Strategizing Artificial Intelligence Transformation in Smart Ports: Lessons from Busan’s Resilient AI Governance Model

Abstract

The global port and maritime industry is experiencing a new paradigm shift known as the artificial intelligence transformation (AX). Thus, domestic container-terminal companies should focus beyond mere automation to a paradigm shift in AI that encompasses operational strategy, organizational structure, system, and human resource management. This study proposes a resilience-based AX strategy and implementation system that allows domestic container-terminal companies to proactively respond to the upcoming changes in the global supply chain, thus securing sustainable competitiveness. In particular, we aim to design an AI-based governance model to establish a trust-based logistics supply chain (trust value chain). As a research method, the core risk factors of AX processes were scientifically identified via text-mining and fault-tree analysis, and a step-by-step execution strategy was established by applying a backcasting technique based on scenario planning. Additionally, by integrating social control theory with new governance theory, we designed a flexible, adaptable, and resilience-oriented AI governance system. The results of this study suggest that the AI paradigm shift should be promoted by enhancing the risk resilience, trust, and recovery of organizations. By suggesting AX strategies and policy as well as institutional improvement directions that embed resilience to secure the sustainable competitiveness of AI-based smart ports in Korea, this study serves as a basis for establishing strategies for the domestic container-terminal industry and for constructing a global leading model.

1. Introduction

1.1. Background

In recent years, the global shipping and logistics industry has experienced a rapid paradigm shift owing to the realignment of maritime alliances, supply-chain restructuring, port automation, and the introduction of artificial intelligence (AI)-based logistics optimization technologies. In particular, the General Data-Protection Regulation (GDPR), Data Act, and AI Act of the European Union (EU) promote the transition to data-driven AI systems [1] and highlight the strategic importance of data management in port and logistics systems [2]. Despite these changes, the digital transformation of the port industry has been slow compared with that of other industries. Only a few smart ports have been implemented on a full scale, which represents the digital transformation (DX) of the port industry, and disparities in the digitization level between countries and ports remain [3]. This gap can result in a global competitive disadvantage in the future. To counteract this, digital connectivity and technology standardization across the shipping and port industry are essential, particularly for a full-cycle AI transformation (AX) that extends beyond simple automation. These international trends and technological demands directly affect the Korean shipping and port industry. Busan New Port’s Pier 6 began automated operations in 2022 by introducing an automated quay crane. In 2024, Dongwon Global Terminal began its AI-based paradigm shift by adopting an automated guided vehicle (AGV)-based system. For domestic container-terminal companies to maintain their competitiveness as global shipping and logistics centers, they must construct smart ports based on AI technology, expand eco-friendly logistics infrastructure, and strengthen strategic cooperation with the Global Maritime Alliance. Additionally, the introduction of eco-friendly logistics solutions to reduce carbon emissions, enhancing operational efficiency through supply-chain integration, and optimizing customer-centric logistics services are emerging as important tasks.

However, this transformation process is accompanied by a complex set of challenges, including legal and social responsibilities, evolving labor-market structures, alignment with global norms, and cybersecurity risks. Without a systematic and proactive response to these challenges, domestic container-terminal companies risk losing their competitive advantage in the global market. Thus, in order for domestic container terminals to undergo a significant transformation through the implementation of AI technology in accordance with the global paradigm, it is imperative to formulate a proactive, resilience-based response strategy to mitigate the various adverse implications that AI will engender.

1.2. Aim of Current Study

This study aims to provide practical response strategies for domestic container-terminal companies to enhance their competitiveness and achieve sustainable growth through a strategic AI paradigm shift when the global shipping and port industry is experiencing a strategic inflection point. In particular, this study aims to devise a method for designing a strategic AI paradigm shift roadmap and an AI governance model with inherent resilience to respond to the legal, social, and technological risks arising from accelerating the construction of AI-based smart ports, with emphasis on Busan Port. The significance of this study is that it provides policy and practical directions for domestic container-terminal companies to spearhead the global logistics supply-chain market and become smart container terminals by specifying an AI-based governance system that can implement a paradigm shift across comprehensive container-terminal operations and policy decision-making, beyond merely introducing physical AI equipment and systems within container-terminal companies.

1.3. Research Gap Analysis

Currently, AI technologies are departing from the hype stage and approaching the stage of creating tangible utility and value in the real world, thus promoting paradigm shifts in existing systems across various industries. In particular, AX is defined as a complex phenomenon that promotes continuous change across an organization’s strategy, structure, people, technology, and processes in response to the demands of external stakeholders such as customers, partners, and regulators [4].

The port industry is similarly reflecting this AI-adoption trend. The integration of digital technologies and AI in port operations is rapidly progressing based on the concepts of PORT 4.0 and smart ports. Various operational outcomes have been reported, including efficient logistics activities, safer working environments, improved vessel-traffic management, and container automation [5,6]. Furthermore, some studies reported that the adoption of AI in container terminals contributed positively to operational efficiency, productivity, and safety, as well as enhanced resilience to external shocks and uncertainties [7,8]. However, the adoption of AI technology is accompanied by sociotechnical risks, along with its positive contributions. Privacy, algorithmic bias, cybersecurity, and data misuse expose vulnerabilities in existing systems, thus reducing the reliability of operations [9,10,11]. Additionally, the potential for errors in AI-based decision making, labor-market instability due to automation, and organizational resistance can adversely affect technology acceptance and corporate credibility [12]. These risks pose significant barriers to the realization of the potential of AI technologies. In response to this, recent researchers have recognized that the adoption of AI technology is not merely a technical decision but an issue that affects corporate governance. Thus, a governance framework that guarantees consistency between AI investments and long-term strategic goals, eliminates algorithmic bias, and ensures ethical use must be established [13,14,15]. Additionally, some researchers suggested that the adverse effects of AI transition can be minimized through increased digital literacy, ethical standards, and comprehensive risk-management strategies [16,17].

From this perspective, it is imperative for container terminal companies to pursue AX while ensuring resilience through changes in corporate governance. In essence, the implementation of AI in container terminals is contingent upon achieving a balance between technological effectiveness and resilience. Organizational-level resilience must be cultivated to ensure the capacity to respond to unforeseen threats. This argument serves to emphasize the following novelty of the present study.

While previous studies have primarily focused on AI adoption and governance in general industries, such as manufacturing, finance, and the public sector, this study explores new directions for AX in the container-terminal industry, which is characterized by a complex operational environment. Moving beyond qualitative approaches, it employs a quantitative and systematic analysis to identify the risks associated with AX and develops scenario-based models that reflect diverse success and failure cases specific to container terminals. Through this approach, this study aims to help domestic and international container-terminal companies proactively identify potential risk conditions during the AX and establish stable and successful transformation strategies. Furthermore, this study comprehensively analyzes the impact of AX on operational and policy decision-making in the container-terminal sector and proposes an industry-specific AI governance model that integrates social, legal, and technical risks. This governance model serves as a strategic management system that enables container terminal companies to respond flexibly to various internal and external digital threats and contributes to securing operational efficiency and resilience using AI technology.

The remainder of this paper is organized as follows. Section 2 presents the theoretical foundations and research methodology of this study, with particular emphasis on the backcasting approach and a visual flow of the research framework. Section 3 describes the results of the empirical analysis conducted. Section 4 discusses the academic contributions of this study, focusing especially on the proposed AI governance framework. Finally, Section 5 provides concluding remarks, including this study’s limitations and directions for future research.

2. Theory and Method

2.1. Conceptualizing AX in Container Terminals

This section clarifies the concept of AX and examines the rationale for embracing a paradigm shift based on AI technology by the container-terminal industry as a key strategy for achieving sustainable competitiveness. In particular, it emphasizes that AX for container-terminal companies is not merely about improving operational efficiency but is a preliminary step toward evolving into a data-driven intelligent enterprise.

First of all, although the term “AX” is often used interchangeably with “DX,” there are clear distinctions between the two. A comprehensive definition of DX describes it as “a fundamental change process, enabled by the innovative use of digital technologies accompanied by the strategic leverage of key resources and capabilities, aiming to radically improve an entity and redefine its value proposition for its stakeholders [18]”. In contrast, the concept of AX has been indirectly inferred through studies on digital transformation. However, some researchers [4] argue that AX should be distinguished from DX, as AX entails a clear shift in cognitive tasks from human agents to computational systems. Accordingly, AX can be defined as “the ongoing change in organizational dimensions (strategy, structure, people, technology, and processes), subject to constraints and interests of external forces (customers, suppliers, partners, competitors, regulators), and manifested in AI readiness”.

Next, the sophistication of digital technologies and the proliferation of data have resulted in the rapid development of technologies that mimic human intelligence, such as machine learning, AI, and natural language processing, all of which have enabled the port and maritime industry to develop customized business models that respond promptly and precisely to the granular demands of cargo and shipowners [19,20].

Under this trend, global container-terminal companies are accelerating their transition to the intelligent enterprise through a strategic paradigm shift centered on AI technology, and domestic companies are preparing for an AI-based strategic paradigm shift to respond to the reorganization of the global supply chain and changes in the technological environment. In particular, they are actively applying AI technology to complex processes such as stevedoring, logistics management, and maritime transportation beyond simple mechanical automation. Additionally, they are establishing mid- to long-term strategies, innovating organizational structures, upgrading technology infrastructure, strengthening human-resource capabilities, and reorganizing operational processes. Korea is accelerating the construction of AI-based smart ports, centered on Busan Port, to respond preemptively to changes in the global logistics supply chain [21].

Moreover, Korea is promoting strategic differentiation through real-time maritime data analysis and prediction-based decision-making to optimize cargo management, optimize ship scheduling, and provide customized logistics solutions [22]. The establishment of such a response system is key for accelerating the AI paradigm shift.

2.2. Case Observations of Global AI Adoption in Container Terminals

In this section, we systematically examine the technologies and operational systems deployed by global container-terminal companies to implement the automation and intelligentization of container terminals using AI technologies to realize AX.

The automation phase of container terminals involves increasing operational efficiency by mechanizing or algorithmizing repetitive rule-based tasks [23]. Major global container-terminal operators are operating driverless trailers and carriers utilizing AI technology [24,25,26,27,28]. Other examples include AI-based port management information systems and automated gate systems. In the intelligentization stage, AI and deep learning are used to perform real-time prediction, optimization, and situational awareness, as well as to enhance strategic decision making [29]. Major container-terminal operators are utilizing AI technologies in areas such as cargo forecasting and maintenance, operational automation, safety management, and customs-clearance system automation [30,31,32,33].

This AI-driven technological transformation does not occur in isolation but in conjunction with institutional governance and workforce empowerment strategies. APM Terminal and DP WORLD are establishing AI ethics guidelines based on ISO 27001 certification [34], whereas PSA International is enhancing data transparency and policy alignment through its PORTNET system [35]. The China Merchant Port is establishing a nine-point smart port strategy and platform-based operations under governmental leadership [36]. Additionally, efforts are expended in education and research to develop specialized human resources and internalize technology [35,36,37,38,39].

These examples show that global container-terminal companies are approaching the transition to AI-based operations as a multilayered system in which organizational strategies, systems, and education are organically connected, instead of merely introducing technology. This suggests that the Korean port industry should focus on technology adoption, albeit in parallel with the governance and internalization of workforce capabilities. In particular, global container-terminal companies are maximizing their cooperation with port authorities to realize AI-based governance systems. These global cases are summarized in

Table 1. Overview of AI adoption by global and domestic container-terminal companies.

Operator	Country	Technology	Policy and Strategy	Education
APM Terminal	Denmark	AI-powered port-management information system Wind-resilience tool CONTROLS Emulation	ISO 27001 AI ethical guidelines	ConQuip AI digital-tool-utilization training and digital-literacy internalization training
DP World	Dubai	CARGOES AVA+ CARGOES IoT+	CARGOES platform	DP World Academy Global education portal/platform
PSA International	Singapore	AI fleet management system OptETruck	PORTNET®	PSA Academy
China Merchant Port	China	L4-level smart-port automated unmanned trailer	Lean Operation and Center of Excellence (COE) Nine smart elements of smart port	5G Smart Port Innovation Laboratory
Dongwon Global Container, Busan	Korea	Digital Terminal Operating System(TOS)	Strategy for localization of automation equipment	Cooperation with overseas software companies (AKQUINET, CyberLogitec, etc.)

below. We will use them as a theoretical basis to infer key implications for a sustainable AX strategy.

2.3. Theoretical Integration of Resilience and Social Control for Sustainable AX

In this study, an effective and sustainable AI-based governance framework is constructed by integrating two pillars of strengthening responsiveness to external risks and establishing an internal ethical control system centered on resilience and social control theories. Currently, the container-terminal industry is confronted by an increasingly complex and uncertain business environment as AX accelerates. Because a container-terminal operating system is affected by various exogenous and endogenous risk factors, one must establish a new governance system that can simultaneously enhance resilience and social acceptance, along with developing technological innovations, to effectively manage risks [12,37].

By applying resilience theory to port operations, this study argues that ports must strengthen their capacity to maintain and recover from unpredictable global risks, such as climate change, pandemics, and cyberattacks [38,39]. Resilience theory emphasizes the ability of a system to absorb, adapt, and maintain or rapidly restore its original function when encountering external shocks or stresses [40]. In particular, real-time maritime data analysis, prediction-based port resource allocation, and the establishment of autonomous management systems suitable for the Port Industry 4.0 era, where digital integration and the construction of AI-based systems are accelerating, are key strategies to fundamentally improve the resilience of ports [41,42].

However, AI technologies are inherently opaque and complex, and new forms of governance challenges must be addressed [13,14,15]. For example, algorithmic bias, opacity in decision making, and security vulnerabilities related to the sharing and movement of maritime data between ports, ships and ports, and ports and warehouses cannot be effectively addressed using the conventional hierarchical control models. Therefore, this study applies social control theory to restructure internal control systems such that autonomy and ethical accountability within organizations can be enhanced during the AI transition. Social control theory proposes a control mechanism that strengthens organizational integration and accountability through the autonomy of individual members and the internalization of common norms instead of through regulatory-oriented external control [43]. In accordance with new governance theory, this study should disregard the existing rigid hierarchical structure and perform decision-making based on a multilayered cooperative network involving various stakeholders, such as the government, port authorities, container-terminal companies, shipping and logistics companies, labor unions, and local communities [44].

2.4. Methodological and Conceptual Framework

2.4.1. Macro Perspective: Backcasting-Based Research Design

This study adopted a backcasting approach to formulate an actionable strategy for the AX of the Korean container-terminal industry from a macro perspective. For the AX of the domestic container-terminal industry, a future-oriented long-term strategic plan is essential. Such a plan should include future scenarios that articulate an ideal vision, and backcasting is often proposed as an appropriate methodological approach for designing long-term development policies. Unlike approaches that analyze existing conditions, backcasting is particularly suitable for planning structural transitions and institutional innovations. It is considered effective for complex socio-technical systems—such as port AX—where multi-stakeholder participation, uncertainty management, and goal-driven policy integration are required. Therefore, this study employs a backcasting approach to develop an action-oriented strategic framework that supports scenario-based planning and stakeholder-centered consensus-building [45,46].

In other words, the backcasting method in this study can effectively define a future state, known as “AX in Korean container terminals,” and then backtrack the current change path to realize that future [46]. For this reason, this study considers the future risk environment that will be encountered by Korean container-terminal companies; this study set “resilient AX” as the final goal and conducted a study based on the following key premises of the backcasting approach.

First, resilience must be normative, with a clear vision of a future that involves not only maintenance and improvement. Second, the AI transition must entail radical and disruptive change; incremental changes alone are insufficient to respond to the changing global logistics landscape [47,48,49]. In particular, we applied the resilience design methodology proposed by Kishita et al. [50] to design a backcasting scenario that first assumes a negative future failure and then overcomes it, as shown in Figure 1. By defining various paths, it simultaneously secures the strategic flexibility to explore multiple paths without assuming a single vision.

2.4.2. Micro Perspective: Text Mining and Fault Tree Analysis

This study applied text mining and failure tree analysis (FTA) to identify and analyze risk factors associated with the AI paradigm shift process in domestic container terminals from a micro perspective, as text mining is effective in quantitatively extracting meaningful information from unstructured text data and deriving relationships between keywords to ground the selection and direction of detailed strategic action tasks [51,52]. This study utilized text mining to identify key risk factors in the AI paradigm shift and then adopted FTA to systematically analyze the structural relationships among the risk factors. FTA is a deductive risk-analysis method that logically traces the path of a specific undesirable outcome (top event) [53]. In FTA, basic, intermediate, and logical gates (AND/OR) are used to clearly model the risk-occurrence structure [54,55]. Therefore, this study structurally identifies the various risk factors associated with the AX of container-terminal companies and provides a basis for deriving resilience-based response strategies.

2.4.3. Analytical Framework of Current Study

The research design of this study integrates text mining, FTA, and backcasting approaches.

(1)

We adopted text mining as a research method because the operation and management of container terminals are characterized by the complex and dynamic interconnectedness of various factors, and meaningful risk factors must be objectively derived from a wide range of unstructured data. Text mining is considered a suitable methodology for extracting keywords from significant amounts of data and for identifying correlations between risk factors [51,52]. Additionally, relying solely on the subjective views of experts may cause key risk factors to be overlooked or biased. Therefore, this study quantitatively analyzes potential risk factors related to AX via data-driven analysis. Specifically, NetMiner (version 4.5.1) [56] was used in this study to perform preprocessing and network analysis for text-mining analysis. NetMiner is a software that specializes in unstructured text analysis and social-network analysis. It is a text-mining analysis program used extensively by researchers as it features a Python-based scripting engine that is highly accessible and allows programmatic implementation.

(2)

We adopted the FTA as a research method because this study required a systematic analysis of the structural relationships between individual elements, as well as a logical understanding of the system vulnerabilities as a whole, instead of merely a list of the identified risk factors. Because hazards in complex systems do not exert their effects independently but are interconnected to form risk pathways, FTA allows for a systematic modeling of the overall system risk by structuring failure scenarios and tracing the logical paths contributing to major top events [53]. This process extends beyond enumerative risk identification and contributes to the clarification of prioritized response areas for resilience strategy formulation. Additionally, we used Microsoft Office Excel 2016 and Python 3.12.3 to calculate a series of probability statistics for FTA construction.

(3)

We adopted the backcasting approach as establishing effective flexible strategies based on current forecasting alone is impossible amid the rapid technological paradigm shift in AI and the uncertainty in the global logistics supply chain. Backcasting involves setting a desired future state, i.e., “resilient AX of domestic container terminals,” as a goal and then tracing the strategic path required to realize the goal backward from the present [46].

This study aims to secure both logical consistency and practical contributions by quantitatively analyzing risk factors via text mining, logically analyzing the risk structure via FTA, and reverse engineering an ideal resilience-based strategy via backcasting. The research procedure is shown in Figure 2. Additionally, this study aims to address the following core research questions (RQs) to strengthen the systematization and logical consistency of each research step:

• (RQ. 1) What are the key risk factors and relationships identified during the AI paradigm shift?
• (RQ. 2) What are the key risk scenarios that can threaten the resilient AI paradigm shift of container-terminal organizations?
• (RQ. 3) What are the effective countermeasures to overcome the identified risks and scenarios, and how can they be prioritized for implementation?
• (RQ. 4) What strategic backcasting-based execution roadmap can be designed to realize a resilient AI paradigm shift?

3. Empirical Results

3.1. Risk Assessment: FTA of AX Failures

First, this study aims to identify and categorize the negative dysfunctions that may occur during the AI paradigm shift in advance. To establish a methodological basis for the categorization of the subject under investigation, this study systematically collected and selected relevant literature using the PRISMA framework [57]. In the identification stage, academic articles published between 1 January 2020 and 31 December 2024 were retrieved from the Web of Science database. The search query, centered on the term "Artificial Intelligence Transformation" and incorporating keywords such as "Failure," "Negative Impacts," and "Adverse Effects" using Boolean operators, yielded a total of 1549 records.

In the screening stage, 344 duplicate records were removed, and conference proceedings and review articles were excluded, resulting in 1205 articles. At the eligibility and inclusion stages, these 1205 articles were retained based on document type (journal articles), language (English), and duplication criteria. Then, in the preprocessing stage, NetMiner’s morphological analysis function was used for word segmentation and filtering. This function was adapted to accommodate the unique characteristics of unstructured text data, ensuring data accuracy and consistency. This process established a robust foundation for subsequent text mining-based topic classification and categorization.

Then, we used NetMiner’s ego-network analysis technique as a preparatory step for the FTA. We set “AI” as an ego-node in the full-text data and analyzed the top 100 keywords with high connection strength to AI to identify risk factors directly related to the negative impact of AX. In this context, a high co-occurrence weight with the keyword “AI” is indicative of a term’s frequent appearance in conjunction with AI. This study specifically selected and analyzed only those high-weighted keywords that carry negative semantic connotations.

Summarizing the results of the final analysis, we derived 13 core risk factors related to data accuracy, employment shift, high implementation costs, efficiency loss, cybersecurity risk, regulatory issues, vulnerability, digital surveillance, ethical concerns, privacy rights, technical malfunctions, digital divide, and bias issues, as shown in

Table 2. Results of identifying key negative keywords for AI transformation.

No.	Source	Target	Weight	Rank
1	AI	Data accuracy	406	6
2	AI	Employment shift	405	7
3	AI	High implementation costs	270	12
4	AI	Efficiency loss	134	19
5	AI	Cybersecurity risk	125	22
6	AI	Regulatory issues	107	27
7	AI	Vulnerability	106	28
8	AI	Digital surveillance	91	32
9	AI	Ethical concerns	84	36
10	AI	Privacy rights	79	37
11	AI	Technical malfunctions	74	44
12	AI	Digital divide	71	49
13	AI	Bias issues	58	72

Source: Cyram NetMiner 4.5.1 (Cyram Incorporated, Seoul, Republic of Korea).

below. These key risk factors can serve as the main analysis target for the next stage of defect tree analysis and as a backcasting-based strategy roadmap.

Subsequently, we conducted an FTA based on the text-mining results to derive the causal structure of the AI paradigm shift failure through deductive logic. Fault trees are used to model the propagation of various failure factors throughout a system toward failure [58]. This study utilized the results of text-based data analysis to construct quantitative fault trees. Quantitative fault trees are used to quantify and analyze the probability of system failure and the contribution of each failure factor, which is different from the approach used in previous studies that primarily addressed hardware failure frequency or failure probability. In particular, to overcome the limitation of existing studies that rely on qualitative approaches in the absence of probability information, we constructed a fault tree based on quantitative figures extracted from text mining. To construct the FTA model, we first determined the event and gate criteria based on the 13 core risk factors derived previously as follows:

(1)

We set “failure of AX” as the top-level event. Intermediate events were derived by analyzing degree, betweenness, and eigenvector centrality indicators for the 13 core risk factors identified in advance. Centrality analysis was performed to measure the strength of connections between nodes, the degree of intermediary roles in the network, and the degree of closeness to influential nodes. Based on the analysis, five items were identified as intermediate events: data accuracy, high implementation costs, efficiency loss, vulnerability, and cybersecurity risk. The remaining eight items were categorized as primary events (

Table 3. Results of quantitative analysis for constructing fault tree.

Keyword	Event Type	Frequency of Occurrence	Total Number of Documents	Fault Probability (FP)	Birnbaum Importance
Failure of AX	Top event	-	-	0.0003386	-
Data accuracy	Intermediate event	316	1205	0.262	-
High implementation costs	Intermediate event	172	1205	0.143	-
Efficiency loss	Intermediate event	136	1205	0.113	-
Vulnerability	Intermediate event	107	1205	0.089	-
Cybersecurity risk	Intermediate event	83	1205	0.069	-
Digital surveillance	Basic event	93	1205	0.077	0.00804
Employment shift	Basic event	72	1205	0.060	0.00280
Technical malfunctions	Basic event	73	1205	0.061	0.00429
Regulatory issues	Basic event	68	1205	0.056	0.00377
Privacy rights	Basic event	67	1205	0.056	0.00266
Digital divide	Basic event	43	1205	0.036	0.00273
Ethical concerns	Basic event	40	1205	0.033	0.00325
Bias issues	Basic event	15	1205	0.012	0.00163

Source: Cyram NetMiner 4.5.1, Python 3.12.3.

). The basic events constituted the actual risk occurrence units, and each intermediate event was organized as a set of basic events based on semantic association.

(2)

We used the results of co-occurrence frequency analysis between keywords to set the gate types. Specifically, keyword combinations with a co-occurrence frequency exceeding 30 were set as AND gates, and those with fewer than 30 were set as OR gates. For example, keyword combinations such as (data accuracy–efficiency Loss), (data accuracy–high implementation costs), (data accuracy–technical malfunctions), (data accuracy–vulnerability), (data accuracy–digital surveillance), and (efficiency loss–high implementation costs) were connected by an AND gate. Based on these criteria, all intermediate and primary events were organized as AND and OR gates, respectively.

(3)

We quantitatively analyzed the constructed fault-tree structure. Specifically, we estimated the occurrence probability of each basic event and then performed failure-contribution analysis. The probability of occurrence was estimated using the results of text-mining analysis to calculate the percentage of document occurrence for each basic event. These probability values were calculated based on the frequency of occurrence of key risk factors within the literature-based unstructured data obtained in this study and were used as the basis for the quantitative defect-tree analysis.

In this study, the failure probability ($FP$) of an event occurring via a literature-based approach was calculated as follows (1):

$${FP}_i={\displaystyle \frac{n_i}{N}}$$

(1)

where ${FP}_i$ represents the occurrence probability of basic event i, $n_i$ the number of documents where event i (keyword) appears, and $N$ the total number of documents.

In this study, the probability logic of the entire defect tree was systematically derived based on the logical structure in which the top event, i.e., AX failure, is connected to the intermediate events by AND gates, and each intermediate event is connected to the basic events by OR gates. This logical system structure is based on the principle that (1) if any basic events occur then the corresponding intermediate events will occur, and (2) if all intermediate events occur, then they will affect the parent event.

The logic-gate probabilities were calculated based on the FP using the following Equations (2) and (3):

$$(\mathrm{Probability} \mathrm{of} \mathrm{AND} \mathrm{gate}) P_{AND}={FP}_1\times {FP}_2\times \dots \times {FP}_n$$

(2)

$$(\mathrm{Probability} \mathrm{of} \mathrm{OR} \mathrm{gate}) P_{OR}=1-(1-{FP}_1)\times (1-{FP}_2)\times \dots \times {(1-FP}_n)$$

(3)

In this study, we applied the Birnbaum importance to quantitatively assess the sensitivity of each basic event to the occurrence of the top event. The Birnbaum importance is a sensitivity-based metric that quantifies the effect of a minor change in the failure probability of a specific basic event on the overall probability of system failure. It is particularly useful for comparing the relative importance of risk factors within a system. In this study, we applied this metric by increasing the failure probability of each basic event in fixed increments (e.g., 0.01) and measuring the resulting change in the probability of the top-event occurrence. Because the probabilities of the top event were derived from literature-based text mining, the analytical structure resembles the classical Birnbaum framework. However, instead of using the conventional system-reliability models, we substituted the conditional probability of system failure with the probability of the top-event occurrence. This allows for a more context-specific evaluation of the effect of each risk factor within the fault-tree structure. The general form of Birnbaum’s importance is written as follows (4):

$${BM}_i={\displaystyle \frac{\partial R\left(S\right)}{\partial R\left(X_i\right)}}=P_r\left(S=0|X_i=0\right)-P_r\left(S=0|X_i=1\right)$$

(4)

In this study, it is adapted as follows (5):

$${BM}_i={\displaystyle \frac{\partial R\left(S\right)}{\partial R\left(X_i\right)}}=P_r\left(TopEventFailure|X_i=1\right)-P_r\left(TopEventFailure|X_i=0\right)$$

(5)

where ${BM}_n$ denotes the Birnbaum importance of basic event $X_i$; $P_r$(TopEventFailure|$X_i$ = 0) is the conditional probability of the top event occurring when $X_i$ is in a normal state; and $P_r$(TopEventFailure |$X_i$ = 1) is the conditional probability of the top event occurring when $X_i$ is in a failed state.

In this study, quantitative analysis via text mining was performed to identify the systematic position of each risk factor in the fault-tree structure while considering the effect of each risk factor on the system failure. This analytical approach provides a logical and objective basis for prioritizing future response strategies.

In summary, this study systematically identified negative dysfunctions in the AX process and presented a probability-based fault-tree model organized based on a deductive logic structure, as shown in Table 3 and Figure 3.

3.2. Scenario Planning: Failure and Success Futures

In this study, a fault-tree model was created by systematizing the various risk factors that can cause the AI paradigm shift to fail at the current level in a logical and deductive manner. We comprehensively included the technical, policy, and social risk factors that may occur during AX. Based on the fault-tree model, we specified the major risk scenarios using a domestic container-terminal company as an example. Finally, we conducted scenario planning to develop strategies for responding to each risk factor and to derive alternatives with inherent resilience.

Specifically, scenario planning aims to proactively identify potential risks, systematically formulate strategies to respond to them, and design resilience-based action paths to achieve the ideal future vision. Using a scenario-based approach, we established an ideal future vision for a domestic container-terminal company to secure resilience and realize sustainable growth through AX.

3.2.1. Envisioning a Failed-AX Future

Based on the fault-tree structure, we systematically designed various risk scenarios that can cause the AI paradigm shift to fail in domestic container-terminal companies centered on AI systems. Each scenario was refined based on the results of previous studies, and the occurrence paths and effects of potential risk factors were derived based on deductive logic.

• The scenario in which AI systems cause data-accuracy problems is as follows: The AI system may fail to accurately obtain data pertaining to cargo stowage, vessel arrival, and departure management because of technical issues [59,60]. Consequently, the sensor signals are incomplete and server instability issues occur, thus resulting in the accumulation of database errors. Subsequently, the surveillance system (digital surveillance) cannot distinguish between normal and error data; moreover, it automatically executes incorrect unloading commands [61], and the lack of a recovery system causes errors in the cascade. Additionally, AI systems obtain additional personal information and terminal user logs during the recovery process, which results in privacy-rights issues [11]. Consequently, shipowners, shippers, and port workers lose confidence in AI systems, terminal unions demand operational overhauls, and port regulators impose sanctions owing to data accuracy and bias issues (Regulatory Issues) [62]. Eventually, AI-based operating systems will collapse, and terminals will be at risk of reverting to conventional manual operations.
• A scenario exists in which AI systems exacerbate the problem of high implementation costs. The system incurs significant fixed and variable costs in performing terminal automation, AGV operation, and IoT infrastructure deployment [63,64]. Additionally, the difficulty in securing high-level digital professionals reduces system reliability, thus resulting in delays and failures in the introduction of AI systems, particularly in small- and medium-sized ports. Consequently, AX is centered on large-hub container terminals, whereas small- and medium-sized ports are not considered for AX, thus widening the digital divide. Simultaneously, AI systems promote human-resource redeployment and employment shifts, whereas intensified data surveillance causes ethical issues and regulatory uncertainties, thus decelerating the overall momentum of AX.
• Scenarios exist where AI systems cause operational inefficiencies, including insufficiently trained or unoptimized AI systems that result in frequent errors in core port operations such as ship scheduling, stevedoring, and yard operation optimization [65,66,67]. Consequently, the vessel waiting time increases, the overall operational productivity of terminals decreases, and the global logistics flows are severely disrupted. Additionally, excessive data acquisition increases transmission delays and exposure to hacking, while AI systems fail to clarify accountability in the decision-making process, thus resulting in a decline in organizational trust and increased conflict among stakeholders.
• AI systems are vulnerable to adversarial cyberattacks and high-dimensional data distortion [68], particularly when biased algorithms are applied, which can adversely affect a particular shipper or shipping company [69]. Such biases can severely undermine system-wide reliability, and cybersecurity vulnerabilities in the digital surveillance-based infrastructure can cause cascading damage, including logistics data breaches, terminal system paralysis, and cargo transportation delays. Additionally, AI systems pose a greater risk of causing worker displacement, widening the digital divide between regions, and spreading social resistance [63,70].
• Scenarios exist in which AI systems increase cybersecurity risk. Systems obtain and process sensitive information such as maritime, cargo, and ship-location data on a large scale, thus rendering them prime targets for hackers [9,10]. Information obtained through digital surveillance systems is exposed to external hacking attacks, and attacks that exploit vulnerabilities in AI systems are common [71]. Additionally, the lack of privacy regulations exacerbates institutional imbalances [72], and the lack of accountability in AI decision-making processes results in confidence loss toward shipping companies, shippers, and logistics companies, which severely threatens the sustainability of AX in container terminals.

3.2.2. Envisioning a Resilient Future for AX

This section presents the following strategic scenarios centered on AI systems to overcome the unfavorable scenarios derived in Section 3.2.1 and to realize the AX of container-terminal companies with resilience. In particular, we categorize and discuss an ideal future vision by deriving specific response directions that can overcome basic and intermediate events based on an established fault-tree structure.

• The strategic scenario for AI systems to overcome the issue of data accuracy is as follows: AI systems operate reliably based on standardized structures and automatic recovery capabilities. This involves real-time error detection and redundant designs that proactively prevent data-quality degradation in the data exchanged between ports and ships, ship-to-ship logistics, port-to-shore logistics, and ship-to-shore logistics. Digital surveillance systems have shifted to a privacy-centric design, thus minimizing the acquisition of sensitive information and resolving privacy-related conflicts through coordination within a governance framework through the participation of various stakeholders. AI systems institutionalize data authentication and audit procedures to gain trust from internal and external stakeholders, which consequently enhances the reliability and sustainability of smart-port operations.
• The strategic scenario for AI systems to overcome the issue of high implementation costs is as follows: The system attracts strategic investments and subsidy support from local governments to ease the burden of initial implementation costs and optimizes the fixed cost structure by jointly utilizing infrastructure among small- and medium-sized ports. Workers engaged in conventional port stevedoring will be transformed into digitally applied professionals through AI-driven retraining and job-transformation programs. Additionally, container-terminal operators and logistics companies can promptly implement AI infrastructure and facility investments based on clear legal and institutional guidelines. This systematic approach will enable AX to expand not only to large ports but also to small- and medium-sized ports, thus reducing digital inequality in the port industry.
• The strategic scenario for AI systems to overcome operational inefficiency losses is as follows: AI models make accurate and rapid decisions through high-quality data-driven learning and recover from errors without cascading delays using automatic recovery and redundancy schemes. Surveillance systems actively adopt permission-based access, encryption, and anonymization technologies to protect privacy and ensure social acceptance. Simultaneously, the AI-integrated management platform is utilized to monitor operations and improve problems based on ethical standards and accountability systems, thereby increasing the efficiency and reliability of digital port operations.
• The strategic scenario for AI systems to overcome the issue of system vulnerability is as follows: The AI system internalizes a multireview algorithmic structure and an ethics-based decision-making system to prevent biased algorithms, whereas the cybersecurity system enhances threat detection, security authentication, and anonymous data-processing technologies to support continuous and reliable port operations. Additionally, port workers adapt proactively to AI systems through job transformation and capacity-building programs, thus reducing labor unrest and resistance. Digital-infrastructure subsidies and transition deferral schemes for small- and medium-sized ports supported by national and public authorities have been realized to promote equity and resilience during the nationwide AI transition.
• The strategic scenario for AI systems to overcome cybersecurity risks is as follows: AI systems adopt “security by design” from the beginning of their development, thus internalizing privacy and access control policies. All data obtained are anonymized and encrypted to prevent leakage and misuse, and surveillance systems are securely operated based on a policy-based design and ethical evaluation criteria. Real-time intrusion detection and response systems are implemented to respond promptly to external cyberattacks, and regulatory agencies are constantly updating AI cybersecurity technology standards to strengthen corporate security certification and accountability structures. Additionally, the AI-integrated management platform prevents ethical risks through internal and external monitoring, thus enabling container-terminal companies to establish a world-class AI-based port-operation system characterized by safety, reliability, and accountability.

3.3. Roadmap: Strategic Roadmapping Toward a Resilient Future

3.3.1. Prioritization of Strategic Measures and Roadmap Structuring

This section synthesizes the results of FTA based quantitative metrics and scenario planning to propose a backcasting-based strategic roadmap for materializing an ideal future vision. To ensure the quantitative validity of the strategy formulation, a quantitative analysis was conducted using the Birnbaum importance to calculate the risk-contribution score of each underlying event, which was then combined with the results of applicability assessment by container-terminal experts to enhance the acceptability of this study. For the analysis, we summed the failure contributions of the associated underlying events for each countermeasure to calculate a risk-contribution score and systematically evaluated the relative importance of each countermeasure in enhancing system resilience.

The risk contribution score (R) of each countermeasure is calculated using the Birnbaum importance as follows (6):

$$R_k=\sum _{i\in E_k}{BM}_i$$

(6)

where $R_k$ is the risk-contribution score of countermeasure k, $E_k$ the set of basic events associated with countermeasure k, and ${BM}_i$ the Birnbaum importance (i.e., the contribution to top-event failure) of basic event i.

Additionally, expert-evaluation scores were derived based on the opinions of 40 key stakeholders from actual domestic container-terminal companies, which were the result of systematically evaluating the practical applicability of each measure in response to risks that may occur in the AX of container-terminal companies. Therefore, all data were normalized to compare the relative importance of the risk-contribution score and expert-applicability score calculated for each countermeasure. Finally, the priority score for each countermeasure was calculated by averaging the normalized risk-contribution and expert-applicability scores.

The final prioritization score for each countermeasure is calculated as follows (7):

$${SCORE}_k={\displaystyle \frac{\widehat{R_k}+\widehat{A_k}}{2}}$$

(7)

where ${SCORE}_k$ is the final prioritization score of countermeasure k, $\widehat{R_k}$ the normalized risk-contribution score of countermeasure k, and $\widehat{A_k}$ the normalized expert-applicability score of countermeasure k.

The results are summarized in

Table 4. Prioritized results with quantitative scoring per countermeasure.

No.	Strategy	R	A	$\hat{R}$	$\hat{\mathrm{A}}$	Score	Step
1	Establish privacy-protection and data-management systems	0.008	3.774	1	1	1	Short
2	Strengthen cybersecurity certification system for port terminals	0.005	3.698	0.614	0.889	0.751	Short
3	Strengthen R&D for cybersecurity solutions	0.008	3.415	1	0.472	0.736	Short
4	Secure initial cybersecurity budget	0.008	3.34	1	0.361	0.681	Short
5	Design AI–human cooperation process	0.004	3.642	0.355	0.806	0.58	Middle
6	Establish a red-team organization to balance decision making	0.002	3.717	0.196	0.917	0.556	Middle
7	Monitor AI systems in real time	0.004	3.528	0.462	0.639	0.551	Middle
8	Introduce AI-algorithm optimization system	0.004	3.528	0.462	0.639	0.551	Middle
9	Expand in-house training opportunities to internalize new technologies	0.003	3.623	0.248	0.778	0.513	Middle
10	Establish a strategy for phased introduction of AI systems	0.003	3.623	0.238	0.778	0.508	Middle
11	Activate existing workforce reassignment and job-transition programs	0.003	3.566	0.248	0.694	0.471	Middle
12	Introduce AI-risk (human rights, safety, security, etc.) analysis system	0.003	3.528	0.27	0.639	0.454	Middle
13	Perform quality control and optimization for AI decision making	0.004	3.396	0.462	0.444	0.453	Middle
14	Establish AI human-rights impact assessments and ethical guidelines	0.003	3.547	0.217	0.667	0.442	Middle
15	Enhance transparency of AI governance activities and expand stakeholder engagement	0.003	3.472	0.312	0.556	0.434	Middle
16	Introduce AI-related insurance and third-party certification systems	0.004	3.321	0.388	0.333	0.361	Long
17	Analyze AI-related legal systems	0.004	3.302	0.388	0.306	0.347	Long
18	Optimize the integration of AI systems with existing systems	0.003	3.396	0.238	0.444	0.341	Long
19	Perform AI-based system standardization and DB management	0.001	3.491	0	0.583	0.292	Long
20	Manage automatic AI error recovery and redundant systems	0.004	3.151	0.462	0.083	0.273	Long
21	Activate infrastructure investment attraction (establishment of funds, funds, etc.)	0.003	3.283	0.238	0.278	0.258	Long

. Based on the results, categorization was performed based on the priority score of each countermeasure. For example, countermeasures with a priority score of 0.6 or more were categorized as short-term strategies, those with a priority score of 0.4 or more but less than 0.6 were categorized as medium-term strategies, and those with a priority score of less than 0.4 were categorized as long-term strategies.

3.3.2. Strategic Roadmap for Resilient AI Paradigm Shift in Container-Terminal Operations

This study synthesizes the results of a quantitative analysis and scenario planning based on FTA to propose a backcasting-based strategic roadmap such that a resilient AI paradigm shift can be realized for domestic container-terminal companies. This is shown in

Table 5. Strategic roadmap for resilient AX of container-terminal companies.

Distinctions		Contents
Vision		Sustainable, innovative, and resilient AX of container terminals based on resilient-AI governance
Goal	Short term	Initially stabilize AX and creation of cybersecurity foundation
	Mid term	Stabilize AI system and establish ethics and governance system
	Long term	Complete AI infrastructure and policy foundation, and strengthen global competitiveness
Step		Strategies
Short term (1–2 years)		[Private] Establish privacy-protection and data-management systems [Public and private] Strengthen cybersecurity certification system for port terminals [Public and private] Strengthen R&D for cybersecurity solutions [Public] Secure initial cybersecurity budget
Mid term (3–5 years)		[Private] Design AI–human cooperation process [Private] Establish red-team organization to balance decision making [Private] Monitor AI systems in real time [Private] Introduce AI-algorithm optimization system [Private] Expand in-house training opportunities to internalize new technologies [Public and private] Establish a strategy for phased introduction of AI systems [Public and private] Activate existing workforce-reassignment and job-transition programs [Public and private] Introduce AI risk (human rights, safety, security, etc.) analysis system [Private] Perform quality control and optimization for AI decision-making [Public] Establish AI human-rights impact assessments and ethical guidelines [Public and private] Enhance transparency of AI governance activities and expand stakeholder engagement
Long term (5–10 years)		[Public] Introduce AI-related insurance and third-party certification systems [Public] Analyze AI-related legal systems [Private] Optimize the integration of AI systems with existing systems [Public and private] Perform AI-based system standardization and DB management [Private] Manage automatic AI error recovery and redundant systems [Public] Activate infrastructure-investment attraction (e.g., establishment of funds)

below

(1)

The short-term strategy aims to initially stabilize the AI system. Privacy-protection and data-management systems should be established urgently, such that AI systems in container terminals can operate reliably and securely. Additionally, a multilayered cybersecurity certification system that considers the specificities of container-terminal operations must be introduced to ensure system security and reliability. Furthermore, cybersecurity budgets must be secured proactively from the early stages, and research and development (R&D) must be conducted on security solutions to strengthen long-term stability. These measures minimize the risks that may arise in the early development stages of AI systems and serve as a foundation for stable digital operations.

(2)

The medium-term strategy aims to optimize AI systems and establish a foundation for human–AI collaboration. AI systems must improve their prediction accuracy and operational efficiency through continuous learning and algorithm optimization based on high-quality data, as well as secure the ability to detect system errors early and respond promptly via a real-time monitoring system. Additionally, a collaborative process between humans and AI systems should be designed to enable flexible responses to unpredictable situations, and an integrated risk-management system should be established in advance to analyze and manage the ethical and social risks that may arise in the AI decision-making process. Moreover, an integrated AI management platform with an independent review system should be introduced to ensure the balance and ethics of AI system operations, and digital human rights and ethical standards should be strengthened to secure the trust of stakeholders.

(3)

The long-term strategy aims to complete the AI paradigm shift in the port logistics industry based on AI systems and secure leading competitiveness in the global market. To achieve this, one must optimize the integration between the existing port operation and AI systems as well as strengthen scalability and connectivity via standardization and a database (DB) management system. Additionally, one must minimize system operation risks by establishing automatic recovery functions and redundant infrastructure (cloud-on-premises integration) when errors occur in AI systems. For example, cooperation between the Ministry of Oceans and Fisheries (government), the port authority, and private container-terminal operators should be encouraged to provide continuous financial support for AI infrastructure. Legal frameworks should be established to support the spread of AI technology across industries, and financial risk-management systems should be established through third-party certification of AI systems and insurance systems for unexpected risks.

In summary, the strategic roadmap presented herein aims to realize a complete paradigm shift in container-terminal companies centered on AI systems instead of merely introducing technology. Thus, domestic container-terminal companies are expected to create sustainable growth models that advance the global port logistics market through technology, policy, and institutional aspects.

4. Discussion

4.1. Proposal for Resilient AI Governance Model and Its Organizational Role

In this section, we propose a resilience-based AI governance system, as shown in Figure 4, to practically implement the strategic roadmap derived earlier and to stably and sustainably promote the AI paradigm shift of domestic container-terminal companies. To effectively promote the AI paradigm shift, container-terminal companies should establish an integrated AI governance system that extends beyond technology adoption and encompasses organizational strategy, human-resource operations, data management, and decision-making systems while considering the following four components.

(1)

Domestic container-terminal companies should establish an AX management system based on compliance and policy responses that faithfully reflect the regulations of the International Maritime Organization (IMO) and global technical standards. This includes the IMO’s MASS Regulatory Scoping Exercise, the EU’s AI Act and Data Act, and each country’s maritime and privacy laws (e.g., GDPR). Additionally, key international and national laws and policies, such as the EU AI Act and Data Act, national maritime laws, privacy laws (GDPR), and the IMO Cyber Risk Management Guidelines, should be proactively integrated and reflected to help container-terminal companies proactively identify and prevent legal risks that may arise during their AI paradigm shift to ensure global regulatory responsiveness and system reliability. The AI governance system should ensure institutional consistency and global acceptance of the AI paradigm shift by simultaneously strengthening the regulatory-monitoring system, policy-response protocols, and external cooperation channels to respond actively to the changing policy environment.

(2)

Domestic container-terminal companies should establish a feedforward-based all-around AX risk assessment system to predict and respond to potential risks that may arise when implementing strategic tasks based on AI governance. AI systems should repeatedly and systematically evaluate risk factors at every stage of port-logistics management, vessel scheduling, cargo handling, and system maintenance, and establish preemptive response strategies through maritime data-based risk modeling. In particular, they should establish a risk-management system based on an integrated AI platform that integrates real-time maritime data acquisition, analysis, and prediction capabilities to detect and prevent system failures, cyberattacks, and ethical issues at an early stage.

(3)

Under the AI governance system, container-terminal companies should establish a proactive early-response system to respond promptly to ethical, human-rights, environmental, and safety issues that may arise during the AI paradigm shift. The early-response system should apply a machine-learning-based early-warning system to detect abnormalities in real time and activate various scenario-based response manuals based on success and failure to prevent the proliferation of problems. Additionally, by accumulating and analyzing recurring errors and risk data to continuously optimize response protocols, the system should strengthen both resilience and sustainability during the AI paradigm shift.

(4)

Container-terminal companies should establish an operating system based on transparency and accountability to establish an AI governance system. The progress of the AI paradigm shift, the manner in which maritime data are utilized, and key decision-making procedures should be clearly disclosed to all stakeholders, and the relevant parties should be clearly defined for each step of the implementation. Additionally, AI literacy should be internalized as an essential competency for all employees, from the CEO of the container terminal to on-site managers and operators. The strategic goals of AX should be communicated and practiced throughout the organization via systematic education and training.

4.2. Implications and Contributions

This study aims to propose strategic directions and implementation frameworks for the successful advancement of AX in domestic container terminal companies. The present study aims to contribute to the theoretical framework and to the practical design of policy by means of an integrative presentation of a resilience-based AI governance model and transformation roadmap. These address legal, social, and technological risks, elements which are often overlooked in existing research. The academic contributions of this study are as follows:

(1)

This study defined the AI paradigm shift of domestic container-terminal companies as a process that fundamentally reorganizes the entire strategy, structure, operation system, and human resource management, instead of merely adopting technology. In particular, it emphasized that the AI system transforms the entire process of port operation into a data-based decision-making system and that container-terminal companies must simultaneously innovate not only the technological infrastructure but also the organizational culture, leadership, and decision-making methods.

(2)

This study revealed that establishing a resilience-based strategy is essential for responding to various risks and uncertainties that may inevitably arise during the AI paradigm shift. Unlike the DX phase, the AI paradigm shift can continuously generate complex and dynamic risk factors (e.g., cybersecurity threats, algorithm bias, and system errors). This implies the necessity to develop a resilience-centered response system with preemptive risk detection and recovery capabilities.

(3)

A resilience-based AI governance system was designed in this study, which is necessary to successfully promote AI transition, thus differentiating this study from previous ones. A global regulatory response system that integrates social control and new governance theories should be developed to internally strengthen ethics and accountability, as well as to externally comply with the IMO’s Autonomous Ship Regulatory Framework (MASS RSE), the EU AI Act and Data Act, the GDPR, and the IMO Cyber Risk Management Guidelines.

(4)

This study emphasized that the AI paradigm shift is not limited to technological innovation but requires fundamental changes in organizational culture and mindset. Container-terminal companies should minimize resistance to organizational transformation and maximize acceptance by implementing a strategy for internalizing AI literacy at the company-wide level, learning from successes and failures through initial pilot projects, and then gradually imparting the knowledge gained throughout the organization.

(5)

This study proposed a method to overcome the limitations of the existing feedback-centered post-response system and integrated a feedforward-based predictive risk management system into an AI transformation strategy. In particular, a structure was systemized to detect and prevent possible failures in an AI system in advance through real-time data analysis, an early-warning system was established, and a predictive maintenance system was introduced.

This study redefines the AI paradigm shift as the structural transformation of port operations, i.e., not merely technology adoption, and proposes a resilience-based AI governance system and feedforward strategy to promote it. AI-based operating systems provide high automation efficiency; however, they pose the risk of system disruption owing to a single point of failure or cyberattacks. Therefore, “cyber resilience” is emerging as a key strategic element in the shipping and logistics system. Additionally, beyond Zero Trust, AI-based security, and supply-chain security, cyber resilience should be the last form of protection in the digital era [73] and a key condition for port competitiveness. This study provides practical and policy contributions to the establishment of digital transformation strategies for the domestic port industry and suggests strategic directions for securing sustainable global competitiveness.

5. Conclusions

In this study, a strategic direction and an implementation system for domestic container-terminal companies were systematically presented to successfully realize AX. Based on the unique nature of the container-terminal industry and the rapidly evolving nature of the global logistics environment, an AX strategy model was established to fundamentally reorganize the entire strategy, organizational structure, operating system, and professional human-resource management beyond the simple introduction of equipment and technology. Additionally, this study scientifically identified the major risk factors that may occur during AI transformation using text mining and FTA, as well as specified a step-by-step implementation strategy that inherently embodies resilience via the backcasting approach. Thus, the complex risk structure of the AI paradigm transformation, which is distinct from DX, was systematically analyzed, and a practical strategic direction was derived in response to it. A resilience-based AI governance system was designed to support the AI paradigm shift by integrating social control and new governance theories. The significance of this study is that it systematically designed a system to simultaneously secure regulatory responsiveness and ethical responsibility by preemptively reflecting global norms, such as the IMO’s MASS regulatory framework, the EU AI Act and Data Act, and the GDPR. The proposed system overcame the limitations of the existing feedback-based risk-management system. Additionally, a feedforward-based predictive risk-management system was proposed for the first time in this study, thereby establishing a structural innovation direction that can prevent AI system failures in advance. This study simultaneously presented a theoretical model and practical application strategies to strategically and systematically support the AI paradigm shift of domestic container-terminal companies. Furthermore, this study systematically structured the relationship between the AI paradigm shift and port resilience from an academic perspective, established an independent theoretical framework distinct from DX, and developed a flexible and adaptive AI governance design model by integrating social control and new governance theories. Additionally, the panopticon control problem that may be inherent in AI systems was critically examined, and a theoretical direction for responding to the side effects of technological innovation was presented by specifying ethical and social control mechanisms. From a social perspective, this study provided a practical and feasible AI transition roadmap and response strategy for domestic container-terminal companies to secure sustainable competitiveness in the global shipping and logistics markets. This increases the possibility of field application through data-based risk analysis, design of ethical AI utilization guidelines, and internalization of AI literacy within the organization, as well as specifying a policy direction that can preemptively accelerate AI-based innovation in the Korean container-terminal industry by comparing major global smart port cases. This study contributes significantly in terms of strategically supporting the AI paradigm shift of domestic container-terminal companies; however, several limitations remain. Considering the rapid development of AI technology and the rapid changes in the global supply-chain environment, continuous and in-depth follow-up research is warranted. In particular, regarding the AI paradigm shift in domestic container-terminal companies, efforts should be expended to systematically analyze and respond to ethical and social risks instead of merely aiming for technical efficiency. In future studies, the issue of panopticonization, which may be caused by AI systems operated by container-terminal companies, should be investigated comprehensively. Follow-up studies should focus on (i) establishing a strategic foundation for realizing the social acceptance of AI systems and an AI paradigm shift by systematically analyzing the effect of AI systems on port operations and worker management; (ii) the possibility of privacy infringement, labor rights deterioration, and imbalances in data governance; and (iii) specifying institutional and ethical response measures to control and prevent these issues.