Artificial intelligence in forward supply chains: insights from fashion industry Open Access

Interesting because: The fashion industry has historically relied on manual processes and has been slower to embrace digitalization compared to high-tech sectors. This study explores how fashion companies adopt AI across their supply chain processes. Research on this topic remains fragmented and often focuses on isolated technologies. This paper addresses this gap by offering a holistic view of AI adoption within Italian FSC.

Theoretical value: The study reveals that AI adoption in the FSC is driven by the pursuit of operational efficiency and advanced analytics. It is also hindered by technological constraints and cultural resistance. The research identifies seven distinct themes shaping this integration. A notable finding is the environmental paradox (the technology is perceived as a tool to advance sustainability goals, yet its high energy and water consumption is recognized as a significant environmental barrier). The analysis demonstrates a divergence in perspectives. It indicates that staff members focus primarily on operational stability. Management priorities center on strategic market responsiveness.

Practical value: Fashion companies should recognize that effective AI adoption requires more than technological investment. Managers need to establish organizational readiness by implementing role-specific training programs to bridge digital skill gaps. Firms must develop formal governance frameworks to ensure responsible use and address data privacy concerns. To mitigate cultural resistance and preserve the value of human expertise, leaders should cultivate a collaborative environment and frame it as an augmentation of human capabilities rather than a substitute.

Artificial Intelligence (AI) is a key paradigm of Industry 4.0, transforming how companies collect, process, and use data. In Supply Chains (SC), AI enables the rapid analysis of large volumes of information, helping companies make better decisions, respond quickly to market changes, and plan more effectively (Dubey et al., 2020; Toorajipour et al., 2021). AI applications span a wide range of SC processes, from demand forecasting and inventory planning to sourcing optimization and SC integration (Ivanov, 2023; Kusiak, 2018).

While AI has been widely adopted in high-tech industries, it has also begun to transform more traditional sectors, such as fashion, which have historically relied on manual processes and been slower to embrace digitalization (Fani et al., 2023). The fashion industry is characterized by fast-changing trends, short product lifecycles, and unpredictable consumer behavior (Bindi et al., 2021). These challenges demand smarter approaches to manage operations, relying on advanced technologies to optimize SC performance (Fani et al., 2025; Lamees and Ramayah, 2025). In recent years, advances in generative AI, combined with greater access to large-scale datasets, have enabled fashion companies to align with consumer preferences (Lee and Kim, 2024).

The fashion industry has been criticized globally for its environmental and social impacts along the SC, which makes the integration of AI particularly relevant under the Triple Bottom Line (TBL) perspective. Fashion is responsible for between 3% and 8% of total greenhouse gas emissions, and its SC is the third biggest polluter on the globe (John and Mishra, 2023; McKinsey, 2024), and generates about 20% of global water pollution (Ahmed et al., 2025; Muntean et al., 2022). At the same time, fast fashion has been associated with severe social issues, including child labor, unsafe working conditions, forced labor, and discriminatory practices (Huang et al., 2025).

In this context, emerging research demonstrates that AI can both support and challenge sustainability efforts in the Fashion SC (FSC). Environmentally, AI contributes to reducing waste, increasing recyclability, and enabling circular economy practices (Sinha et al., 2024). Socially, AI can enhance transparency, identify labor risks, and support responsible consumption (Karimova et al., 2025).

The implementation of AI in SCs also opens opportunities for new business models. Ventura and Meirelles (2025) propose a typology for structuring business models in Industry 4.0, emphasizing how advanced technologies such as AI reshape value creation, configuration, and appropriation processes. In this context, initiatives like Industry 5.0 further integrate human-centric approaches with digital technologies. Fani et al. (2024) show how Lean practices focusing on employee empowerment and engagement, when supported by technologies such as cloud manufacturing and business data analytics, can foster a Lean 5.0 paradigm in sectors like Italian fashion, reducing errors and boosting employee involvement.

Despite these advances, there are still significant barriers to AI adoption. Khin and Kee (2022) identified key drivers, enablers, and barriers influencing the adoption of Industry 4.0 technologies in manufacturing, including expected benefits, customer demands, resource availability, technical challenges, and workforce readiness. Moreover, while tools like ChatGPT have shown potential benefits in SCs, challenges remain related to accuracy, maturity, and the irreplaceable role of human expertise (Haddud, 2024).

While the fashion industry is often associated with creativity, it is fundamentally a manufacturing-intensive sector characterized by complex global production networks, rapid product cycles, and increasing technological integration (Quadras et al., 2025a). The adoption of AI within these networks directly affects core manufacturing processes such as production planning, process automation, and quality control (Quadras et al., 2025b).

However, although academic and practical interest in AI is growing, research on its application in the FSC remains fragmented, often focusing on isolated processes or technologies without providing a holistic view of adoption dynamics (Hossain et al., 2024; van Hoek, 2024). This lack of integration underscores the need for studies that systematically explore the purposes, drivers, barriers, and enablers of AI implementation in this context.

To address this gap, the present study employs a Grounded Theory (GT) approach. GT was chosen for its methodological rigor and its ability to move beyond descriptive findings commonly generated by interviews or case studies. GT supports the construction of robust, theoretically grounded explanations, enhancing both credibility and generalizability (Gioia et al., 2013). Based on this framework, this study investigates the following research question (RQ): How is AI being incorporated into the FSC operations, and which organizational and contextual mechanisms influence this process?

This paper is structured as follows. Section 2 presents the theoretical background regarding the AI applications within the FSC. Section 3 outlines the methodological procedures detailing the steps of the GT. Section 4 shows the results and the developed theory. Section 5 discusses the findings. Lastly, Section 6 concludes the paper with suggestions for further development.

AI is reshaping different echelons of SCs. Specifically, in fashion literature, AI stands out as a support for design activities. The main purpose in developing AI-based solutions within fashion design departments is to generate new content, trends, and patterns aligned with consumer expectations. Companies can generate popular synthetic images for the fashion industry based on user preferences (Nezhad et al., 2023). Literature shows that it is possible to improve diversity (Nezhad et al., 2023), capture consumer needs, select appropriate fabrics, and improve the efficiency of selection (Qiu and Ma, 2021). Also, AI can address the limitations of traditional pattern making and hand-drawn designs, generating clothing styles with high resolution and better quality (Ke and Wang, 2024).

Different cultures and anthropometric parameters can be addressed by AI, as it can help in the development of specific models that respect traditions. Such systems can adapt to unique body characteristics, optimize material selection by balancing aesthetics and comfort, and create diverse and trendy designs enriching the distinctiveness of ethnic apparel (Deng et al., 2023). Also, it can design modern patterns while respecting the history and the culture of a community and anticipating global trends (Jung and Suh, 2023).

The production process may use AI to drive the process optimization. Smart approaches can lead companies to the next step by improving the inner material characteristics, and accurately represent with high accuracy the mechanical properties (Youn et al., 2024). For material development, it can be used to predict the breaking strength of cotton and blended woven fabrics (Berkhan Kastaci et al., 2024), footwear slip resistance (Lau et al., 2025), and optimal dye recipes (Souissi et al., 2024).

Indeed, the production line could be entirely reshaped to implement new automatic processes. Computer vision can be used to automate the top-stitching process in garment manufacturing, detecting seam lines, generating stitching paths, and performing automated sewing, eliminating the need for human intervention (Ku et al., 2023). It can also automate grasping operations, reducing hardware investment and lowering labor costs (Chiu and Yang, 2024).

In sales, forecasting can leverage the application of neural networks to sales data, selling price, holiday periods, and seasonal sequences to increase the efficiency of sales prediction, leading to better planning (Kizgin et al., 2025). Within this context, physical attributes such as color, material, style, sales, expectations, size, store, trend, family, weather, promotional activities, and the opinions of domain experts can be used to train agents and achieve better results (Negre et al., 2024). AI stands out for customized suggestions for online shopping based on different types of products, taking into consideration different parameters such as color, consumer preferences, and anthropometric data, which is also integrated with consumer feedback. (Deldjoo et al., 2024; Hamad et al., 2023).

The integration of AI in the fashion industry has revolutionized how brands interact with customers. Chatbots in the fashion industry are AI-powered conversational agents used to enhance customer experience. These chatbots can take several forms, including text-based, voice-based, and embodied agents (Manzo et al., 2024). Text-based chatbots interact with users through messaging platforms, while voice-based agents use speech recognition. Embodied agents are virtual characters that interact more realistically with users. Chatbot systems can improve response times, reduce human effort, and maintain high accuracy in handling customer queries (Ngai et al., 2021). They are also able to provide personalized recommendations and create immersive brand experiences (Joy et al., 2022).

AI evaluates clothing compatibility to suggest cohesive outfits (Balim and Ozkan, 2024). Alongside this, virtual try-on tools allow consumers to visualize garments without physical trials. Such systems generate realistic images by transferring clothing appearances between pictures (Ghodhbani et al., 2023). They achieve this through semantic segmentation (isolating body and clothing regions) to enable precise image manipulation (Liu et al., 2024). As a result, users can digitally experiment with new styles and colors to assess compatibility with their existing wardrobe (Jeong and Sohn, 2020).

Integrating AI into the fashion SC advances sustainability objectives (Quadras et al., 2024). Intelligent systems optimize energy use and material flows to lower greenhouse gas emissions and reduce manufacturing waste (Wu et al., 2025). Data analysis guides the design of products with lower environmental burdens and supports networks focused on recycling (John and Mishra, 2023). Technology also improves waste management by predicting generation rates and refining collection routes (Badran et al., 2025). Within reverse SCs, real-time sensor data helps identify opportunities for material recovery and circular economy practices (Ali et al., 2024). Automated systems enhance recycling operations by reducing human error and increasing the recovery of valuable textiles (Jaczynska and Mehta, 2025).

This study employed a GT approach to investigate how fashion companies apply AI across the SC. GT is a technique used to generate or discover theory “grounded” in the observed data, aiming to identify emerging concepts and their interconnections through analysis, interpretation, and iteration (Schumm and Niehm, 2024; Kumar et al., 2026). This methodology is divided into two steps: (1) data collection and (2) data analysis and coding.

The study collected data through four rounds of semi-structured interviews conducted between July 2024 and January 2026 with professionals from eleven fashion companies in Italy. Participants included managers (44%) and team members (56%) from different departments: Production (30%), Design (15%), IT (15%), Sales (15%), Finance (11%), Logistics (11%), and Marketing (4%). The interviews aimed to explore company structures, business models, and the role of AI in the FSC. The first round focused on understanding the company context and identifying perspectives and concerns related to AI. In the second round, participants were asked to validate and expand on the initial findings. The third round followed specific AI implementations in progress within selected companies. The final round aimed to explore how the participants perceive the impacts of AI on the company waste management and sustainability. Interviews lasted between 45 and 60 minutes and were conducted in person or via video calls. All interviews were recorded and transcribed for analysis. Table 1 presents the list of interviewees.

The data analysis followed an iterative and systematic approach, conducted progressively after each round of interviews and adapted the procedures proposed by (Corbin and Strauss, 1990; Gioia et al., 2013). Data collection consisted of four rounds of semi-structured interviews. Following each round, textual data were analyzed progressively. During the open coding phase, transcripts were divided into segments and labeled with preliminary codes reflecting emerging concepts, categories, and sub-categories, maintaining an inductive stance that allowed new insights to emerge. Axial coding was subsequently performed to explore relationships between codes and to group them into broader second-order categories, with particular attention to recurrent and dominant themes. In the coding phase, these categories were distilled into overarching themes, forming the foundation of a coherent conceptual framework.

In the first round, all participants provided a general overview of AI adoption in their organizations, following a structured interview protocol (available in Table A1 in  Appendix). It is important to highlight that the questions used in the interviews were intentionally left open so that participants could freely express their thoughts and perspectives, without any suggestions or information that might influence their answers. These interviews were transcribed and analyzed using open coding, with three researchers independently coding the data. This process of triangulation served as an internal validation mechanism, ensuring reliability and reducing individual bias in the identification of preliminary codes. In the second round, a single open-ended question was posed, focusing specifically on ongoing or planned AI projects, and was applied only to the two companies that had already implemented AI initiatives. The third round involved external validation with the participating companies, providing the opportunity to verify, discuss, and adjust the categories and interpretations derived from previous rounds. In the fourth round, specific questions were introduced to assess the relationship between AI and sustainability. This phase investigated the role of technology in supporting environmental goals to manage both digital and green transitions (Table A1).

This study has revealed seven distinct themes: (I) Data-Driven, (II) Technology, (III) Trust, (IV) Culture, (V) Support Structures, (VI) Market Dynamics, and (VII) Environment. These categories are described in the following sub-sections. Interview excerpts, presented in italics, are used to illustrate participants' perspectives.

Across the organizations interviewed, the interest in AI is strongly connected to its potential for improving decision-making, forecasting, and operational performance. Managers and staff see AI as a valuable tool for optimizing how data is used across departments. The coding process for the category Data-Driven is presented in Figure 1.

The category Data-Driven reflects how the systematic use of data can shape both operational and strategic practices. Axial coding revealed several dimensions that illustrate the pervasive role of data in guiding decisions across an industry defined by short product lifecycles, fast-changing trends, and high demand volatility. Forecasting and planning practices rely on predictive models that process past sell-out performance, consumer browsing behavior, seasonal patterns, and trend signals from social media to generate more accurate projections for upcoming collections. This is particularly critical in fashion, where errors in demand estimation quickly translate into leftover stock and markdowns.

This connects directly to inventory and resource optimization, where data enables precise allocation of stock and materials, mitigating risks of overproduction and shortage. By modeling color-size ratios, identifying slow-moving SKUs, and anticipating replenishment needs, companies reduce the risks of overproduction and stockouts. A logistics professional emphasized the expectation that “It is expected that AI can support production planning and help prevent under- or over-stocking” (Logistics Staff at Company 7), while another highlighted the desire to “optimize stock levels” and “reduce excess inventory” (Logistics Staff at Company 1), reflecting the sector's pressure to align supply more precisely with rapidly shifting consumer demand.

The analytical potential of data also extends to sales and performance monitoring, allowing companies to identify consumer behavior patterns, evaluate market responsiveness, and refine pricing and promotional strategies. This responsiveness is central in markets where trends can peak and decline within weeks. From a strategic perspective, leaders see data as essential for shifting focus from repetitive reporting tasks to higher-level decision-making. As one finance manager explained, “More insightful financial analysis would enable a stronger focus on strategy rather than on repetitive tasks” (Finance Manager at Company 5). Enhanced consumer insight and personalization, supported by richer datasets, are seen as key outcomes of advanced data practices, enabling brands to tailor recommendations, improve product–market fit, and increase the success rate of new designs and capsule collections.

The category Technology reflects the level of infrastructural readiness organizations exhibit in integrating AI into their operations. Figure 2 illustrates the coding process.

The category Technology highlights the structural and infrastructural conditions necessary for effective AI deployment. Axial coding revealed that integration with legacy systems and existing operational practices represents one of the most persistent challenges. These obstacles are intensified by fragmented technological ecosystems across brands, suppliers, manufacturers, and logistics partners, where varying levels of digital maturity limit interoperability and restrict the flow of product and data. The effectiveness of AI is further constrained by inconsistencies in information and by limited traceability data from upstream suppliers, both of which reduce the reliability of AI-generated insights.

A recurring theme concerns the technical complexity of integration, often described as a significant barrier to adoption. As one IT staff member observed, “Integrating AI with existing systems is complex and often disrupts established routines, as it is not a plug-and-play process” (IT Staff at Company 1). This complexity is amplified by the sector's reliance on multiple seasonal calendars, capsule collections, and rapid design-to-store cycles, which require tight alignment between digital tools and established workflows. Beyond integration, the usability of tools was frequently mentioned, particularly by non-technical staff, who emphasized that many AI applications are not sufficiently user-friendly and require significant adaptation. A marketing manager underlined this point: “There is a need for user-friendly AI-powered tools to support data analysis and recommendations” (Marketing Manager at Company 6). Given the creative and cross-functional nature of fashion work, steep learning curves and training demands further slowdown and effective adoption.

Infrastructure requirements also emerged as a central concern. Fashion firms rely on a constant flow of data from stores, e-commerce platforms, manufacturing partners, and trend-monitoring tools, which intensifies the need for robust and scalable systems. IT professionals emphasized that AI initiatives require adequate computing capacity, stable integrations, and uninterrupted data pipelines to process high-frequency sales data, design files, and production schedules. As one IT manager explained, “it's necessary to ensure the AI systems have the necessary computing power and that the data flows correctly” (IT Manager at Company 5). This emphasis connects directly to the broader issue of technological readiness, where upgrading existing infrastructure and bearing the related costs are viewed as essential yet challenging prerequisites for the widespread adoption of AI within the FSC.

The category Trust captures the psychological and ethical hesitations that permeate the adoption of AI within FSC. Figure 3 shows the coding process.

The category Trust underscores the psychological, ethical, and organizational uncertainties surrounding AI adoption. Axial coding revealed that confidence in algorithmic systems is undermined by three main factors: the perceived unreliability and opacity of outputs, concerns over ethics and data governance, and anxieties about workforce disruption. These issues are particularly relevant in fashion, where decisions depend on tacit knowledge, aesthetic judgment, and close coordination across design, merchandising, and production teams.

The lack of transparency in decision-making was mentioned as a barrier. A discomfort with recommendations was described to feel arbitrary. As one design manager stated, “I don't understand how the AI decides” (Design Manager at Company 4), while production staff added, “AI results are manually double-checked because they are not yet fully trusted” (Production Staff at Company 8). This sense of algorithmic unclarity limits reliance on AI, as outputs often appear detached from the specific context in which decisions are made.

Second, concerns over bias, fairness, and accountability amplify distrust. The risk that algorithms could reinforce discrimination or distort customer insights if trained on biased datasets also emerged: “AI can reinforce bias and prejudice behaviors” (Design Staff at Company 3). A logistics staff member stressed, “It may affect fairness in decisions, especially when the details of its training process are not transparent” (Logistics Staff at Company 9).

Privacy also emerged as a pressing issue, with several respondents questioning how sensitive data would be used and safeguarded. As a finance manager asked, “What about data privacy? It's hard to know how secure the data really is because there's not much transparency” (Finance Manager at Company 5). The absence of clear organizational guidelines and robust compliance measures intensifies these ethical and governance concerns.

Finally, job insecurity and displacement anxiety were frequently reported, reflecting fears that automation might gradually reduce professional stability. Logistics and sales staff expressed worries that “AI might replace them” (Logistics Staff at Company 1) or “make some roles redundant” (Sales Staff at Company 11) fueling resistance and skepticism. These emotional responses demonstrate that trust in AI is not solely a technical matter but is deeply tied to perceptions of fairness, creative autonomy, professional security, and the broader impact of technology on fashion's identity as a human-driven industry.

The category Culture highlights the socio-cultural tensions and psychological discomforts that arise when AI tools are introduced into the work routines of employees. Figure 4 shows the coding process for this category.

The category Culture reflects organizational attitudes and workforce perceptions toward digital transformation. Axial coding revealed that resistance to change, limited AI literacy, and concerns over professional identity emerged as central cultural barriers shaping the adoption of AI in the FSC.

Resistance to digital transformation emerged as a recurrent theme, rooted in skepticism and detachment from AI initiatives. Employees across both creative and operational roles described a detachment from digital initiatives, noting that established ways of working often prevail over new technological practices. As one sales employee observed, “some colleagues are skeptical. They won't change their ways easily” (Sales Staff at Company 7). what illustrates how entrenched practices hinder openness to experimentation and adoption.

Usability challenges and low AI literacy were highlighted as obstacles. Limited familiarity with digital tools and the technical nature of AI created barriers to meaningful engagement, particularly for non-technical staff. As an IT employee remarked, “AI feels too technical for most staff” (IT Staff at Company 10), suggesting how complexity in interfaces prevents broader workforce inclusion and reinforces perceptions of exclusion from the digital transformation process.

Concerns about professional identity and the devaluation of human expertise emerged as a deeper cultural apprehension. These concerns were particularly pronounced in creative functions. A designer noted, “AI is replacing human creativity” (Designer Staff at Company 4), reflecting anxiety that algorithmic systems might weaken the creative core of fashion work. Similarly, a production manager warned, “there is concern about becoming dependent on machines” (Production Manager at Company 6), expressing fears that automation could erode human agency and diminish the value of tacit knowledge accumulated through experience.

The integration of AI within FSC is not solely contingent on technological infrastructure but is profoundly shaped by organizational preparedness and the existence of structured support mechanisms. The coding process for the category Support Structures is presented in Figure 5.

The category Support Structures emphasizes the organizational and institutional mechanisms that provide stability and direction for AI adoption. Axial coding revealed that leadership endorsement, external partnerships, skill development, and governance mechanisms might facilitate effective integration of AI initiatives.

Leadership commitment emerged as a key factor in legitimizing AI adoption and fostering organizational alignment. Active engagement from decision-makers signals strategic priority and reduces internal skepticism. A designer staff member at Company 10 noted that the team only considered using AI for generating sketches after they had support from leadership, illustrating how top management involvement can drive experimentation and acceptance.

External support and partnerships surfaced as critical to overcoming internal capability gaps. Collaborations with technology providers, consultants, or research institutions provided expertise and resources that were otherwise lacking. As a finance manager explained, “outsourced expertise helped to get started on using AI” (Finance Manager at Company 5). Such partnerships enabled organizations to overcome early-stage capability deficits and facilitated knowledge transfer.

Skill development also emerged as a pillar for preparing the workforce. Structured training sessions and collaborative learning might help to mitigate fear and build competence. A designer manager at Company 4 observed that “learning together helped acceptance”, demonstrating the importance of shared learning experiences in fostering readiness for digital transformation.

Finally, governance mechanisms might ensure responsible and compliant AI integration. Concerns arose over the absence of clear institutional policies guiding AI use, particularly regarding ethical standards and data protection. IT staff members at Company 10 highlighted the need for formal protocols, stating: “there should be clear policies in place for AI use” and “there should be rules about responsible use”. Such quotes suggest that organizational readiness depends not only on resources and skills but also on codified standards.

The decision to explore or adopt AI is shaped by a confluence of external pressures that redefine organizational priorities and influence the pace of innovation. The coding process for the category Market Dynamics is presented in Figure 6.

Market Dynamics capture the external pressures that shape organizational behavior, with particular intensity in the fashion sector. Axial coding showed that innovation pressure pushes fashion companies to explore new technologies to remain competitive in an industry defined by short product cycles, volatile demand, and constant aesthetic reinvention. This pressure is amplified by rapidly shifting customer expectations: consumers seek hyper-personalized styles, quicker product drops, and seamless omnichannel experiences, making digital transformation not optional but essential.

Competitive imitation also emerged as a strong driver. Firms perceived the need to adopt AI in response to peer activity, with benchmarking and sectoral mimicry accelerating change. As a finance manager observed, “other companies are using it” (Finance Manager at Company 5), revealing how, in fashion, a trend-sensitive and reputation-driven industry, awareness of competitors' digital strategies heightens urgency regardless of internal preparedness.

Participants also highlighted the growing influence of consumers in the fashion market, where demands for personalized services and faster availability directly affect operational decisions. A production staff member at Company 11 noted that adoption of AI-enabled systems is shaped by the fact that “customers expect more personalized services” (Production Staff at Company 5), reflecting how fashion brands must respond quickly to evolving preferences to retain loyalty.

Finally, respondents positioned AI not merely as a technological add-on but as a strategic necessity for long-term competitiveness and innovation, central concerns in the fashion sector's transition toward more efficient, transparent, and environmentally conscious practices. An IT manager emphasized that “future competitiveness depends on how companies use AI daily” (IT Manager at Company 5), signaling a broader recognition that, in fashion, AI is becoming integral to survival, differentiation, and responsible growth.

The category Environment highlights the perceived potential of digital technologies to drive sustainability and reduce the ecological footprint within the FSC. Figure 7 shows the coding process.

This category represents the significant role that AI is expected to play in the fashion sustainability. Axial coding identified three primary areas of impact: waste reduction, resource management, and eco-compliance.

The potential for waste reduction through virtualization was a central theme. There is a clear understanding that replacing physical processes with digital alternatives can lower material consumption. It was explained that digital tools may “allow to visualize virtually, preventing the waste of materials rather than actually making and then discarding” (Design Staff at Company 3). Similarly, the ability to improve the sampling process was recognized as a relevant step toward minimizing the industry's pre-consumer waste, with respondents noting that “digital samples may speed up the process and reduce physical samples and therefore less fashion waste” (Design Manager at Company 4).

The results point to a recognition that technology can optimize resource management during production. Beyond reducing samples, there is an understanding that intelligent systems can maximize the utility of raw materials. It was observed that “software helps reduce waste during the placement phase” (Production Staff at Company 8), which helps to minimize cuts in the manufacturing phase. Furthermore, the integration of physical hardware with digital monitoring was highlighted, with the realization that “sensors could monitor water usage and the CO2 emissions” (Logistics Staff at Company 1), ensuring that natural resources could be used more efficiently.

Eco-compliance emerged as a new opportunity in which AI is seen as a solution to data complexity. As regulations become stricter, there is a consensus that manual tracking is no longer sufficient. It was suggested that advanced tools are essential “to identify impacts and internalize impact analysis through management software” (IT Manager at Company 5). This capability is understood as essential because “AI may help to integrate and merge data from very different sources” (IT Staff from Company 10), transforming environmental data from a compliance burden into a strategic asset.

However, alongside these benefits, a paradox emerged regarding the environmental cost of the technology itself. While AI is viewed as a tool that may increase sustainability, there is a rising consciousness about the resources required to power these digital systems. Concerns were raised that “the water consumption of data centers could offset the advantages” (Sales Staff at Company 11), highlighting a fear that the digital transition might increase the environmental impacts burden rather than reducing it.

Building on the evidence emerging from the results, the seven categories derived from the grounded theory were analyzed into the dimensions of purposes, barriers, enablers, and drivers that shape companies' adoption of AI within the FSC. This analytical framework facilitates the interpretation on how different organizational actors perceive these changes. In particular, the analysis accounts for the perspectives of both staff (“S” in the table) and managers (“M” in the table). Accordingly, Table 2 sums up the main insights discussed in the following paragraphs. Specifically, Section 5.1 discusses the findings related to the purposes of applying AI in the FSC; Section 5.2 explores the main barriers to its implementation; Section 5.3 examines the key drivers that motivate its adoption; and finally, Section 5.4 presents the enablers for AI adoption.

The purposes driving AI adoption in the FSC reveal patterns that both converge with and diverge from findings in other sectors. Within the Data-Driven category, staff primarily associate AI with forecasting accuracy, operational stability, and inventory optimization. This operational focus aligns with trends reported in logistics, where AI is applied to improve operational efficiency, enable predictive analytics, and optimize logistics and procurement processes (El Jaouhari et al., 2025; Maghroor et al., 2025). Similarly, fashion professionals view AI as a tool capable of improving operations, forecasting accuracy, and resource allocation. However, the importance attributed to demand planning and inventory control appears more pronounced in fashion. This sector's short product lifecycle, high demand uncertainty, and strong exposure to overproduction make staff more sensitive to operational risks than workers in other SC contexts. Management, by contrast, understand AI as a strategic enabler for better decision-making, improved analytical insight, and alignment with market trends. This perspective is similar to literature, where AI is associated with decision support systems, real-time analytics, and predictive modeling that enhance managerial reasoning and planning capabilities (Maghroor et al., 2025). Likewise, the use of AI for resilience testing, scenario planning, and strategic response to environmental changes has been highlighted (El Jaouhari et al., 2025). Nonetheless, fashion-focused interviews reveal a stronger emphasis on personalization compared to other sectors. Fashion managers see personalization as a strategy to reinforce brand positioning and respond to fast shifts in consumer behavior.

Regarding the Market Dynamics, staff and managers agree that AI adoption is partly a reaction to competitive pressure and benchmarking. Yet, managers more explicitly highlight customer expectations in shaping AI's purpose. While recent SC studies emphasize AI's ability to strengthen resilience, enhance SC visibility, and mitigate vulnerabilities (Maghroor et al., 2025), the fashion context reveals a stronger orientation toward market responsiveness and consumer-facing value creation. In sectors such as logistics, AI-driven initiatives often focus on operational coordination, sourcing evaluation, and risk reduction across supply networks (El Jaouhari et al., 2025). By contrast, the fashion sector's fast and competitive environment amplifies AI's external strategic purpose.

The Environment category highlights a significant evolution in how TBL principles are integrated into AI adoption. Recent SC research frames AI as a tool for sustainability optimization, resource efficiency, and environmental compliance (El Jaouhari et al., 2025; Maghroor et al., 2025). In logistics, AI applications include automated sustainability reporting, carbon footprint modeling, and environmentally optimized routing decisions (El Jaouhari et al., 2025). Similarly, studies on sustainable SC management emphasize AI's contribution to energy optimization, circular economy initiatives, and waste reduction (Maghroor et al., 2025). In fashion, Waste Reduction emerged, pointed out by both staff members and managers, where digital tools are used to visualize samples virtually, reducing physical waste during the design phase. Staff members specifically emphasized resource optimization, pointing to software that optimizes fabric placement to minimize off-cuts and sensors that monitor energy and water consumption in real time. Furthermore, both groups highlighted compliance tracking and reporting as an interesting purpose.

The barriers identified in this study present parallels with challenges documented in other sectors, but important differences emerge when considering the fashion context and the distinct perspectives of staff and management. Within the Technology category, staff members highlighted reliance on legacy systems as a major obstacle. This aligns closely with barriers reported in literature, where integration complexity, system incompatibility, and limited digital infrastructure hinder the deployment of AI technologies (El Jaouhari et al., 2025; Han and Li, 2025). Similar challenges related to data quality, data governance, and the difficulty of managing large volumes of operational data have also been emphasized as key technological constraints in AI adoption (Maghroor et al., 2025). However, the operational implications appear particularly intense in fashion, where legacy platforms constrain not only data integration but also responsiveness to rapidly changing consumer trends. For staff, the lack of intuitive, accessible interfaces further limits adoption, similar to studies in which limited AI literacy and a lack of personnel capable of interpreting AI-driven emerge as major human and technological barriers to implementation (Han and Li, 2025; Maghroor et al., 2025).

Management perspectives reinforce these structural barriers but emphasize different dimensions. High infrastructure costs align with the financial constraints identified in sustainable SC and logistics studies, where high investments and uncertain return on investment often delay AI adoption (El Jaouhari et al., 2025). Managers also stressed accessibility issues, recognizing that adoption is limited when systems are too technical for widespread use. Similar concerns have been raised regarding the lack of technological maturity and limited scalability of AI solutions, which create gaps between advanced technological potential and the operational readiness of organizations (Maghroor et al., 2025).

Barriers related to Trust also emerged. Staff placed strong emphasis on opacity, bias, and fears of job displacement. Concerns about algorithmic transparency, explainability, and the reliability of AI outputs closely mirror issues identified in literature, where lack of transparency and potential misuse of AI systems raise ethical and operational concerns (Maghroor et al., 2025). Fear of job displacement and stakeholder mistrust are also reported as social barriers to AI adoption, reflecting anxiety about the redistribution of work and the disruptive consequences of automation (Han and Li, 2025). Anxiety about job security is especially prominent in fashion, where creative and operational roles are highly exposed to automation and where replacement fears are amplified by the industry's employment structures. Management, on the other hand, cited data privacy and the lack of governance protocols, a pattern consistent with regulatory and data security concerns reported in logistics-focused research (El Jaouhari et al., 2025). Yet unlike highly regulated sectors, fashion respondents described governance gaps as organizational rather than compliance-driven, reflecting the sector's less formalized data culture.

Cultural barriers further differentiate the fashion sector from the industries examined in the comparative studies. Staff described resistance to change and low AI literacy, consistent with organizational and human barriers reported in the SC literature, including workforce resistance, limited AI competencies, and low awareness of AI benefits (Maghroor et al., 2025). However, the perceived threat to creativity emerged as a fashion-specific barrier. While other sectors primarily discuss operational resistance or organizational inertia, fashion staff specifically articulated fears that AI could dilute the originality and tacit knowledge essential to creative roles. Management acknowledged these tensions and recognized the cultural value of preserving human-driven creativity.

An environmental barrier emerged that challenges previous assumptions about the sector's focus. Recent studies in logistics recognize the environmental footprint of AI technologies themselves, highlighting concerns about the high energy consumption and water usage associated with training and operating large-scale AI models (Han and Li, 2025; Maghroor et al., 2025). While comparative literature typically frames sustainability challenges in terms of regulatory compliance or environmental performance, the present study uncovered a paradox regarding the Environment. Contrary to the expectation that sustainability would only act as a driver, staff members explicitly identified the environmental cost of the technology itself as a significant barrier. This reflects a situation in which the digital tools intended to drive sustainability are analyzed for their own resource consumption. This worry regarding the ecological footprint of digital infrastructure represents a barrier visible in environmentally conscious sectors such as fashion, suggesting that the environmental cost of AI may limit its potential benefits.

The drivers identified in this study reflect pressures and motivations that are partly shared across sectors and partly shaped by the specific dynamics of the fashion industry. Within the Data-Driven category, staff emphasized operational efficiency as a primary driver. This motivation aligns with drivers identified in literature, where organizations pursue AI adoption to achieve cost savings, enhance operational efficiency, and improve risk management capabilities (El Jaouhari et al., 2025). However, the fashion-sector emphasis is more tightly tied to the daily challenges of forecasting uncertainty and short product lifecycles. These operational complexities intensify the perceived need for AI as a practical enabler of more stable workflows. Management, in contrast, stressed the need for enhanced analytics. This perspective aligns with findings in logistics, where AI adoption is driven by the increasing importance of data-driven decision making, SC transparency, and advanced optimization (Han and Li, 2025). In fashion, the strategic driver is within the perceived insufficiency of current data practices, which limits foresight in trend prediction and market responsiveness.

In the Market Dynamics category, competitor activity emerged as a shared drive among both staff and management. This finding aligns with broader SC studies that identify market competition as a fundamental force motivating organizations to adopt AI in order to maintain strategic positioning and operational performance (Maghroor et al., 2025). However, in fashion, the influence of competitor activity is amplified by fast lifecycles and the speed at which trends propagate, creating a sense of urgency compared to slower-moving sectors. Staff interpret competitor actions as a signal that inefficiencies may soon become unsustainable, while management understands them as a strategic necessity to avoid falling behind. Shifts in market trends represent another common driver. Both staff and management observed that rising consumer expectations are transforming industry standards. Comparable dynamics are visible in logistics and sustainable SC research, where technological adoption is stimulated by the need for more transparent, responsive, and data-driven supply networks (Han and Li, 2025). Yet the fashion sector differs in the centrality of personalization as a driver, reflecting its reliance on differentiated consumer experiences.

Comparing staff and managerial perspectives reveals distinct motivational emphasis. Staff focus on immediate operational improvement, highlighting efficiency, workflow stability, and the reduction of routine workload. Management focuses more strongly on strategic and analytical insufficiencies that threaten long-term competitiveness, including the need for improved insights and the capacity to respond to unpredictable market dynamics. These differences mirror patterns observed across SC research, where operational personnel tend to emphasize efficiency gains while leadership highlights competitive positioning and strategic adaptability in increasingly volatile environments (Maghroor et al., 2025). Despite these differences, both groups agree that external pressures such as competition, technological progress, and evolving market expectations act as unavoidable drivers, generating a shared sense of urgency to accelerate AI exploration.

The enablers identified in this study echo many of the facilitators reported in other sectors, but they also reflect the particular organizational structures and capability gaps that characterize the fashion industry. Within the Support Structures category, leadership endorsement emerged as a foundational enabler. This aligns with recent SC research emphasizing organizational readiness and strong leadership commitment as central conditions for successful AI adoption (Han and Li, 2025; Maghroor et al., 2025). In the fashion context, staff understand leadership signals as essential for granting legitimacy to experimentation and reducing uncertainty around the role of AI.

Workshops and training initiatives also stood out as important enablers for staff. Similar patterns appear across sectors, where workforce readiness and structured reskilling programs are identified as fundamental conditions enabling organizations to effectively implement AI technologies (Maghroor et al., 2025). In fashion, the role of workshops is particularly important due to wide disparities in digital skills between creative, operational, and managerial functions. Workshops demystify AI, foster shared understanding, and build confidence among employees who may perceive AI as inaccessible.

Hiring AI specialists is another enabler that emerged from staff perspectives. While other industries often benefit from stronger in-house analytical capabilities, recent research emphasizes the importance of skilled human capital and cross-functional AI implementation teams to translate technological potential into operational practice (Han and Li, 2025; Maghroor et al., 2025). Fashion respondents described a high reliance on newly recruited or external expertise, reflecting the sector's effort to build internal analytical competencies. External consultancy was another shared enabler. For staff, external experts provide support that reduces the learning curve and helps translate abstract concepts into practical workflows. For management, consultants offer strategic guidance, facilitate technology selection, and accelerate adoption without overburdening internal teams.

Across both staff and management perspectives, guidelines and governance frameworks were consistently identified as enablers. This finding aligns with the emphasis on standardized AI governance frameworks, regulatory coordination, and ethical oversight (Maghroor et al., 2025). Similarly, studies on AI adoption in logistics underline the role of supportive regulatory frameworks and institutional clarity in reducing uncertainty and encouraging technological experimentation (El Jaouhari et al., 2025; Han and Li, 2025). In the present study, governance frameworks played two roles: they reduced ethical and operational ambiguity for staff, and they helped management ensure accountability, transparency, and alignment with emerging regulatory expectations.

Distinct from these organizational supports, staff members also highlighted a specific technological enabler: sensors for monitoring. This technology can act as a bridge between physical operations and environmental sustainability by enabling the precise tracking of resource consumption and production efficiency. While comparative literature often emphasizes broader technological readiness and infrastructure as enabling conditions (Han and Li, 2025), the fashion interviews illustrate how specific monitoring technologies can operationalize AI by generating the real-time data necessary for reduction of energy and water consumption.

Comparatively, enablers frequently discussed in other sectors, such as highly advanced digital infrastructures or fully mature AI ecosystems, were not highlighted with the same intensity by the respondents of this study. This difference reflects the maturity level of AI adoption within the fashion industry. While logistics and SC studies often assume a degree of technological readiness, fashion organizations appear to remain in earlier stages of capability development, where leadership support, workforce preparation, governance frameworks, and collaborative partnerships play a more immediate role in enabling AI adoption.

This study investigated the adoption of AI in the FSC using a GT approach, based on interviews with twenty-seven professionals from eleven Italian companies. The findings show that AI adoption is shaped by a complex relationship between technological, cultural, organizational, and environmental factors, indicating that the introduction of AI represents a broader socio-technical transformation rather than a purely technological shift.

Seven interrelated analytical categories were identified: Data-Driven, Technology, Trust, Culture, Support Structures, Market Dynamics, and Environment. The analysis was structured across four dimensions: purposes, drivers, barriers, and enablers. Such categories provide insight into why companies pursue AI, what accelerates or hinders its adoption, and which mechanisms facilitate implementation. The study also compared staff and managerial perspectives, revealing a multi-level adoption process in which operational and strategic viewpoints interact to shape how AI initiatives are interpreted and implemented within organizations.

From a theoretical perspective, this study contributes to the literature by proposing an industry-specific framework that extends traditional technology adoption models. It integrates technological, cultural, organizational, and environmental elements into a coherent model, highlighting that AI adoption is iterative and adaptive. From a practical perspective, the results indicate that successful AI adoption requires more than investment in infrastructure or advanced technologies. Companies should adopt a systemic approach that includes employee upskilling, governance frameworks to guide strategic and ethical decision-making, and leadership engagement to support organizational change.

Some limitations should be acknowledged. The study focuses on Italian companies, which may reflect context-specific characteristics. Moreover, although appropriate for GT methodology, the sample size does not allow statistical generalization. Finally, the research focuses explicitly in the forward SC within the fashion industry, not exploring AI adoption in reverse logistics nor in other low-tech sectors.

Future research could expand the analysis to companies in different countries, SC positions, and market segments. Including the perspectives of customers and end-users could also provide insights into how AI-driven innovations are perceived. Further studies may develop quantitative models to assess the economic and operational impacts of AI initiatives in FSCs and examine how AI adoption interacts with emerging paradigms such as Industry 5.0 and circular economy practices. Nevertheless, future studies should acknowledge the application of AI in the circular economy and closed loop in the fashion industry.