong support for the interaction between AI and human reasoning when formulating strategies [78]. This finding is compatible with the dual process theory, which states that the combined use of intuition and logic can integrate holistic components into information processing, thus making it more effective when dealing with complex situations [36, 37].

By contrast, solution 3 illustrates an AI-averse solution, indicating that the adoption of AI and traditional rational analysis can complement each other. That is, digital business model innovation does not universally rely on AI or the use of big data. Traditional data such as financial and customer data stored on corporate dashboards could also contribute to strengthening a firm’s competitive advantage if effectively used.

An ambidextrous strategy for responding to crises

In addition to the TOE framework, we also integrated firms’ strategic responses to COVID-19 as the firm-level strategy. We included retrenchment and digital business model innovation as two strategies needed to be aligned with the technological adoption. Our findings show that an ambidextrous strategy that focuses on both innovation and efficiency (i.e., solutions 2a, 2b, and 3) is equally applicable to a business’s sustained competitive advantage. Managers should not fall into the innovation trap, but should instead consider how to restructure the company, for example, through retrenchment, while adopting an innovation mindset to transform the business model to better suit the digitalization shift. It is worth noting that data quality and digital business model innovation are sufficient and necessary conditions for a stronger competitive advantage both during and after COVID-19. This reaffirms that consumer behavior changes, and this digitalization trend is unlikely to reverse post-COVID. However, this does not imply that all of a company’s efforts should be devoted to innovation. Our findings echo those of Leppänen et al. [47] in that business model innovation is considered a driver for value creation (“growing the pie”), while efficiency contributes to value capture (“getting a large slice”).

In addition, the absence of intuition and the presence of rational analysis were evident in firms that prioritized an ambidextrous strategy (solutions 2a, 2b, and 3). This indicates that the use of rational analysis is more appropriate. This may be caused by balancing the degree of retrenchment to contribute to business model innovation, rather than inhibiting it. Rational analysis is essential for analyzing the synergies between these two diametrically opposed strategies. The solutions indicate that high-quality data and strong analytical capability are the necessary conditions for the successful implementation of a rational thinking style and ambidextrous strategy. However, businesses can focus on AI adoption (solution 2a) or AI aversion (solution 3), as long as synergies develop between the innovation and retrenchment strategies, and the use of rational thinking contributes to this synergistic condition.

Theoretical contributions

Our study contributes to the recent discussion on the digitalization trend during COVID-19 and the strategic responses pertaining to it. This study contributes new insights to the literature on information system, strategies, and decision-making processes in the context of China. The TOE framework emphasizes the importance of integrating technological applications into the organizational context and the external environment. However, the mechanism by which they are successfully integrated remains unclear. Our study addresses this gap by emphasizing configurational thinking for AI adoption and sustained competitive advantage. We find a strong indication of strategy equifinality, indicating that multiple configurations help strengthen a firm’s competitive advantage. We stress the significance of organizational resources and capabilities critical to the input and output of AI adoption, without which adoption may be unsuccessful.

Our findings also reveal that AI adoption is not the only success formula, as firms employing a relevant traditional approach to data analysis and digital business model transformation can be equally successful. We extend the dual process theory by investigating human interaction with AI adoption, whereas the discussion on combining intuition and rationality is mostly theoretical [36]. Echoing Leppänen et al. [47], equifinality also applies to innovation-focused and ambidextrous strategies. We confirm that organizational ambidexterity can be applied during uncertain times, when firms face fierce competition and dynamic changes.

Practical contributions

The configurational view of AI adoption and the strategic responses to crises offer three managerial implications. First, in the age of big data, acquiring high-quality data, and having a team equipped with domain knowledge and data analysis skills are essential assets for a firm’s competitiveness at the global scale. This also brings attention to actively involving and effectively communicating with a team of experts including both experienced managers and data scientists. Second, by understanding the equifinality of organizational contexts and decision-making styles, we provide a practical lens that decision-makers can use to intentionally adjust their decision styles based on a holistic assessment of organizational context and strategies [79]. Finally, our findings reveal the importance of digital business model innovation and an ambidextrous strategy for firms to sustain their competitive advantage after COVID-19. Simultaneously, implementing a retrenchment approach is not the best course of action. Businesses must think more strategically and proactively about restructuring their business models and updating their resources and capabilities to better adapt to digitalization.

Limitations

The limitations of this study provide scope for further research. First, although this study was carefully designed and integrated within a holistic framework, the results are constrained by the solutions revealed in our survey sample from China. More configurations of AI adoption and factors both internal and external to a firm may be identified in other countries. Second, it would be interesting to conduct a qualitative study to better understand the mechanism of the strategic responses before and during the COVID-19 pandemic. Finally, the findings of this study are based on cross-sectional survey data. A next step could be investigating the configurations over a period and the lasting effects of the pandemic. A longitudinal research design can provide more support for the robustness of the causal relationships.

Appendix. Survey questions

Survey questions	Variables
Please assess your relative strength in comparison to your competitors regarding the following:	Competitive advantage
1. Innovative offerings 2. Quality 3. Technological capacity 4. Reputation 5. Service 6. Financial success 7. Market share
How often is AI-based global information system used in your organization for the following?	Adoption of AI
1. Visually present business processes 2. Support the design and creation of new business processes 3. Support streamlining and scheduling processes 4. Automate business processes 5. Provide information about what human and other resources are needed for business processes 6. Provide real-time information about resource availability
To what degree you agree with the following statements?	Data quality
In my organization, data used in data analytics: 1. Is reliable 2. Has an appropriate level of details 3. Is secure 4. Is timely 5. Is relevant to the task at hand 6. Is accurate
To what degree do you agree with the following statements?	Domain knowledge
In my organization, there is a high level of knowledge of the 1. External environment (e.g., government, competitors, suppliers, and customers) 2. Organizational goals and objectives 3. Core capabilities of the organization 4. Key factors for the organization to succeed
To what degree you agree with the following statements?	Analytic capability
In my organization … 1. Our data analytics users are knowledgeable when it comes to utilizing such tools 2. Our data analytics users possess a high degree of data analytics expertise 3. Our data analytics users are skilled at using data analytics tools
To what degree do you agree with the following statements?	Rationality
1. To what extent did the decision-makers gather relevant information for making this decision? 2. How extensively and thoroughly did the decision-makers analyze relevant information for making this decision? 3. How successful were the decision-makers at focusing their attention on crucial information when making this decision? 4. To what extent did the decision-makers use analytic techniques to analyze the data?
To what degree do you agree with the following statements?	Intuition
1. To what extent do senior managers in your company rely on pure judgment when making important decisions? 2. In your company, how much emphasis do senior managers place on experience when making important decisions? 3. On many occasions, senior managers do not have enough information, and must make important decisions based on a “gut-feeling” 4. Senior managers did not have time to decide analytically, so they relied on their experience and expertise 5. Senior managers made a connection between the situation at hand and similar situations in the past, and decided accordingly
To what degree do you agree with the following statements regarding your industry as a whole?	Environmental dynamism
1. Environmental changes in our industry are intense 2. Our clients regularly ask for new products and services in our industry, changes are taking place continuously 3. In a year, many aspects have changed in our industry 4. In our industry, the volumes of products and services to be delivered change fast and often
To what degree you do agree with the following statements regarding your industry?	Environment competition
1. Competition in our local market is intense 2. Price competition is a hallmark of our local market 3. Anything that we can offer, our competitors can match easily 4. Our clients usually purchase from multiple suppliers and there are many promotion wars
To what extent do you agree with your business model regarding digital technology adoption?	Digital business model innovation
1. Our digital business model offers new combinations of processes, products, services, and information 2. Our digital business model attracts a lot of new customers 3. Our digital business model attracts a lot of new suppliers and other business partners 4. Our digital business model brings together internal and external participants in novel ways 5. Our digital business model is revolutionizing the way business deals are made 6. We frequently introduce new ideas and innovations in our business model in the context of digital technology adoption 7. We frequently introduce new processes, routines, and norms in our business model in the context of digital technology adoption 8. In the context of digital technology adoption, we are pioneers with our business model 9. Overall, and in the context of digital technology adoption, our business model is novel

Authors’ contributions

All authors contributed to the study conception and design. Material preparation, data collection, and analysis were performed by Lili Mi, Wei Liu, and Yu-His Yuan. The first draft of the manuscript was written by Lili Mi, and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Funding

This study received financial support from Griffith Business School, Griffith University in 2021.

Data Availability

The full data is available to the immediate research only as per the ethics approval protocol on confidentiality, storage and sharing of the data.