Harnessing business value from artificial intelligence through dynamic capabilities and sociotechnical microfoundations

Artificial intelligence (AI) has evolved from narrow applications in specific tasks, such as rule-based predictions of customer behaviours to addressing various business challenges (Koreis et al., 2025; Fosso-Wamba et al., 2024a). While AI promises automation, enhanced decision-making and operational optimisation, real-world implementations fall short of expectations, with many organisations struggling to translate AI into sustained strategic value (Enholm et al., 2022). This gap between AI's potential and realised outcomes is particularly pronounced in asset-intensive organisations, such as those in Transport and Logistics (T&L), where technological change must be accommodated within tightly coupled operational systems, fragmented legacy infrastructures, regulatory constraints and established work practices (Fosso-Wamba et al., 2024a, b; Richey et al., 2023). Hence, in addition to calls for research on how organisations across industries leverage AI to generate business value (Ooi et al., 2025; Enholm et al., 2022; Richey et al., 2023), the contextual, empirical and theoretical problems discussed below motivated this study.

In terms of context, T&L organisations, projected to exceed $13.7 trillion in value by 2027 (Statista, 2025), play a critical role in inter/national economies. Yet these organisations face escalating costs, workforce shortages, sustainability imperatives and disruptions (Richey et al., 2023; Shen and Sun, 2023). Although AI is widely recognised as a potential response to these challenges, existing research in the T&L domain remains predominantly conceptual or experimental (Richey et al., 2023; Koreis et al., 2025; Li et al., 2025). As a result, we lack empirical understanding of how T&L firms implement AI, what capabilities they develop, and how these efforts translate into business value (Lee et al., 2023). Scholars have recognised this need and called for empirical research at the intersection of AI and logistics/supply chain management (Richey et al., 2023). Moreover, while studies in manufacturing and B2B settings have reported positive relationships between AI and organisational performance (Fosso-Wamba et al., 2024a, b; Mikalef and Gupta, 2021), such research often assumes firms possess the sociotechnical readiness to adopt advanced technologies. These assumptions contrast with the long-standing reluctance of T&L organisations to invest in advanced systems and digital skills development (Molla et al., 2025; Gekara et al., 2019). Consequently, the lack of empirical studies demonstrating AI benefits and the contextual conditions required to realise them limits our understanding of how AI-based capabilities are developed and deployed in specific business contexts (Pan and Nishant, 2023).

Addressing these contextual and empirical challenges requires moving beyond questions of AI adoption or technical performance to examining how organisations build, adapt and mobilise AI-related capabilities in technologically dynamic and disruption-prone environments. Unlike adoption or resource-based perspectives, which explain whether AI is acquired or possessed, the dynamic capabilities (DCs) theory (Teece, 2007) provides a useful analytical lens to unpack how firms leverage digital technologies to transform business models, enhance agility, respond to environmental turbulence and enable innovation (Fosso-Wamba et al., 2024a, b; Mikalef et al., 2021; Shen and Sun, 2023). However, existing research has paid limited attention to AI as a distinct technological domain, and even less to the sociotechnical microfoundations through which AI-based DCs, as a specialised subset of organisational capabilities, are enacted in operationally constrained contexts such as T&L. While DCs are understood to rest on microfoundations (Chen et al., 2023; Schilke et al., 2018), prior studies have predominantly highlighted social and behavioural microfoundations such as training programs, employees' risk orientation and collaboration routines (An et al., 2024; Enholm et al., 2022). This social-centric orientation tends to overlook technological microfoundations, including AI-specific infrastructures, data pipelines and algorithmic routines. Yet both researchers and practitioners underscore the necessity of a sociotechnical understanding to “select, orchestrate, and leverage AI-specific resources” (Mikalef and Gupta, 2021, p. 2), encompassing operational processes, workforce skills, governance mechanisms and technological assets (An et al., 2024; Enholm et al., 2022). Clarifying these sociotechnical microfoundations is essential for understanding how AI systems can be effectively aligned with business objectives and integrated into organisational architectures.

In general, despite a growing interest in AI research, there is a need for contextually grounded empirical studies that distinguish among different forms of AI-based DCs, identify their microfoundations, situate them within specific organisational environments and clarify the mechanisms through which AI generates business value, defined as realised operational, environmental and strategic outcomes including efficiency gains, safety improvements, sustainability outcomes and enhanced customer experience (Enholm et al., 2022; Pan and Nishant, 2023). Hence, this study contributes to the following questions:

To address these questions, we focused on the organisation as the primary unit of analysis and conducted a qualitative study using a theory elaboration strategy. Theory elaboration involves extending existing conceptual ideas or preliminary models to generate new theoretical insights by refining, contrasting or specifying constructs in light of empirical evidence (Fisher and Aguinis, 2017). Guided by the microfoundations of DCs, we analysed qualitative data from Australian T&L organisations to develop a granular understanding of how AI is harnessed for operational and sustainability outcomes. Australia provides an appropriate context because the T&L sector is a major economic contributor and faces pronounced capability-reconfiguration challenges (Molla et al., 2025; Gekara et al., 2019).

Our findings show that T&L organisations are adopting both generic AI tools and industry-specific AI-enabled systems, such as autonomous guided vehicles, robotic forklifts and collaborative robots. When combined with AI-related skills, organisational routines and governance processes, these technological and social elements form the microfoundations of AI-based DCs for sensing, seizing and transforming transport, warehousing and delivery operations. These capabilities enable firms to enhance worker safety, operational efficiency, customer experience and environmental sustainability. Building on these insights, we develop a model and propositions that explain how AI assets, skills, governance and process microfoundations interact to create AI-based DCs and generate business value.

The remainder of the paper proceeds as follows. We first review AI research in T&L and outline the theoretical foundations of DCs and their microfoundations. We then present the research design, method, empirical findings and the study's theoretical and practical implications.

This study is motivated by interconnected contextual, empirical and theoretical issues in understanding how AI creates value. To frame the study, we integrate insights from three complementary literature: AI studies in T&L, AI capabilities and the microfoundations of DCs.

T&L organisations include firms engaged in warehousing, postal and parcels and transport across maritime, rail, road and aviation (Tay et al., 2021). These firms have progressed through various stages of technological change, with AI becoming central to digital transformation initiatives (Fatorachian and Kazemi, 2025; Bawack et al., 2021). Two broad research streams characterise the AI in T&L literature. Computational studies develop models and simulations to demonstrate AI's potential in process optimisation, monitoring the flow of perishables, inventory management and intelligent warehousing (Balster et al., 2020; Tu et al., 2021). While these studies report efficiency gains and cost reductions, they typically rely on simulated environments and do not validate real-world implementation or examine mechanisms for building AI-based capabilities.

Behavioural studies, closer to our focus, examine user intentions and organisational adoption-for example, truck drivers' intention to use AI assistants (Loske and Klumpp, 2021), AI affordances for personalised urban transport and infrastructure management (Li et al., 2025), AI adoption in warehousing (Koreis et al., 2025) and agri-food logistics (El Bhilat et al., 2024). These works note benefits such as reduced travel time, improved pick rates and enhanced customer satisfaction, alongside concerns about potential misuse. Other research highlights AI-enabled digital platforms that help firms respond to global supply chain disruptions (Shen and Sun, 2023) and AI's role in automating logistics activities, as well as improving information sharing and decision synchronisation (Garg et al., 2025). However, with a few notable exceptions (e.g. Shen and Sun, 2023; Garg et al., 2025), the literature largely emphasises AI's potential or performance outcomes rather than empirically examining AI-based DCs and how firms capture value. To address this limitation, we next turn to the broader AI capabilities literature to clarify what capabilities comprise and how they underpin value creation.

AI capabilities can be viewed from technological and organisational perspectives (Fosso-Wamba et al., 2024b; Mikalef and Gupta, 2021). Technologically, they encompass perception (e.g. sensing via sensors and image recognition), comprehension (reasoning and understanding), action (autonomous execution) and learning (pattern recognition from data) (Bawack et al., 2021; Schuetz and Venkatesh, 2020). Organisationally, AI capabilities represent “a firm's ability to structure, bundle, and leverage its AI-based resources” (Mikalef and Gupta, 2021), integrating technological assets with skills, processes and structures. When organisations develop higher order capabilities to effectively deploy AI resources in sensing environmental shifts, seizing opportunities and transforming business processes, they create AI-based DCs that enhance performance (Abou-Foul et al., 2023; Fosso-Wamba et al., 2024a, b). For instance, JD.com's intelligent digital platforms, combined with an agile and collaborative business model, improved forecasting, replenishment and distribution and strengthened resilience and performance during COVID-19 (Shen and Sun, 2023). Although the literature outlines both constituent elements of AI capabilities and the potential value of AI-based DCs, it offers limited insight into what these capabilities mean in T&L contexts and how T&L organisations develop and mobilise them for strategic value. To explain these development processes and their contextual enactment, we draw on DCs theory and its microfoundations.

DCs theory explains how firms gain and sustain advantage in uncertain environments (Schilke et al., 2018; Steininger et al., 2022). Although there are varying interpretations of what DCs entail, the dominant view follows a procedural conceptualisation of sensing, seizing and transforming (Teece, 2007). The theory has been extended to specify digital DCs-a subset of a firm's overall DCs – where digital sensing leverages technologies to discover opportunities and threats via digital scouting, scenario planning and exploration; digital seizing orchestrates digital and non-digital resources to reconfigure processes and models; and digital transforming entails ongoing reconfiguration of structures and assets to adapt and thrive (Chatterjee et al., 2022; Steininger et al., 2022; Warner and Wäger, 2019). Firms that are adept in digital DCs can easily transform their business models, foster agility and enhance innovation and customer experience (Steininger et al., 2022; Warner and Wäger, 2019).

While digital DCs clarify what firms must do to adapt, their microfoundations specify how these capabilities are enacted (Warner and Wäger, 2019; Kroh et al., 2024; Chatterjee et al., 2022). Teece (2007, p. 1319) defines microfoundations as “distinct skills, processes, procedures, organisational structures, decision rules, and disciplines”. Subsequent work classifies them into people, structures and processes (Chen et al., 2023; Schilke et al., 2018), emphasising that both DCs and their microfoundations are highly contextual, shaped by specific technologies and organisational settings (Chatterjee et al., 2022; Mikalef et al., 2021). Accordingly, some studies emphasise socially oriented microfoundations, such as employees' skills, psychological attributes and alignment with digitalisation initiatives (Chatterjee et al., 2022), decision-making protocols, loyalty-building mechanisms, co-specialisation and organisational learning (Eriksson et al., 2022). Others examine digitally oriented microfoundations that enable sensing, seizing and transforming through digital focus, innovation processes, mindset, networks, technology capabilities and data management (Kroh et al., 2024, p. 1). Furthermore, AI-specific DCs have also been explored in B2B marketing (Mikalef et al., 2021) and manufacturing (Fosso-Wamba et al., 2024a, b). However, two gaps remain. First, most studies examine settings with strong digital foundations and digital platforms (Shen and Sun, 2023; Mikalef et al., 2021), offering limited understanding of how microfoundations and AI-based DCs function in asset-intensive T&L organisations characterised by legacy systems and skill shortages. Second, while scholars acknowledge the need for sociotechnical integration, the technical microfoundations of AI-based DCs and their interactions with social elements remain underspecified. Table 1 synthesises the theoretical framing and its implications for our study, highlighting the gaps that inform our research questions.

To address the research questions, we employed a multi-phase, sequential, qualitative methodology (Creswell and Poth, 2023), combining a systematic review of grey literature in Phase 1 with key informant interviews in Phase 2 (see Figure 1). The primary unit of analysis is the T&L organisation. Data were collected from organisational documents and managers as key informants, but all coding, interpretation and theorisation were conducted at the organisational level. Individuals and AI initiatives informed organisational capabilities rather than serving as independent units of analysis.

We follow a sequential design, well established in qualitative research (Creswell and Poth, 2023), as it provides the ability to generate a comprehensive and rigorous understanding of an emerging phenomenon. Because AI adoption in T&L is rapidly evolving and only sparsely documented in academic sources, grey literature provides early, practice-based evidence of AI applications, organisational responses and emerging capability configurations. At the same time, its limitations, such as self-reporting bias and potential outdatedness (Adams et al., 2017; Zhang et al., 2021), necessitate further empirical validation. The sequential design therefore enables Phase 2 interviews to corroborate, refine or contest Phase 1 findings, mitigating grey-literature limitations while supplying in-depth, context-specific insights into organisational AI use and the sociotechnical mechanisms underlying capability development.

Our design also involves a theory elaboration approach (Fisher and Aguinis, 2017), in which existing constructs guide the analysis while remaining open to empirical extension. Using the seed concepts of sensing, seizing and transforming and the sociotechnical microfoundations of people, process, structure and technology, we applied a common analytical framework to both phases. Phase 1 involved abductive analysis of grey literature to derive initial AI-based DCs, their microfoundations and business value mechanisms. Phase 2 employed a deductive orientation, examining interview data in light of these preliminary insights to confirm, enrich and contextualise them. This sequential structure generated a multivocal understanding (Zhang et al., 2021), enabled triangulation across data types Bogner et al. (2021) and supported data saturation (Fusch and Ness, 2015). Overall, the two-phase design ensured a progression from generic to empirically refined theorisation of how T&L organisations develop the sociotechnical microfoundations underlying AI-based dynamic capabilities and value creation.

Grey literature refers to materials not “published in a traditional format” and “lacking bibliographic control” (Levin in Adams et al., 2017). It includes diverse sources such as press releases, blogs, web pages, organisational and government reports, videos, newspapers and emails (Adams et al., 2017; Zhang et al., 2021). A grey literature review was adopted in the first phase because it (1) helps to assess the impact and relevance of emerging technologies (Zhang et al., 2021), (2) offers timely reflections on the rapidly evolving use of AI in organisations (Bawack et al., 2021) and (3) provides practical insights without recall bias from stakeholders otherwise difficult to reach (Zhang et al., 2021; Molla and Biru, 2023). These sources were analysed to identify organisation-level AI assets, routines and capabilities, consistent with the study's organisational unit of analysis.

The evidence collection and screening followed the systematic literature review procedure described in Figure 2. It encompasses the identification, screening, eligibility and inclusion stages. Although presented in sequence, these steps were executed iteratively from June to August 2024.

We identified the top 20 logistics companies listed by the Australian Logistics Council (2024) and reviewed their publicly available AI grey literature. We identified five firms with traceable AI initiatives- DHL, Ceva Logistics, Toll Group, Australia Post and Linfox and excluded the rest. We then conducted an advanced search on Google and YouTube using the organisations' names and the term “Artificial Intelligence OR AI” delimiting our search to 2015–2024 to align with the pace of significant AI advancements. This produced over 406,000 Google results and 21,000 YouTube videos. Acknowledging the challenge of working with grey literature to sift through vast amounts of data (Molla and Biru, 2023), we refined our search following the norms established in prior literature (Zhang et al., 2021) to the first 30 results from each platform, totalling 150 documents and 103 videos.

To establish the quality and credibility of sources, we followed recommendations for grey literature screening and evaluated the authority and source credibility, corroboration of information, topic coverage, objectivity and publication date (Adams et al., 2017; Molla and Biru, 2023; Zhang et al., 2021). Accordingly, we first inspected the title and content of each source and eliminated 58 papers and 77 videos due to irrelevant content, duplicates, advertisements or technical issues such as broken links. We then meticulously assessed the remaining 118 items and removed sources (18 documents and 3 videos) that lacked sufficient organisational context, transparency regarding AI applications, corroborating evidence, substantive AI coverage or credible authorship. Ultimately, we selected 97 materials: 74 papers and 23 videos produced by recognised organisations, industry associations and established media platforms; contained concrete descriptions of AI initiatives, outcomes or implementation practices; and triangulated across multiple sources whenever possible. Table 2 summarises the evidence corpus.

The Gioia methodology (Gioia, 2021) was employed to analyse the data using NVivo 11. The process began with a thorough review of relevant documents, focusing on AI technologies, their features, integration processes, challenges and outcomes. One of the co-authors, with close supervision and guidance from other co-authors, conducted first-order coding using the seed concepts of microfoundations (people, process, structures and technology), DCs (sensing, seizing and transforming) and business outcomes. At this stage, theory elaboration occurred through inductive expansion by adding new codes which produced a total of 57 open codes (Figure 3). Subsequently, we entered an interpretive phase to develop second-order themes using the theory elaboration technique of refining conceptual definitions. This involved analysing the first-order codes for patterns and similarities aligned with our research questions and the theoretical framing. We grouped similar concepts from various organisations into broader theoretical themes, such as “AI Literacy”, which is linked to “workforce upskill” and “team expertise”. Higher-level aggregate dimensions were then derived, representing overarching patterns. During this process, we maintained weekly meetings to discuss and review coding decisions and emerging themes. Any disagreements regarding code interpretation, category assignment or theme development were resolved through discussion and consensus. Figure 3 illustrates the analytical structure.

The following findings describe how T&L organisations develop AI-based dynamic capabilities through specific sociotechnical microfoundations.

The findings show that T&L organisations use both generic AI such as chatbots, virtual assistants and machine learning and industry-specific AI, including AI embedded in long-haul trucks, autonomous guided vehicles in warehouses, robotic forklifts and product identifiers, collaborative robots and digital twins. The analysis also identified key AI-based DCs, enabling microfoundations and business value. These findings are summarised in Table 3.

T&L organisations leverage the inherent features of AI technologies, such as perception, comprehension, automation and augmentation, to develop capabilities for sensing market opportunities, responding effectively and transforming their business processes.

The five T&L companies are developing the capabilities of using AI to identify opportunities in the market, which Fosso-Wamba et al. (2024a, b) refer to as AI-enabled sensing. For example, DHL employs AI chatbots and virtual assistants to monitor customer behaviour, recognise data-driven changes and gain insights into user actions, which enhances their ability to sense customer needs. As one of DHL's Senior Vice Presidents noted:

DHL also harnesses interactive and generative AI capabilities to obtain user feedback by engaging in complex customer conversations. Similarly, Linfox leverages machine learning to assist its clients in understanding seasonal patterns among end-users. Australia Post uses AI's cognitive capabilities to analyse customer feedback continuously, enabling the company to provide tailored recommendations and delivery solutions. Australia Post's Machine Learning Specialist and Senior Developer shared:

For Ceva, the self-driving truck's ability to perceive its surroundings has improved community safety. In partnership with Kodiak, they have equipped their long-haul trucks with cameras that generate and process perception data. Linfox uses AI capabilities for advanced data analytics to harness its data pool and inform decision-making, ultimately driving better results.

In dynamic capabilities literature, seizing involves organisations mobilising their resources in response to changes in their environment (Fosso-Wamba et al., 2024a, b). The five T&L organisations have used AI to adapt to changing environments and maintain competitiveness. This includes optimising and improving decision-making, automating core business operations and enhancing various business activities. DHL implemented an AI “vision picking technology” to automate workflows and order management. This reduced task completion time and allowed employees to focus on value-added activities. Ceva adopted autonomous guided vehicles in its warehouses to enhance operations without human intervention. Likewise, Australia Post's automated robotic forklifts operate seamlessly, allowing them to restructure their underlying processes and reassign workers to more productive roles. Both DHL and Ceva deployed collaborative robots (cobots) to automate cart-picking tasks alongside humans in warehouses.

The case organisations have also harnessed the augmentation capabilities of AI to support staff and managers across various business processes. DHL employs AI to determine the most efficient routes within the warehouse and for last-mile delivery. This reconfiguration of resources optimises “workload distribution for warehouse staff and help prioritise more time-critical shipments” (Pohl, 2020). Furthermore, Ceva leverages AI integrated within digital twins to replicate its ground operations and optimise its operations in real time.

The findings indicate that AI technologies are transforming the warehousing and transport operations. For instance, AI has either replaced or enhanced human intelligence in distribution centres. Ceva Logistics and Toll Group utilise a Robotic Product Sortation and Identification (RPSI) system and robots, respectively, to autonomously sort parcels in their warehouses: “The RPSI system can handle challenging parcels such as polybags, tubes, padded mailers, and envelopes that are common in eCommerce shipping” (Moralez, 2021). This integration has resulted in seamless and efficient processes, with Toll Group achieving up to a 70% increase in productivity (Geek+, 2024). Similarly, Australia Post has enhanced its sorting and packaging business processes by incorporating robotic arms, improving accuracy and reducing accidents. Meanwhile, DHL's autonomous mobile robots optimised cart-picking activities and transformed complex processes that were traditionally performed manually.

Another important application of AI is in improved delivery planning and routing. Australia Post has deployed AI systems to enhance the accuracy of delivery predictions. DHL also utilises machine learning to predict defects and damages in delivery vehicles earlier, avoiding delays due to vehicle breakdowns.

In relation to the second research question, our results show that enhancing employees' AI literacy, improving organisational AI management processes and establishing AI governance are the most important microfoundations for enabling AI applications and capabilities.

AI literacy encompasses the essential skills, knowledge and training required to integrate and utilise AI technologies. This includes both the development of new skills and the realignment of existing skills to build AI capabilities. For instance, DHL's experience shows that “knowledge gaps may hinder the smooth adoption of interactive AI technology” (DHL, 2022). Similarly, in Ceva, “All staff pickers underwent a two-week training induction during the implementation” of autonomous robots (CEVA Logistics, 2023).

In addition to training, social interaction among individuals is vital for developing AI literacy at the group level. This is reflected in developing collaboration and shared expertise among different teams within T&L organisations. For example, Ceva created synergies between its logistics and technology teams to deploy AI: “This was a complex project with new types of data that had to be managed in real time, presenting a whole new experience in terms of design. Given these constraints, success can only be achieved if the best people work together towards a single shared objective, which is exactly what we had here” (Guilhaume Leroy-Méline and Rabaud, 2023). Toll Group “established multidisciplinary teams of data scientists, AI engineers, and business leaders to align what's needed with what's possible” (Nephy, 2023). These examples underscore the significance of team collaboration and social interaction in enhancing AI capabilities.

The five T&L organisations have adopted new routines and processes to respond to the implementation and integration of AI. For example, Linfox emphasised agility as a key factor in deploying AI capabilities. Similarly, in Australia Post, “The infinite capabilities of machine learning, combined with the agile methodology, allow Australia Post to continuously refine a product or service quickly, using relevant data provided by end users” (Fisher, 2017).

Another crucial aspect of effectively leveraging AI involves forming partnerships. Linfox implemented secure enterprise data warehousing and insightful analytics across the organisation by selecting cloud-based solutions from tech partners like Snowflake and Microsoft. DHL has collaborated with Dorabot, an AI-driven robotics solutions provider, to introduce the DHL Bot. This initiative followed a successful pilot program that significantly enhanced parcel sorting productivity.

AI-enabling governance, notably organisational structure and culture, is critical in deploying AI systems. The case organisations are actively working to cultivate a culture that supports AI utilisation while addressing implementation challenges, including “resistance from regulatory bodies and workforces affected by automation”. Building trust among workers regarding the outcomes and recommendations provided by AI systems is a vital aspect of governance.

This approach ensures that human oversight remains a central element in both AI training and decision-making processes. To effectively govern AI deployment, Toll Group, DHL and Ceva have also re-evaluated their governance structures to prioritise IT transformation and integration and effectively guide the development and implementation of AI capabilities.

The five T&L organisations have reported advantages from the AI-enabling and AI-based capabilities, including improved operational and environmental sustainability outcomes.

In terms of operational performance, the integration of AI has notably enhanced innovation, customer experience and overall efficiency. For instance, Australia Post “create new products for customers that would not have been possible otherwise” (Fisher, 2017). Additionally, these organisations have successfully improved the quality of existing products. While DHL's AI capabilities “improve the customer experience and increase the first-time delivery rate” (DHL, 2023), Linfox's led to “confidently design optimal customer solutions that can be onboarded rapidly and efficiently” (Harvey, 2021). Furthermore, the five T&L organisations have experienced substantial improvements in efficiency. For example, Toll Group's AI capability has resulted in significant cost savings associated with time and labour. Similarly, Ceva's AI capabilities have reportedly led to a 400% increase in operational speed (CEVA, 2023). DHL achieved 99% sorting accuracy, minimising secondary sorting and saving time and labour.

The T&L organisations have also reported sustainability outcomes due to fuel efficiency and reduced waste and emissions. DHL's AI capabilities minimise wastage by “reducing outer packaging made of cardboard by up to 50%, thereby significantly lowering CO₂ emissions per shipment”. Toll Group employed AI to decide the minimal carton size to fit a package, which minimises material waste and environmental impact. DHL optimised “vehicles' travel time per stop, thus reducing fuel consumption relative to the volume managed”.

A similar trend has been observed across Ceva, Australia Post and Toll as they utilise route optimisation platforms to develop the most efficient routing plans, effectively reducing fuel consumption. AI has also contributed to improving worker and community safety. Ceva's collaboration with Kodiak on autonomous self-driving trucks has been pivotal in “improving road safety for everyone, ultimately saving lives and enhancing the quality of life for truck drivers” (Robotics, 2022). Additionally, while Australia Post's capabilities have significantly lowered strain-related accidents, Linfox has created a workplace environment where accidents are reduced, and pedestrian safety is prioritised.

This analysis of the grey literature indicates that T&L organisations are utilising both generic AI tools and industry-specific technologies, which provide them with unique affordances for perceiving, comprehending, acting and learning (Bawack et al., 2021). To harness these potentials and improve operational efficiency, customer experience and sustainability goals, these entities are integrating the technologies with socially oriented microfoundations to develop AI-based sensing, seizing and transforming capabilities (Fosso-Wamba et al., 2024a). Successful AI deployment requires attention to governance, organisational alignment and human factors and failure to do so may generate implementation problems and unintended outcomes. To triangulate these findings and enhance the study's validity, we conducted follow-up key informant interviews.

To triangulate and enhance the reliability of the Phase 1 findings, ten interviews, each lasting an average of an hour, were conducted with managers from seven T&L organisations in Phase 2. The 10 interviewees came from seven T&L companies, two of which were part of the five companies involved in Phase 1. Consistent with the organisational unit of analysis, managers were interviewed to get insight into their organisations' AI-related practices, decision processes and capability development. This, however, did not prevent reflections on similar developments in the wider sector.

Using purposive sampling, an established method in qualitative research (Creswell and Poth, 2023) and prior AI studies (Mikalef et al., 2021), 53 T&L managers with expertise and knowledge of AI initiatives in their respective organisations were recruited from various sources (e.g. LinkedIn, company websites, referrals), of which 10 consented to the interview. Researchers have differed on the adequate number of interview participants for qualitative research. While saturation is often used to justify sample size (Saunders and Townsend, 2018; Vasileiou et al., 2018; Fusch and Ness, 2015), its meaning and appropriateness are debated. Some argue that saturation alone may not establish sufficiency (Saunders and Townsend, 2018) and factors such as study scope, topic complexity, data quality and design should also be considered (Morse in Vasileiou et al., 2018). In some cases, even one participant may suffice (Saunders and Townsend, 2018).

In this study, since the interviews were intended to triangulate the grey literature and corroborate the findings from Phase 1, rather than develop a new theory, we deem the number of interviews adequate for our purpose. Furthermore, we selected a relatively homogeneous group of managers experienced in AI, prioritising quality over quantity. According to Vasileiou et al. (2018), the adequacy of interview participants should conform to the study's nature. Table 4 provides a basic profile of the companies and interviewees ^[1].

Using the Phase 1 codebook as a deductive framework, we analysed the interview data by examining whether interviewees' accounts confirmed, contradicted or extended Phase 1 findings. We applied theory elaboration through conceptual specification (refining definitions) and relational structuring (explaining interactions and enabling conditions). For example, while the grey literature result identified “predictive maintenance” as an AI application, interviews revealed the specific microfoundations required: data integration capabilities across legacy systems, governance protocols and change management processes. We ensured consistency, reliability and internal validity by using a common theoretical framework across both analysis phases, maintaining a thorough coding audit trail and conducting regular team discussions for reviewing coding and analysis decisions and interpretations.

The interview results reinforce the findings from Phase 1 that T&L organisations have embedded AI components such as computer vision, machine learning (ML), deep learning (DL), natural language processing (NLP) and large language models (LLM) in various systems (Table 5). These components and systems afford T&L organisations the ability to perceive T&L assets and environments, enhance comprehension, automate repetitive tasks and generate bookings and other related activities (Li et al., 2025). However, all interviewees agreed that their companies are at the early stages of AI-experimentation. To use the words of one interviewee: “If I had to put a number on it, I'd sort of say we're probably 30% of what the art of the possible could be”. Another concurred and noted that “Amazon … make use of a lot of robotics when it comes to warehousing. Very few organisations have gone to that level. I would say up to 5%”.

As indicated in Table 5, interview participants mentioned several applications of AI in the seven companies. In Companies A, B, C and F, AI is applied in fleet operations. The systems allow real-time monitoring of journeys, activities, behaviours, asset characteristics (e.g. fuel, mileage, wear and tear) and potential risks such as road hazards and traffic. For instance, AI is used to monitor behaviour and awaken drivers during micro-sleep episodes, ensuring safety. Beyond driver behaviour, the geo-fencing and geospatial ability of AI is used for route planning and optimisation, which is paramount amongst all organisations.

AI bots are increasingly used in operations and customer relations. For example, Company A use a transactional bot, which relies on RPA and ML to monitor and book transport availability automatically. Company C has implemented conversational bots based on Generative AI to interact with customers and address their queries. Several companies, including A, D, E, F and G, use AI-enabled data and ERP platforms such as SAP HANA, Microsoft Dynamics 365 and Navision to enhance their analytics. For instance, SAP HANA has a Generative AI function that allows users to query data. These technologies are helping businesses to build AI models, automate their processes and develop predictive solutions for strategic planning.

Robotics is the least adopted AI application. Company A uses it in the warehouse for sorting, picking, truck loading and detecting staff dress code compliance. Likewise, Company E is using vision-picking technology for pallet picking and loading. Other technologies being trialled include ICR for automating administrative tasks, RPA for invoice processing, traffic management to prioritise trucks at traffic lights and SAP's HR systems for CV screening.

Consistent with our research questions and the purpose of the interviews to validate Phase 1 findings, we investigated how the adoption of AI technologies and systems, as discussed in the preceding section, enabled T&L companies to transform operations. There are both commonalities and differences among the companies in terms of how they leveraged AI to enable sensing, seizing and transforming.

Some of the T&L companies are using AI to enhance their market-sensing capability and identify opportunities through market scanning, competitor analysis and customer behaviour patterns. An interviewee from Company C noted: “We use AI to detect market changes by monitoring news, social media and other statistics sources. This allows and assists us in strategic planning towards meeting customer needs”. In Company A, there is an advanced use of AI to sense market opportunities and seize them autonomously: “I guess another solution we've implemented involves an automated booking system. Essentially, this system monitors transport availability in the industry. If another party cancels a particular transport booking, the system instantly secures it for [Company A], which is critical in a highly competitive market” [Interviewee A1].

In addition to sensing, the seven companies are increasingly making T&L processes optimisation, resource (such as vehicle utilisation) and planning decisions based on insights derived from AI. Although some view AI as “a critical enabler of digital transformation” (Interviewee F1), most interviewees emphasised that they “don't just give AI the 100%… [and] don't take the 100% response” [Interviewee C2]. Instead, they use it as a guide to build on what they already have with human judgment. For instance, the companies are applying machine learning to their own, clients and other environmental and traffic data to create a responsive and efficient delivery route: “… we look at area of focus, integrating AI-driven solution to optimise route planning, … we leverage on predictive analytics to anticipate logistics in terms of ensuring timely delivery, while also minimising costs” [Interviewee F1].

AI insights are also used to guide warehousing processes and decisions in Companies A, B, D, E and F. These decisions extend beyond the T&L companies into notifying their clients “about the trend in terms of their products that is kept in our warehousing” [Interviewee F1]. Interviewees D and E reinforce this: “we have the warehouse data where you keep the historical data when it comes to sales, production, and some other market analysis. So if you are to now look at it based on this trend, … it comes very fast” [Interviewee D1]. “We have a lot of huge data on all these 6,000 customers, on their history, … so we can proactively define certain routines for our customer services” [Interviewee E1].

The smartification of asset utilisation and maintenance is another capability the seven T&L companies are building at different levels of maturity. This ranges from managing to ensure “vehicles were no longer being overloaded or under loaded” (Company G), predicting the maintenance requirements of vehicles (Companies A, B, C, E, F), enabling “automated storage and retrieval system in the warehouse” (Company B), having “an eye” on trucks while they are on the road (Company A, B), monitoring driver behaviour (Companies A, C, E, F), to managing “the turnaround time of trucks” [Company C]. In all these decisions, the companies are using AI to build on “what [they] have got and optimise as much as [they] can” [Interviewee A2] and as a “guide how to plan” [Interviewee E]. This is a continuous process of learning and improvement, as noted by Company F's interviewee:

Regarding the AI-based transformation capability, the T&L companies are developing a relatively new value proposition to their clients based on AI-derived insights and operational transformation of workplace safety. As indicated earlier, Companies A, D, E and F have strategically harnessed client stock data from their warehouses, integrated this information with various market data and sophisticated machine learning algorithms and provided value-added services to their clients to make informed decisions, as suggested in the following quote:

How companies engage and ensure the safety of their workers, particularly drivers of long-haul trucks, is being transformed with the application of AI. Driver's work is very hard. Because of Australia's geography, some drive more than 1800 kilometres, very long hours and in very tough road conditions. The companies are deploying AI to monitor driver's behaviour and ensure their safety. This involves “implementing machine learning models in identifying high-risk drivers and patterns” (Company F), ensuring that drivers are not diverted from the nominated route (Company C) and ensuring drivers' safety (Company A). For instance, Company A combined telematics and AI to monitor the driver's blinking habits to manage micro-sleeps. If the driver blinks for too long, the seat vibrates gently to wake them, and a call is logged to the control room to check in. If the driver appears tired frequently, they are advised to visit a medical centre. This has led to discovering health conditions and reassigning drivers to shorter routes.

Different from the other six companies, Company A is also developing a capability to improve truck movement on roads. They have partnered with a major Australian company to give priority traffic signals for their trucks in regional cities. In this trial, if a truck approaches a green light that turns red, the light remains green to allow the truck to pass. This solution helps maintain smoother traffic flow. Company A also transformed one of its warehouses into a fully automated facility, integrating automated guided vehicles (AGVs) to move inventory: “… we have some warehouses in our network that are fully automated. Robots do the picking. And they're basically picking pallets, full pallets of product and basically getting it loaded onto trucks” [Interviewee A2].

Overall, the interview results corroborate the findings that T&L organisations are leveraging AI technologies and other resources to sense market and environmental trends and improve certain critical processes. They are also enhancing their ability to diversify their offerings and the microfoundations that facilitate the development of AI-dynamic capabilities. The next section presents the related findings.

We examined the microfoundations necessary for organisations to develop AI-based dynamic capabilities across the seven companies.

All companies seem to prioritise AI technical skills, including data analytics, machine learning and data intelligence for workforce training. Interviewee C1, for example, noted “it wasn't easy to get AI at first” due to the challenge in accessing AI expertise. Interviewee F1 explained, “we need skilled people, so we organise nearly six months of training focused on data science, data analytics and machine learning”. Ongoing post-training monitoring is crucial for validating knowledge, as highlighted by Interviewee D1, “After training, it's important to ensure understanding”. A recurring theme is the importance of partnerships with technology companies for tailored AI skills. Interviewee A2 emphasised this collaboration: “We work with key industry partners who specialise in AI, as its a rapidly evolving field”.

In addition to skills, organisational readiness and change management are crucial for building AI capabilities. Company A emphasised the need for slack resources for AI deployment, noting the high upfront costs of automation. Similarly, an interviewee from Company F stated, “We are well equipped, and the support of management in terms of investment and finances is crucial”. Interviewee C3 stressed the importance of a clearly defining AI needs, along with a business strategy and cost-benefit analysis to secure management support. Interviewee D discussed the importance of change management actions in navigating new AI systems, a point echoed by Interviewee F, who noted “We face resistance to change, but we address it through training programs to help employees understand AI as a tool for support rather than a replacement”.

To develop AI capabilities, the companies enhanced existing processes. Companies A and C adopted agile methodologies to support iterative feedback and customer service operations. Company D identified process inefficiencies for AI applications, such as automated order picking and packaging. Partnerships with vendors were deemed essential for AI integration, according to Company A, B and F interviewees. Additionally, AI and data governance play vital roles in this development. An interviewee from Company F highlighted the need for a data governance framework to ensure accuracy and accountability across departments, stating, “Data accuracy is something that is very, very critical. So, we put in place something we call a data governance framework in order to get clean data across various departments”.

In addition to data, the companies have implemented AI governance to guide development. An interviewee from Company A outlined, “We are not a technology company; we pursue AI only when it drives business value and ensures the company's safety”. AI ethics and privacy are equally critical, with dedicated teams ensuring compliance with regulations and standards: “Our team ensures that our AI initiatives align with regulatory guidelines, protecting the personal information of both our staff and clients as we analyse data” (Interviewee F). An interviewee from Company A pointed out the importance of governing AI outputs’ morality, noting, “The system may not recognise what is morally correct, so we implement rules to ensure its outputs stay within our moral boundaries”. This underscores the need for active management of AI's ethical implications.

The seven companies have reported environmental and operational benefits from AI capabilities. They have reduced waste and carbon emissions by optimising delivery routes, minimising the frequent stop-and-start cycles of trucks (a major contributor to emissions), and implementing preventive maintenance to reduce gas leaks and harmful pollutants. Moreover, as highlighted by interviewee C2, integrating AI with safety controls enhanced drivers' safety: “Logistics and safety work hand-in-hand. AI has helped us to ensure safety, protect lives, and ensure happiness. Therefore, safety is a key benefit of using AI”.

Additionally, the companies have boosted operational efficiency, reduced errors and improved on-time deliveries. Proactive maintenance has cut truck downtime and repair costs, while route optimisation has lowered fuel expenses. Companies A and F reported greater market responsiveness, increased client satisfaction and higher revenue, with one interviewee stating, “AI has transformed how my company engage with our customers. AI-powered chatbots offer round-the-clock assistance, from tracking shipments to handling returns, thereby improving response times and boosting customer satisfaction” [Interviewee F1].

The interview findings corroborate the insights drawn from Phase 1, providing additional evidence of the AI technologies being implemented and the AI-based capabilities in sensing, seizing and transforming business processes. Similarly, the use of AI for sensing freight operations, network tracking and orders was enabled by the availability of data and analytical reports.

Key factors that enable AI capabilities include enhancing employee AI literacy, improving organisational processes and establishing robust data and AI governance. These elements contribute to greater operational efficiency, safety and environmental performance. In some cases, they also lead to strategic benefits, such as new sources of revenue and improved client retention. Although the primary focus of this study is AI value creation rather than AI harms, the importance of governance, sociotechnical alignment, data quality and ethical oversight in AI deployment points to potential unintended consequences, including biased outputs, implementation failures and loss of reputation, which warrant further investigation in future research.

AI is increasingly viewed as a transformative force in T&L, offering the potential for improved operational efficiency, innovation, resilience and sustainable outcomes (Koreis et al., 2025; Garg et al., 2025; Shen and Sun, 2023; Richey et al., 2023). Although AI technologies like machine learning, computer vision and large language models present significant opportunities (Bawack et al., 2021; Li et al., 2025; Schuetz and Venkatesh, 2020), their benefits are frequently not fully realised. Based on the empirical results, we contribute a process theory (Figure 4) explaining how organisations develop AI-based DCs through sociotechnical microfoundations to generate business value.

The microfoundations have gained traction following calls for research to address the sources of dynamic capabilities (Chen et al., 2023), with scholars in the field defining the concept with a focus on individual, organisational and process-level elements that enable firms to sense, seize and transform their capabilities (Chen et al., 2023). With the advent of digital transformation, these socially oriented microfoundations have been expanded to incorporate technological elements as sources of DCs (Kroh et al., 2024; Steininger et al., 2022; Warner and Wäger, 2019). In this study, to fully account for the development of AI-based DCs, we consider sociotechnical microfoundations encompassing AI assets, individual/people, organisational and process-level elements.

AI-based DCs require a mature digital architecture, which, in addition to existing digital and data analytics platforms integral to developing digital innovation capabilities (Kroh et al., 2024; Warner and Wäger, 2019), requires creating a range of AI-specific assets, which include systems, technologies and data (Bawack et al., 2021; Schuetz and Venkatesh, 2020). In terms of systems and technologies, T&L and other organisations are adopting AI technologies such as NLP, ML, sensors and computer vision, both as AI-native systems and by integrating them into existing systems (for example, in telematics systems that monitor driver behaviour) (arrow i in Figure 4). The adoption of AI systems and technologies enhances an organisations' digital assets and resources as well as its capacity for perception, comprehension, action and learning (Bawack et al., 2021; Schuetz and Venkatesh, 2020). Untapping this capacity requires appropriate data and infrastructure to deploy the systems and the ability for the system to perform cognitive tasks, support decision-making, augment users and be autonomous (arrows ii). The availability of accurate and high-quality data is essential to training AI systems and driving reasonable AI outcomes (Black et al., 2024). Data are also essential to create the AI capacity to perform cognitive functions such as learning, comprehending, acting and perceiving (arrows iii).

In the process of developing AI assets, T&L organisations also learn to enhance the social microfoundations of AI-based DCs, including AI literacy, processes, routines and governance (arrow a), which is in line with a mutually shaping interaction between the technical and social elements of microfoundations (An et al., 2024; Enholm et al., 2022; Kroh et al., 2024). As an organisation's AI journey is evolving, continuous training and upskilling of the workforce and uplifting of AI governance routines are essential. AI technologies, data infrastructure and skilled human capital are sources of dynamic capabilities and competitive advantage (An et al., 2024; Chatterjee et al., 2022). Organisations need to accumulate these necessary AI assets (data platforms, models and people) and integrate them with other organisational processes that are vital in building the right capabilities to derive value. This is supported by evidence from marketing (Mikalef et al., 2021), manufacturing and supply chain and logistics industries (Fosso-Wamba et al., 2024a, b; Mishra et al., 2024). Hence, the AI assets are critical microfoundations to enact AI-based sensing, seizing and transforming (arrow b). Firms that use large datasets powered by AI can sense their environment quickly, make informed decisions and respond to market changes (El Bhilat et al., 2024). Furthermore, T&L organisations building AI assets, such as machine learning, have reported business value, including operational and environmental efficiency (arrow c). This has also been evident from prior literature on the impact of AI adoption on performance in manufacturing (Abou-Foul et al., 2023), marketing (Mikalef et al., 2021) and logistics (Mishra et al., 2024; Fatorachian and Kazemi, 2025). Thus, our first theoretical proposition is:

Social-oriented microfoundations of DCs refer to the individual, structure and process-level elements that enable firms to develop, deploy and reconfigure their dynamic capabilities (Chen et al., 2023). Likewise, based on our evidence, social-centric microfoundations of AI-based DCs operate at different levels. At the individual level, research highlights the importance of an appropriate skillset as vital for building AI capabilities (Mikalef et al., 2021). The ability to engage with AI will be driven by continuous learning and role-specific upskilling of employees. At the structure level, the need for appropriate governance structures is vital, as it guides how AI-based DCs and technical microfoundations are built (Mikalef et al., 2021; Black et al., 2024). At the process level, routines such as data acquisition and cleaning, model development and deployment workflows provide the operational path through which AI is embedded in day-to-day operational activities (Black et al., 2024).

Our empirical results indicate that the social elements of microfoundations, i.e. AI literacy, organisational structures and governance, not only serve as a source of AI-based DCs (Enholm et al., 2022; Mikalef et al., 2021) but also co-influence one another (arrows iv, v, vi). For instance, AI literacy and AI-enabling organisational processes impact AI-enabling governance, which is vital to AI capabilities development (Enholm et al., 2022; Mikalef et al., 2021). A workforce proficient in technical knowledge, industry-specific skills and managerial experience is essential for building, deploying and maintaining AI systems (Mikalef et al., 2021). Hence, social-centric microfoundations have “ripple effects” (Enholm et al., 2022) on the success of AI implementation (arrow d) and on AI-based DCs building and deployment (arrow e). T&L organisations' anticipation of needs, allocation of resources, collaboration with technology providers and formulation of data and AI governance frameworks have facilitated their effectiveness in leveraging AI to sense and interpret changes within their operational environment, seize opportunities and transform their business processes (arrow e). Governance and decision-making protocols provide the framework within which T&L companies integrate AI, aligning with their ethical standards and functions. Emerging studies have also highlighted how AI-specific microfoundations in B2B marketing (Mikalef et al., 2021) and the manufacturing industry (Fosso-Wamba et al., 2024a) have facilitated the development of AI capabilities. Thus, our second proposition is:

In contrast to operational capabilities, which are vital for a firm's daily activities, DCs are developed through the process of building and integrating new competencies (Teece, 2007). In order to harness AI's value, organisations must move beyond simply implementing AI technologies and instead develop and orchestrate the human, structure, process and technological microfoundations to develop essential capabilities to operate and compete in an increasingly AI dominated business ecosystem (Fosso-Wamba et al., 2024b; Abou-Foul et al., 2023). We define AI-based DCs, as a special subset of an organisation's DCs to embed AI in sensing–seizing–transforming routines. Our data reveal that the sociotechnical microfoundations serve as critical pathways for companies to develop AI-based DCs, enabling them to adapt to changing environments and deliver value. The AI-based sensing, seizing and transforming capabilities interact recursively. The ability to use AI for sensing and responding allows organisations to identify opportunities and threats while reconfiguring their resources and processes to stay competitive and sustain their advantages (arrow vii, ix) (Shen and Sun, 2023). Additionally, AI-based sensing and seizing help T&L organisations refine and transform their routine business processes (arrow viii).

AI becomes valuable when embedded in DCs rather than when used as a stand-alone tool (Fosso-Wamba et al., 2024b; Abou-Foul et al., 2023). For instance, when T&L companies leverage the entanglement of sociotechnical interactions, they identify potentials, reshape value creation opportunities and transform business processes. Together, these DCs contribute to reduced fuel consumption, lower emissions, decreased toxic waste and improved operational efficiency and worker safety (arrow f). Additionally, these advancements create strategic opportunities for generating new revenue streams and enhancing client loyalty (arrow f). These findings align with recent research in manufacturing organisations, which has shown that AI-based dynamic capabilities have a positive impact on environmental performance, ultimately leading to improved manufacturing performance (arrow x) (Fosso-Wamba et al., 2024a). Furthermore, studies indicate that various AI-based dynamic capabilities contribute to different aspects of organisational performance, including operational performance (Fosso-Wamba et al., 2024a), environmental performance (Fosso-Wamba et al., 2024b) and business model transformation (Black et al., 2024). Thus, we formulate the third proposition as follows:

In this study, by analysing how T&L companies can harness AI's value, we respond to the call for domain-specific AI empirical research (Richey et al., 2023; Pan and Nishant, 2023). We elaborate on the theory of DCs microfoundations to advance a process model for the development of AI-based DCs as critical value drivers. Furthermore, we provide a sociotechnical microfoundations perspective on the formation of these capabilities, making several contributions to theory and practice.

First, this study extends research on AI by highlighting their strategic relevance for capability development and sustainability performance, moving beyond either conceptual (Richey et al., 2023; Ooi et al., 2025; Pan and Nishant, 2023; Lee et al., 2023), experimental (Balster et al., 2020; Tu et al., 2021; Loske and Klumpp, 2021) or adoption-focused (Fatorachian and Kazemi, 2025; Mishra et al., 2024). Previous research has shown the potential technical capabilities of AI like machine learning, computer vision and large language models (Mikalef et al., 2021; Chatterjee et al., 2022). Although these AI technologies present significant opportunities for perceiving, learning and acting (Bawack et al., 2021; Li et al., 2025), their benefits are frequently not fully realized due to lack of strategic considerations. Our work enhances this understanding by further explaining how these foundational technical capacities are integrated into organizational capabilities to improve performance. Therefore, the study contributes to strategic perspectives on how firms can harness AI's value.

Second, we contribute to the literature on DCs, particularly studies examining how AI transforms and reconfigures organisational routines to sense, seize and transform in response to evolving market requirements while maintaining economic and environmental performance (Fosso-Wamba et al., 2024a, b; Mikalef and Gupta, 2021). Drawing on Fisher and Aguinis's (2017) concept of theory elaboration, we develop a model and a set of propositions for analysing the value of AI and the mechanisms through which this value is generated. Our model is “specifically linked to AI” (Black et al., 2024, p. 13) and extends DC theory by addressing the specific challenges organisations face in translating AI's potential into realised benefits. This elaboration involves conceptualising AI-based DCs as a subset of organisational DCs, identifying a set of sociotechnical microfoundations and linking these to organisational and environmental benefits through specific relationships, sequences, interacting elements and propositions. Prior studies on DC microfoundations have primarily emphasised social-oriented elements rooted in people, structure and processes (An et al., 2024; Chen et al., 2023). We extend this perspective by adopting a sociotechnical lens, identifying how social and technological microfoundations jointly underpin the development of AI-based DCs. Through these MFs, organisations learn, adapt and renew their capabilities to harness AI's benefits. This contribution addresses the lack of emphasis on sociotechnical elements in AI research, unpacks AI technologies, data and systems and specifies how these interact with social-centric microfoundations and AI-based sensing, seizing and transforming processes, explaining the AI-business value nomology. Furthermore, instead of treating DCs' microfoundations as orthogonal, we theorise recursive relationships that explain how organisations accelerate their AI journey. The propositions can be used to analyse potential organisational challenges explicitly related to AI skills, governance and structure for effectively integrating AI into sensing, seizing and transforming processes.

Third, our study also responds directly to the empirical gap in much of the T&L AI literature, which mostly enumerates what AI can optimise (routing, inventory and emissions reporting) without specifying the conditions that enable the deployment of AI capabilities to drive business value in practice. The empirical findings address the lack of domain-specific research on T&L (Richey et al., 2023; Pan and Nishant, 2023). They offer insights for researchers who might be interested in undertaking “horizontal or vertical contrasting” (Fisher and Aguinis, 2017, p. 444) of AI's value and value-generating mechanisms. Unlike studies developed primarily in manufacturing (Fosso-Wamba et al., 2024a) and marketing (Mikalef et al., 2021), our findings unravel T&L-specific technologies and practices such as fully automated warehouses, where robots perform tasks previously done by humans, highlighting the significant evolution in T&L operations. This example shows that these facilities may eliminate the need for human pickers by relying entirely on robotic systems to handle tasks like pallet picking and product loading. While this shift can be viewed positively in terms of enhancing the efficiency of logistics processes, reducing labour costs and minimizing human error, it nonetheless raises concerns about the future of work and the workforce.

Moreover, we contribute to AI studies in the Australian context. Considering the historically hesitant and slow approach of Australian T&L to technology uptake (Gekara et al., 2019), our findings reveal a broader application of AI technologies and shed light on how some Australian T&L companies are evolving into more advanced, AI-driven environments. The results show changing attitudes towards capability building by prioritising workforce training, refining organisational routines, forming strategic partnerships and enhancing data governance. Through these capabilities, Australian T&L companies are positioning themselves for a transition to an increasingly automated and data-centric industry landscape.

The study has implications for T&L managers. As AI diffusion accelerates, managers must move beyond operationally and technically focused considerations in their AI strategies and instead cultivate AI-based DCs that strategically integrate AI assets with AI literacy, AI-enabling processes and AI governance. Managers should therefore prioritise organisation-wide AI literacy programmes that are role-specific to enable non-technical employees to interact with AI assets and provide feedback for model improvement, thereby accelerating routinisation and assimilation across functions.

Managers should also endeavour to make their AI value chain explicit. The findings show that AI assets create better business value when they are routed through social oriented microfoundations and AI-based sensing, seizing and transforming routines. This means that AI investment and resource allocation decisions should balance technological, human, structural and process considerations rather than focusing solely on technology. To develop AI-based DCs, managers need to identify key AI requirements and develop an AI investment value realisation model. It also requires co-locating AI teams with operational units (transport, warehousing, last-mile, customer contact) and moving AI assets from satisfying narrow operational uses (e.g. routing, demand prediction) to strategically align with the sensing, seizing and transforming capabilities.

Despite the above contributions, the paper has limitations. A key limitation of this study is that the empirical evidence is drawn from a limited number of organisations, primarily large firms within the Australian T&L sector. While this focus enabled in-depth analysis of AI initiatives in asset-intensive environments, the findings may not fully capture the experiences of small and medium-sized enterprises, which face distinct resource constraints, operational dynamics and challenges. Future research could examine how AI-based DCs and their microfoundations manifest in smaller firms or in less capital-intensive logistics settings. In addition, although the transport sub-sector encompasses maritime, rail, road and aviation (Tay et al., 2021), our study covered only road freight. Given the differing regulatory environments, asset configurations and operational dynamics across sub-sectors, further research is needed to explore how AI-based dynamic capabilities develop and produce value in maritime ports, rail logistics or air freight contexts.

While our findings focus primarily on the capability formation and the mechanisms through which organisations create value from AI, AI adoption may also generate unintended consequences (Enholm et al., 2022). For example, increased automation can alter workforce roles and potentially displace certain tasks; algorithmic bias may affect decision quality and fairness; while extensive data collection may raise surveillance and privacy concerns (Enholm et al., 2022; Lee et al., 2023; Ooi et al., 2025). Similarly, reliance on proprietary AI platforms and external technology partners may create forms of vendor dependence. Although such issues emerged only indirectly in our data through themes of workforce resistance, trust-building, governance and ethical oversight, they deserve more focused investigation. Future research could examine the tensions between AI-enabled capability development and these potential social and organisational risks.

Furthermore, the interviews were purposive, meaning that participants may have had a more positive experience with AI than a broader sample. The insights gained from our study may reflect a more optimistic view of AI's role in T&L and downplay concerns or negative experiences. Therefore, researchers should approach the findings with caution. Finally, this study advances a conceptual model and propositions that have not yet been empirically tested. Future research could operationalise the proposed constructs and relationships using quantitative or mixed-method approaches. Such research would enable validation of the model across industries and national contexts.

This study examined how organisations harness AI to generate business value by adopting a dynamic capability and sociotechnical microfoundations perspective. Drawing on qualitative evidence from T&L organisations, it shows that AI's business value does not stem from technology deployment alone. Instead, value emerges when AI technologies are embedded within organisational sensing, seizing and transforming routines, supported by appropriate skills, processes and governance mechanisms.

The study contributes a process model that clarifies the mechanisms, i.e. AI-enabling sociotechnical microfoundations and AI-based DCs, through which organisations translate AI investments into operational and environmental outcomes. In doing so, it shifts the conversation on AI's business value away from adoption towards the organisational conditions under which AI becomes strategically consequential. Overall, the study positions AI not as a standalone technological solution, but as a capability-building endeavour, offering a grounded explanation of how firms can move from AI experimentation to sustained value realisation.

During the preparation of this work, the authors used Grammarly and Copilot to assist in improving the readability and grammatical soundness of the manuscript. In all cases, the authors retained full intellectual responsibility and reviewed and verified all outputs.