Innovators and transformers: Enhancing supply chain employee training with an innovative application of a large language model

Effective employee training capabilities are essential for maintaining responsive supply chain functions (Brockhaus et al., 2023). However, traditional training methods are often time-consuming and costly. According to a recent industry survey, the average training budget for supply chain teams can range from 1,000 USD to 2,500 USD per employee annually and can go up to 10 million USD for large companies, depending on the complexity of the operations and the level of expertise required (Freifield, 2023). Despite these investments, many companies still struggle with ensuring that their employees retain and effectively apply the knowledge they acquire during training sessions. Recent studies have shown that AI-driven training programs can reduce training costs by up to 30% while improving knowledge retention by 20% (Garg et al., 2021; Long, 2023). Given the significant potential of AI to enhance efficiency, reduce costs, and improve decision-making processes, its adoption in supply chain management is essential for staying competitive in a dynamic market landscape (Richey et al., 2023; van Hoek, 2024). This paper demonstrates how LLMs can enhance training practices for supply chain employees by improving efficiency, reducing costs, and providing personalized learning experiences.

AI is a rapidly evolving field that has the potential to transform various domains, including SCM (Hendriksen, 2023; van Hoek, 2024). Possible opportunities for the applications of AI in SCM are presented in Table 1 (Durach and Gutierrez, 2024; Guida et al., 2023; Hendriksen, 2023).

Among the different types of AI, generative AI (GenAI) is a powerful paradigm that can create novel and diverse content, such as text, images, code, and music, based on large amounts of data and user queries. GenAI has gained significant attention around the world after the introduction of ChatGPT in December 2022. ChatGPT, a prominent example of GenAI, is a chatbot API based on large language models (LLMs) that can generate realistic and coherent text responses.

ChatGPT has been applied in various fields, such as education, healthcare, finance, journalism, and climate science, but its implications for SCM remain underexplored (Doshi et al., 2023; Fosso Wamba et al., 2023). ChatGPT is only one of the many LLM-based platforms that are identified as disruptive innovations (Lund et al., 2023). LLMs have shown remarkable capabilities in various natural language processing tasks, such as text summarization, translation, question answering, and sentiment analysis. In this paper, we explore the potential applications and benefits of LLMs for SCM and the challenges and limitations of this emerging technology. In particular, we focus on an intervention that involved the development of an LLM-based application for improving the efficiency and reducing the costs of training supply chain employees.

ChatGPT is based on a “generative pretrained transformer” (GPT) model that uses a neural network architecture and continuous learning to generate high-quality, plausible, human-like written responses (Rahimi and Talebi Bezmin Abadi, 2023). ChatGPT has attracted more than 100 million users since its release, making it the most popular chatbot in recent times. It uses a specifically fine-tuned GPT-based model for human-like conversation. For building customized chatbots, OpenAI provides an application programming interface (API) that allows developers to use other coding platforms to interact with OpenAI’s GPT-based LLMs.

LLMs could impact SCM in three main ways: knowledge management, data analysis and process automation (Dale, 2022; Gagliardi et al., 2023). Knowledge management with LLMs would involve using these models to capture and transfer institutional knowledge and best practices that are often lost due to high turnover, stress or lack of documentation (Fosso Wamba et al., 2023). LLMs could potentially be trained on a company’s internal data sources, such as emails, documents, training materials and transactional history, and then used to answer queries, explain procedures, provide guidance or suggest solutions based on the context and the user’s needs.

Data analysis with LLMs would involve using these to augment or replace traditional data processing methods. In the context of SCM, LLMs could be used to extract insights, trends and patterns from large and complex datasets, such as demand forecasts, inventory levels, transportation routes or supplier performance (Bahrini et al., 2023). LLMs could also be used to generate natural language summaries or reports that highlight the key findings and recommendations for decision making (Edureka, 2022). Process automation with LLMs would involve using LLMs to execute or trigger actions based on user prompts or predefined rules. LLMs could also be used to automate tasks such as order taking, invoicing, returns processing or after-sales service. These models can also interact with other systems or devices to perform actions such as scheduling deliveries, updating inventory or sending alerts (Dale, 2022; Rathore, 2023).

The strength of LLMs lies in their capacity to seamlessly integrate diverse data sources, such as sales data, social media trends, and market research reports. Additionally, these models can be automatically optimized based on the integrated data, facilitating more efficient deployment (Aguero and Nelson, 2024; Liang et al., 2024). Some of the benefits of using LLMs in SCM include reduced costs, increased efficiency, improved customer satisfaction, and enhanced innovation. However, there are also some challenges and risks associated with LLMs, such as data quality, ethical issues, security threats and the lack of adequate human oversight (Lund et al., 2023).

LLMs can predict demand for products by analyzing historical sales data, seasonal trends, weather patterns, and economic indicators (Hendriksen, 2023; O’Leary, 2023). LLMs can also be used to analyze inventory levels and recommend actions to optimize inventory, such as adjusting reorder points or safety stock levels based on demand forecasts and lead times. Furthermore, LLMs can be utilized to identify potential risks in the supply chain, such as disruptions in transportation or production, and recommend strategies to mitigate those risks (Bahrini et al., 2023). This can help companies to proactively manage potential disruptions and minimize the impact on their operations. Lastly, LLMs could improve customer service by providing real-time responses to customer inquiries and complaints (Rathore, 2023). Table 4 in the  Appendix provides concise explanations for key LLM-related terms and concepts relevant to this research.Table A1 

The research approach involved an intervention undertaken at an eyewear company during which a solution was designed for improving the effectiveness and efficiency of training supply chain employees. The solution involved using a LLM (GPT 4.0) along with the LangChain and Faiss libraries in Python. LangChain is a robust framework that facilitates seamless integration of various language models, ensuring flexibility and scalability (Chase, 2023). The Faiss (Facebook AI Similarity Search) vector database was chosen for its efficient similarity search capabilities, which are crucial for handling large volumes of data and providing relevant training content (Lee et al., 2024). The deployment phase included iterative testing and refinement based on employee feedback to fine-tune the effectiveness of the methodology. Evaluation was conducted through continuous feedback loops with employees, which helped in identifying areas for improvement and ensuring the SCM training goals of improving efficiency, reducing costs, and providing personalized learning experiences wee effectively addressed.

This study uses an intervention-based research (IBR) approach. This is a research strategy that is implemented in a real world setting that leverages interventions to test or develop theories (Oliva, 2019). While IBR shares some similarities with other research methodologies used in business, such as experimental research and quantitative research, this approach offers a distinct advantage by offering a systematic framework for conducting and evaluating interventions, thereby enhancing the validity and generalizability of research findings. The interventionist research approach focuses on a problem situation (S) and involves identifying a relevant theoretical framework (T), building a methodology (M), and implementing M that is based on T to get to an improved situation (S′). Using this approach, researchers focus on learnings about T, M and S from the use of M (Anand et al., 2021; Oliva, 2019). Figure 1 illustrates this methodology. In this study, the goal is to improve the efficiency of the knowledge acquisition process for new supply chain employees in an organization.

The case company, Eyewear Co., is a manufacturer of spectacle frames in Australia. While the company remains anonymous for confidentiality reasons, it is a real-world organization. The researchers identified the company as a suitable candidate due to its rapid growth, making it an exemplary case study for organizations facing challenges while expanding their supply chain functions. With an annual turnover of around 33 million USD in 2023, Eyewear Co. operates with 48 full-time employees. Within its supply chain function, a dedicated team of 7 individuals manage critical operational activities such as procurement, logistics, and inventory control. Notably, the company allocates an overall training budget of 41,000 USD per year for its supply chain staff. Additionally, Eyewear Co. has experienced double-digit growth over the past 9 years, further increasing the need for effective training to ensure that employees have the necessary knowledge and skills to meet the changing operational needs. The company has been onboarding new personnel to support a growing supply chain function. The training process for new supply chain employees can be lengthy and complex, as it involves familiarizing them with the company’s products, processes, and suppliers.

To understand the training related challenges, interviews were conducted with managers and other supply chain employees. Onboarding newly hired supply chain staff at Eyewear Co., which operates in Australia and sells through online channels in Canada, Australia, and New Zealand, can be particularly complex due to the global nature of the company’s operations. New supply chain employees need extensive training in international logistics, customs regulations, and inventory management processes.

It is essential for employees to be familiar with the eyewear industry’s unique supply chain challenges, such as the delicate nature of products and the importance of timely and safe shipping. Moreover, it is important to ensure that employees across different operational locations receive a consistent onboarding experience while also considering region-specific needs. A thorough understanding of the company’s values, compliance standards, and commitment to customer satisfaction is essential to maintaining the company’s reputation in various markets.

Eyewear Co. has a training repository that is composed of various materials in text format. When this project was initiated, there were 22 different training documents for standard operating procedures. These documents included a total of 9211 sentences. It was a cumbersome process for the new employees to search for information regarding how to respond or act in specific circumstances.

Organizational learning theory focuses on how organizations acquire, interpret, and apply knowledge to improve performance over time (Cangelosi and Dill, 2016; Qian et al., 2023).This theory highlights the importance of fostering a learning culture throughout all levels of the organization, facilitating knowledge sharing, and influencing overall efficiency in production and operational processes (Argote et al., 2021).

Organizational learning is the dynamic process of acquiring, disseminating, interpreting, using, and storing information within organizations, yielding new knowledge or insights that impact organizational strategies. This multifaceted learning process, encompassing stages such as search, knowledge creation, retention, and transfer, is further extended by inter-organizational learning along the supply chain, commonly referred to as “supply chain learning” in the literature (Bessant et al., 2003; Zhu et al., 2018). While this extension enhances a firm’s knowledge base and provides valuable market and strategic insights, it also introduces challenges, such as the potential unintended and undesirable knowledge transfer, risking the dilution of competitive advantage (Qian et al., 2023). In the context of onboarding, it can provide a framework for understanding how employees and the organization can learn and adapt to the complexities of the eyewear industry and the specific challenges associated with the company’s operations.

Following the initial round of interviews to gain a high-level understanding of the problem situation as described in Section 3.2, further interviews were conducted to understand the common issues in more detail and design a solution approach with supply chain staff. The staff onboarding process was challenging, often falling short of expectations due to the volume of information required for new employees and the time needed to prepare and update training materials. The primary issue identified was that while the training materials covered routine tasks comprehensively, they lacked specific guidance for tackling emerging problems. Creating an application-specific chatbot emerged as the consensus solution for providing users with efficient access to insights from Eyewear Co.’s training repository. This consensus was fuelled by the fact that the company had recently started allowing staff to explore OpenAI’s ChatGPT platform and were keen on identifying practical applications incorporating OpenAI’s API. The chatbot solution for increasing learning efficiency for new supply chain staff is shown in Figure 2.

The methodology chosen for this project was to use LangChain, Faiss vector database, and OpenAI’s API to build an LLM-powered application for training supply chain personnel. The methodology has two workflows: document handling workflow and query handling workflow. In the document handling workflow, the first step involved converting the training documents (PDF) into chunks of text of specific word length and storing them in vector format. This was done using the LangChain framework, a powerful tool that can be used to build various applications that use LLMs (Chase, 2023). First, LangChain was used to split the documents into smaller chunks at the word level. Then these chunks were converted into vector embeddings. In this project word2vec word embedding was used for this purpose. Once the LLM (GPT 4.0) was fine-tuned, it was stored in a Faiss vector database which can be used to store and query the data.

The query handling workflow involved creating a chatbot capable of receiving messages from personnel and responding with appropriate training materials or feedback. The chatbot was integrated with the Faiss vector database to allow it to access the necessary information. This was done using the Langchain framework which facilitates the construction of an intricate, hierarchical series or chain of actions such as a question-answering (QA) chain. The QA chain was built to respond to enquiries rooted in each collection of documents. It accomplishes this by executing a similarity search, matching the input question with embedded documents, and subsequently employing a model (GPT 4.0) to formulate an answer rooted in the most pertinent documents. The application was deployed to a cloud-based server so that it could be accessed by personnel around the world.

After designing the solution, the next step was to develop and implement the system which would allow new supply chain employees to query training materials efficiently. The development process involved fine-tuning the LLM (GPT 4.0) with the textual data from the training materials and developing guidelines for prompt engineering to facilitate effective retrieval of relevant information. Prompt engineering is a strategy for creating efficient neural network-based Automatic Question Generators (AQGs) when dealing with limited data sets (Budhwar et al., 2023). This approach involves offering hints or prompts to a language model, assisting it in comprehending the task and producing the intended outcomes. By leveraging LLMs, individuals can attain the desired results even with small datasets and without an extensive technical background, if precise and suitable prompts related to the domain are supplied (Lee et al., 2023). In this step of the project, several training sessions were held for the supply chain staff. The purpose of these sessions was to help the staff members learn to write better prompts to obtain more useful outputs from the chatbot. The solution was integrated with the existing learning management system, enabling seamless access to training material.

The final step of this IBR approach was to evaluate the solution. This involved measuring the impact of the solution on the efficiency of the training process.

Table 2 shows the results for the performance indicators related to post-training tests taken by supply chain employees before and after the intervention. Employees are expected to complete this training annually using the company’s e-learning system. The pre-intervention results are related to the last round of training prior to the intervention. After the intervention two additional employees who were new hires completed the training and the post-training test alongside five existing employees. The average test score after the first attempt increased from 75% (pre-intervention) to 85% (post-intervention) and the average test duration dropped from 17 min to 14 min. Furthermore, the training completion rate (i.e. average proportion of training content viewed by an employee) within the given time deadline increased from 62% to 93% and the test completion rate (i.e. percentage of employees completing the test) also increased from 60% to 86%.

The research team also collected qualitative feedback from the supply chain employees to understand their experience with the LLM-based training system in terms of ease of use, relevance of retrieved content, and overall satisfaction. Qualitative feedback from these trainees indicated that the LLM-based system made information retrieval faster and also improved their understanding of key issues and processes, leading to better job performance.

The preliminary results of this IBR study showed that using an LLM-based application for querying training material improved the efficiency of the training process. The study found that new supply chain employees were able to find the information they needed more quickly and easily using the application. This led to a reduction in the amount of time required for training and a corresponding improvement in the quality of the training. While the novelty of using a chatbot may have contributed to initial engagement, the improvement in test scores and reduced test duration suggest a genuine enhancement in training efficiency. Furthermore, new employees reported feeling more confident and competent in their roles within a shorter period, as evidenced by their ability to complete tasks independently sooner than those trained without the LLM-based system. Example queries and corresponding answers are shown below.

Table 3 shows examples of queries used by the employees. Answering these questions requires an intricate understanding of the text within the training material. For example, it would be impossible to get an instant summary of the training material, as required by the first query without using an LLM. Furthermore, without the use of an LLM it would be impossible to quickly extract the key performance indicators that are important for the supply chain team, as these performance indicators were spread across different documents and in different contexts. Therefore, with the power of an LLM it was possible for employees to gather necessary information more efficiently than before the intervention.

Historically, training involved a combination of online and in-person sessions, costing around 900 USD per employee. With the chatbot’s introduction, there was a marked reduction in the time spent reviewing and updating training materials and running face-to-face question-and-answer sessions. In the past, training sessions for supply chain employees at the company required considerable resources. Many of these sessions were conducted on-site, necessitating the use of outsourced meeting rooms, refreshment provisions and time commitment from cross-functional managers charged with delivering modules.

The implementation led to a notable reduction in the time allocated to training material preparation, delivery, logistical arrangements and coordination of multiple managers involved in the training process. This optimization led to a consolidated training cost of around 500 USD per employee, reflecting a substantial 44% cost reduction.

Beyond initial training, the chatbot continues to support employees by providing ongoing assistance and resources. Employees can access information quickly, allowing for more efficient problem-solving and knowledge acquisition. This accessibility fosters a culture of continuous learning within the organization.

This study’s impact extends beyond Eyewear Co. by showcasing the benefits of integrating AI-driven systems in training processes. The demonstrated cost and time savings offer a model for other companies seeking to optimize their training methodologies. Additionally, this study aligns with organizational learning theory, emphasizing the importance of adaptive learning mechanisms for organizational growth. The chatbot’s implementation serves as a testament to the theory’s practical application in enhancing employee learning experiences and organizational efficiency. Moreover, this study provides a blueprint for leveraging technology in training practices, illustrating its broader applicability across industries. The demonstrated benefits highlight the potential of AI-driven systems to revolutionize learning methodologies and enhance organizational efficiency.

In this section, we discuss the implications of the intervention for both research and practice. We also acknowledge the limitations of the study and suggest some directions for future research.

Our findings demonstrate that LLMs and AI technologies significantly enhance SCM employee training by improving efficiency and providing personalized learning pathways. These technologies enable faster knowledge acquisition and better retention, which are crucial for SCM advancements. The practical significance of these results lies in the ability to tailor training programs to individual needs, thereby increasing overall training effectiveness and reducing costs.

Before the implementation of the LLM-based solution, trainees often had to spend a significant amount of time searching through the training manuals and documentation to find specific information. With the LLM-based system, a process was implemented whereby the system was continuously trained with new documents. This assisted the employees in submitting natural language queries and receiving instant responses with relevant content. This resulted in faster knowledge acquisition and reduced time-to-competence for staff members. The pre- and post-intervention training processes are shown in Figure 3.

The solution allowed the training experience to be tailored based on individual trainee needs. By analyzing the queries made by trainees, cross-functional managers could identify areas where more support or additional resources were needed. This personalized approach enhanced the effectiveness of training and catered to diverse learning styles.

Ongoing employee engagement and feedback play a pivotal role in refining AI-driven training tools. Continuous feedback loops ensure that the training content remains relevant and effective, positioning this approach as a cornerstone of best practices for future SCM innovation. User feedback and performance metrics provided valuable insights for making iterative improvements to the LLM-based system. A key benefit of the system was the improvement in the training completion rate from 62% to 93%. This demonstrates the system’s efficacy in engaging employees and facilitating comprehensive learning experiences. Furthermore, employee satisfaction with personalized content contributed to a better training environment, fostering greater knowledge retention. As the system has been successfully trialed with the supply chain team in Australia, the company plans to utilize this solution in its operations in Canada and New Zealand as well.

Implementing the LLM-based solution to support training also presented some challenges. One primary concern was ensuring the accuracy and reliability of retrieved content. It is important to acknowledge the inherent limitations and potential pitfalls associated with AI-driven systems, particularly in relation to the accuracy and reliability of responses generated. Conducting rigorous audits and implementing corrective measures to rectify inaccuracies are crucial steps in mitigating the risk of misguiding or misinforming employees, thereby ensuring the efficacy and trustworthiness of the training platform. The research team addressed this by providing employess with guidance on prompt engineering. This involved creating training materials for staff on how to write better prompts.

Another challenge was the need for data privacy and security. The company’s supply chain and ICT team collaborated on implementing robust privacy protocols to ensure that sensitive information was not exposed during the querying process. While the intervention demonstrated promising results in enhancing the training process at Eyewear Co., the study would benefit from a more rigorous evaluation through control measurements or A/B testing to isolate the specific impacts of the LLM-based chatbot. This would provide more substantial evidence of causality and enable a more precise assessment of the intervention’s effectiveness.

The study’s exploration of LLM-powered tools in the context of SCM opens avenues for further research. Future investigations could delve deeper into the impacts of LLMs on SCM processes, examining factors such as scalability and adaptability in different organizational contexts, and the potential evolution of these tools over time. The proposed methodology, centred around the integration of LLMs into the training process for a company’s supply chain team, sets a foundation for broader research endeavours. Future research could broaden the use of LLM-based tools such as chatbots to diverse contexts, including student learning. These chatbots, available even after work hours, can provide contextual answers, offering students flexibility and a dynamic educational experience (Gilbert, 2024).

The methodology employed in this study contributes to organizational learning theory in various ways. In alignment with the foundational principles of organizational learning that emphasize problem recognition and resolution, the initial phase involved a thorough identification and understanding of the challenges faced by a company in onboarding new supply chain employees. By selecting IBR as the research methodology, the study embraces an active approach to organizational improvement, consistent with the continuous learning approach. Furthermore, the detailed analysis of the organizational complexities associated with global operations, logistics challenges, and compliance standards resonates with the organizational learning theory’s emphasis on understanding and navigating the intricacies of the organizational context. In essence, the methodology addresses specific organizational challenges and contributes to the broader organizational learning theory by incorporating technology, adaptability, and a proactive problem-solving orientation. The specific outcomes that were obtained and shown in the previous section would not be possible if an LLM-based method was not used.

Specifically, this study identifies the integration of LLMs into functional knowledge acquisition as a key refinement to the organizational learning theory. By incorporating LLMs into SCM activities such as demand forecasting, inventory management, order processing, and supplier management, organizations can enhance their ability to learn from data, adapt to changing market conditions, and make more informed decisions. In essence, this study extended organizational learning theory by incorporating generative AI for facilitating the development of a continuously learning sociotechnical system where the learning abilities of the technical component enrich the knowledge acquisition experience for the organization.

To advance research in generative AI for supply chain management, several actionable recommendations can be pursued. First, developing scalable AI frameworks is essential, focusing on modular and customizable solutions that cater to organizations of varying sizes and supply chain complexities, ensuring both accessibility for small enterprises and scalability for larger ones.

Expanding the application of AI beyond training is another critical area, with opportunities to explore use cases such as demand forecasting, supplier performance analysis, and enhanced decision support systems. Equally important is prioritizing employee engagement by studying the impact of personalized, AI-driven training modules and implementing robust feedback mechanisms to keep the tools user-centric and effective. Promoting ethical AI adoption through frameworks emphasizing transparency, data security, and trust is vital for encouraging widespread acceptance. Finally, researchers should quantify the cost and efficiency benefits of AI through case studies across various organizational contexts, demonstrating the adaptability and value of these innovations.

For practitioners in the SCM field, the paper offers actionable insights into the practical implementation of LLMs. Organizations inspired by the approach taken at the case company, can consider adopting similar methodologies to enhance the efficiency of their training processes. This implies a shift towards more interactive and LLM-integrated training programs to address specific challenges associated with employee onboarding.

The three key areas of LLM applications—knowledge management, data analysis, and process automation—present tangible opportunities for SCM practitioners. Organizations can strategize the integration of LLMs to capture and transfer institutional knowledge effectively, addressing issues related to employee turnover and inadequate documentation. Practical implementations of LLMs in extracting insights, identifying trends, and automating processes align with the broader goal of improving SCM efficiency.

Practitioners can leverage LLMs to enhance data processing and visualization, thereby facilitating more informed decision-making. The potential for process automation using LLMs in tasks such as order processing and after-sales service presents a pathway for organizations to streamline their supply chain operations, reduce errors, and save time.

This paper demonstrates that relevant stakeholders within a company can work together to identify opportunities for the application of GenAI. They can then collaborate with researchers or consultants or work with their own developers to implement the LLM-based methodology described in the paper. By doing so, they can provide personalized, on-demand learning experiences that can lead to improved knowledge retention and productivity.

This study underscores the transformative potential of LLMs in SCM, opening the way for more efficient, responsive, and innovative practices. Beyond their demonstrated utility in training, LLMs can inspire a broader range of applications, including workflow optimization, decision support, and knowledge management. By processing vast amounts of unstructured data and generating actionable insights, LLMs provide organizations with tools to streamline complex operations, enhance real-time decision-making, and foster organizational agility.

The transformative integration of AI into SCM lies in its ability to streamline complex operations while fostering a more agile and responsive supply chain ecosystem. LLMs can personalize user interactions, democratize access to information, and bridge the gap between technical and non-technical users. This transformation enhances operational efficiency and also supports organizations in achieving strategic objectives.

To fully harness these capabilities, best practices for ongoing employee engagement are essential. Personalized training modules, interactive learning tools, and consistent feedback loops can ensure employees remain engaged, informed, and empowered to use AI-driven systems effectively. Additionally, integrating ethical safeguards, such as robust privacy protocols and bias mitigation frameworks, will be critical to building trust and fostering widespread adoption.

The research provides a foundation for advancing AI applications in SCM and emphasizes the need for collaboration among researchers, practitioners, and organizations. By scaling these innovations responsibly and inclusively, LLMs can drive a new era of efficiency, adaptability, and innovation in supply chain management.

Based on this study, there are several areas for future research. First, LLMs can be refined and fine-tuned to align with the nuanced requirements of SCM. While LLMs can process vast amounts of data and generate natural language responses, their effectiveness hinges on their ability to understand the intricacies of SCM contexts. Further research should focus on domain-specific training and customization to enhance the contextual relevance and accuracy of LLM-generated insights. Additionally, efforts to address the limitations of LLMs, such as susceptibility to biased training data and lack of emotional intelligence, can pave the way for more ethical and sensitive SCM applications. Furthermore, while this study focused on a mid-sized company, the potential benefits of LLM-based chatbots for knowledge management could be even more pronounced in larger organizations with extensive data and document repositories. Langchain and the vector databases have been shown to be applicable for larger datasets (Durach and Gutierrez, 2024; Lee et al., 2024), suggesting that the LLM-based methodology developed in this study will be scalable for use in larger organizations.

Traditional deep learning and LLMs have fundamental similarities as both depend on neural networks. A promising avenue for future research is the exploration of hybrid models that combine the strengths of LLMs with other AI-based approaches and optimization techniques. Integrating LLMs with machine learning, deep learning, or reinforcement learning algorithms can lead to more robust SCM solutions. For instance, in a hybrid solution, deep learning could be used for forecasting demand and supply, while the LLM-based component could allow users to retrieve relevant information without delving into the numbers. Furthermore, research into the development of automated decision-making frameworks that leverage LLM-generated insights to optimize supply chain processes in real time is a promising area for future exploration.

Data privacy, security, and ethical considerations surrounding LLMs in SCM should remain at the forefront of future research. As LLMs handle sensitive business information and customer data, robust data protection and privacy protocols must be adopted and continuously updated. Additionally, detecting and mitigating algorithmic biases within LLMs is essential to ensure fair and unbiased decision-making in SCM applications.

Integrating LLMs into SCM represents a transformative leap towards more efficient and responsive SCM processes. LLMs transform the learning experience by providing personalized, real-time, and adaptive learning pathways tailored to individual needs. They enable rapid access to vast amounts of information, facilitating quicker understanding and retention of knowledge. Furthermore, LLMs support interactive learning through natural language processing, fostering increased engagement and comprehension. As the synergies between AI technologies and SCM continues to evolve, the potential for innovation and improvement in the field remains unlimited. Researchers, practitioners, and organizations should collaborate to harness the full potential of LLMs in SCM and address the challenges that lie ahead.