Latest / 1 More Paper / Knowledge Management in Manufacturing: Current Practices, Barriers, and Automation Potential for LLM-Supported Systems † — Finkel & Wurster (2026)
Transcript
- You’re listening to “Knowledge Management in Manufacturing: Current Practices, Barriers, and Automation Potential for LLM-Supported Systems †,” by P. Finkel and P. Wurster. Published in 2026. Abstract. Knowledge management (KM) is increasingly becoming a critical success factor in Ger- many’s manufacturing industry due to demographic change, the shortage of a skilled workforce, and the growing need for flexible and resilient production systems. This study contributes empirical evidence on current KM practices in manufacturing and derives practice-oriented design implications for future LLM-supported KM systems. Two consecu- tive survey rounds involving six companies in Survey 1 and five companies in Survey 2 were conducted in order to identify current KM practices, recurring barriers, and design implica- tions for large language model (LLM)-supported KM. The results show that KM is perceived as highly relevant, but is implemented only incompletely in practice. Across both datasets, central themes such as fragmented documentation practices, reliance on interpersonal trans- fer of tacit knowledge and uneven integration of digital KM tools recur consistently. Based on the identified practices, the paper further derives areas in which LLMs may support or augment existing KM processes, particularly with regard to semantic retrieval, contextual- ization, onboarding, and the preservation of tacit knowledge. The findings also highlight that successful implementation of artificial intelligence (AI)-enabled KM in manufacturing will depend on technical feasibility, trust, usability, and organizational acceptance. 1. Introduction. The importance of knowledge management (KM) in the manufacturing industry has increased in recent years for multiple reasons. The approaching demographic shift and the current shortage of skilled workers pose fundamental challenges for Germany’s manufacturing industry. The retirement of many skilled workers is potentially leading to a major loss of know-how. A study published by the Cologne Institute for Economic Research (IW) in 2024 forecasts that 16.5 million people in Germany will reach the statutory retirement age by 2036, while only 12.5 million individuals of working age will enter the labor market over the same time. This will intensify the already existing shortage of a skilled workforce. Moreover, this means that approximately 36% of the current (as of January 2025) 45.6 million workers living in Germany will retire by 2036. The knowledge Academic Editor: Murray Jennex held by these leaving employees is crucial to the success of their companies. Without tar-geted measures, there is a high risk of losing valuable practical knowledge. In addition, the manufacturing industry is facing an increasing need for enhanced adaptability in re-sponse to constantly changing conditions. This need is driven by the growing globalization of markets, multiple global crises, the emergence of new information and communication technologies, evolving business models, and shifts in customer consumption behavior. The consequences include more complex products and processes, global and volatile value creation networks, and intensified competition along the value chain. In order to be able to respond to these inevitable changes with agility, companies are increasingly re-quired to deploy their workforce more flexibly across different production-related tasks and contexts. Given the shortage of skilled labor, this also implies that less experienced or newly assigned personnel must increasingly take on more demanding tasks. As a result, the need for effective knowledge transfer, including training existing and new employees, is gaining importance. KM is seen as a strategic response to this challenge, with the aim of preserving the organizational knowledge held by employees and ensuring its transfer to the next generation. In short, securing the knowledge and expertise of the workforce is of highest priority in order to maintain a company’s competitiveness in the face of the demographic change and skilled labor shortage. From a research perspective, this creates a timely need to better understand how manufacturing companies currently preserve, transfer, and operationalize organizational knowledge before AI-enabled KM systems can be meaningfully designed and evaluated. Since the 1990s, KM has become a widely recognized field of research and practice within companies. In academic discourse, the term “knowledge” is often explained according to its hierarchical relationship with data, information, and experience. According to Ackoff, data consists of symbols that describe the characteristics of objects, events, and their environments. When data is interpreted by means of description, it becomes information. To translate this information into actionable instructions, knowl-edge is required. Knowledge relates information to experience and a relevant con-text. This knowledge can be either extracted from personal experience or acquired via instruction from those who already possess it. In the following, the term knowledge refers to a context-related and experience-based understanding that enables action, while information refers to interpreted data and experience refers to the practical basis through which knowledge is contextualized and applied. Within the research field of knowledge, a key distinction is made between tacit (implicit) and explicit knowledge, particularly in the context of effective knowledge transfer. The undocumented and often invisible experiential knowledge held by individual employees is referred to as tacit knowledge. It is lost when employees leave the organization. At the same time, explicit knowledge which is described as knowledge documented in books, manuals, or digital repositories, may become difficult to interpret without a clear understanding of the context, conditions, and expertise associated with its original application. Earlier work has therefore devel-oped models and methods for preserving knowledge. Recent research suggests that KM is not a temporary trend but rather a long-term necessity for organizations. This is especially due to the afore mentioned demographic shift in Germany, as well as increasingly complex and information-driven business processes. While this study focuses on the German manufacturing sector, KM is increasingly recognized as a global priority. International research likewise highlights the risks of knowledge loss due to de-mographic change, digital transformation, and evolving production models particularly in industrialized nations. This underscores the relevance of examining how companies manage knowledge not only nationally, but also with awareness of international develop-ments. Existing research has therefore emphasized the importance of structured knowledge transfer, the identification of key knowledge holders, and the preservation of experiential knowledge in response to demographic change and increasing process complexity. At the same time, recent developments in the field of generative artificial intelligence (AI), particularly the emergence of powerful large language models (LLMs), are increasingly becoming the focus of current research. LLMs are AI-based systems trained on vast amounts of textual data (e.g., books, articles, websites, and wikis) enabling them to read, write, and communicate in a human-like manner. Based on user or program input, known as text-based prompts that may consist of words, letters, special characters, numbers, links, or documents, these AI-powered text generators automatically produce the requested output. As a result, earlier LLMs such as ChatGPT 3.5 by OpenAI, Inc. or DeepSeek-R1 by the Chinese start-up Hangzhou DeepSeek Artificial Intelligence Co., Ltd. have already opened up entirely new possibilities for extracting and making knowledge accessible in companies and have been continuously further developed since then. Experts predict that LLMs will serve as an interface between humans and machines in manufacturing, helping to bridge knowledge gaps. Initial approaches, such as intelligent KM tools, which combine LLMs with proprietary corporate data sources, could significantly enhance employee productivity and efficiency. While existing literature explores conceptual models for KM in the era of generative AI, there remains a lack of empirical insight into how manufacturing companies currently handle the retention of both explicit and tacit knowledge. In order to assess the potential of LLM-supported internal KM, it is essential to first analyze the current state of how organizations manage internal knowledge. To address this gap, the present study draws on direct feedback from the manufac-turing sector. Two consecutive survey rounds conducted with medium- and large-sized manufacturing companies allow a structured description of current KM practices and an assessment of whether recurring challenges, requirements, and opportunities for LLM-supported KM emerge consistently across both empirical rounds. Based on the literature background, four aspects require further empirical clarification. First, while prior research emphasizes the strategic relevance of KM, it remains necessary to understand how man-ufacturing companies currently perceive this relevance under the specific conditions of demographic change and skilled labor shortage in Germany. Second, because tacit knowl-edge is often embedded in experience-based production routines, current practices for managing and transferring such knowledge need to be examined empirically. Third, the literature points to persistent barriers in KM implementation, but it remains unclear which of these barriers are visible in current manufacturing practice and where LLMs may pro-vide meaningful support. Fourth, before developing LLM-supported KM applications, it is necessary to examine whether recurring patterns emerge across different companies and empirical rounds. Accordingly, the present study seeks to answer the following research questions: RQ1. How is the perceived importance of knowledge management in manufacturing companies in Germany, and which current needs related to KM can be inferred from the survey findings? RQ2. How do manufacturing companies currently manage organizational knowledge, particularly tacit knowledge in production-related contexts? RQ3. Which barriers to the transfer and documentation of tacit knowledge can be identified in the surveyed manufacturing companies and where could LLMs support? RQ4. To what extent do two consecutive survey rounds reveal converging findings that can inform the design of LLM-supported knowledge management applications? 2. Materials and Methods. In order to investigate the current state of KM in the manufacturing industry and to extract present-day practices, an empirical survey approach was chosen. In preparation for further research, two standardized questionnaire-based surveys were conducted. The design of the questionnaire was informed by a literature review on KM challenges in indus-trial settings, with a focus LLM-supported systems and the gap between awareness and implementation. The questionnaire consisted of four core questions. The first three were closed questions, requiring participants to provide ratings on a scale from one to ten. These questions focused on the participants’ perception of the importance of KM within a business context, the anticipated change in the relevance of KM at their own company over the next five years, and the current efforts regarding KM within their organization. The fourth question was open-ended and aimed at systematically cap-turing current KM practices within the respective companies. The open-ended responses to Question 4 were analyzed by grouping the mentioned KM practices into recurring categories. The categorization was performed inductively based on the content of the responses and was subsequently checked against the discussion notes from the collective reflection with participants. The resulting categories were not intended to represent sta-tistically generalizable constructs, but to structure recurring empirical patterns across the participating companies and to derive preliminary design implications for LLM-supported KM. The questions were reviewed and refined in collaboration with academic and industry experts to ensure content validity and practical relevance. The chosen approach provides a snapshot of current KM activities and qualitative insights into the informal and social dynamics shaping knowledge transfer in manufacturing contexts. Unlike prior works that focus on digital systems or AI applications from a technical standpoint, this study captures lived practices and perceived challenges in real-world settings, forming the basis for practice-oriented development of LLM use cases in KM. In addition to the descriptive analysis of current KM practices, the open-ended responses and the subsequent joint discussions were interpreted analytically in order to derive preliminary implications for LLM-supported KM and potential areas for the augmentation of existing KM practices. To improve the statistical transparency of the comparison between the two survey rounds, the responses to the closed questions were additionally summarized using mean and me-dian values. Means are reported to reflect the overall tendency of the ratings, while medians are included as a measure of central tendency for the ordered 10-point response scales. These descriptive indicators are used to compare the general response patterns of Survey 1 and Survey 2 across the individual responses and the company-weighted averages. Two consecutive surveys were conducted in order to compare recurring patterns across both empirical rounds and to assess whether the principal findings converged suffi-ciently to suggest recurring thematic patterns for LLM-supported KM in this domain across two empirical rounds. Survey 1 had nine experts from six medium-sized companies while Survey 2 involved ten experts from five different medium- and large-sized German companies. One company (U3/C2) was represented in both surveys but with different experts from different production facilities, thereby broadening the empirical perspective while avoiding duplicate respondent contexts. All experts gave their consent to the pub-lication of the results. However, the participants requested that no direct connections to their person or companies be made in the publication. Therefore, the individual responses to question four from the first survey are cited below using the following unique company identifiers (U1-U6) of the participating companies shown in Table 1. The individual responses to question four from the second survey are cited below using the following unique company identifiers (C1–C5) of the participating companies shown in Table 2. The companies included in Survey 1 and Survey 2 were selected through purposeful sampling based on the diversity of industries, reliance on engineering expertise, and geographic proximity. The latter enables potential future collaboration, including joint concept development and on-site testing in production environments. Common to all participating companies is their reliance on the expertise of skilled professionals such as engineers and the challenge of systematically managing this knowledge in established industrial enterprises with complex technical products. In addition, care was taken to ensure a heterogeneous expert setting across both survey rounds. The participants were between 30 and 60 years of age, thereby covering different career stages and professional perspectives relevant to production practice in Germany. Three participants (15.8%) were female and the remaining participants were male. This composition appears plausible in light of the broader German manufacturing labor market: in 2024, women held 29.1% of leadership positions in Germany, while women accounted for only 15.0% of industrial occupations such as machine operation and assembly work. Both surveys were conducted in written and anonymous form on-site in person at a university in Bavaria, Germany. The results of the open-ended fourth question were subsequently discussed collectively. The present data provide valuable insights into KM in German manufacturing companies and form the basis for future, more extensive studies on KM and the application of generative AI-based solutions in industrial KM settings. Given the exploratory character of the study, the findings should be interpreted as analytically informative rather than statistically representative. 3. Results. The following section presents the results of the Survey 1 and 2. To begin with, the responses to questions 1 to 3 are visualized and discussed using individual bar charts (Figures 1–3). Each chart displays both the individual responses of each participant (first survey in grey; second survey in green) and the weighted responses of participants from the same company (first survey in grey with dotted lines; second survey in green with dotted lines). For instance, if two individuals from the same company participated, each response is weighted at 0.5, collectively representing a single company-level response (company average). Each figure also shows the total individual responses (blue) and the total company averages (blue with dotted lines). In addition to the graphical comparison, the closed-question responses were also summarized descriptively using mean and median values for the total responses, the company-weighted averages, and both survey rounds separately. In the following subsection, the current practices identified in the open-ended fourth question are summarized and visualized in Figure 4, focusing on current KM approaches within the respective companies. The total number of mentioned current practices are colored blue while the mentions in the first survey are colored grey and in the second survey are colored green. Building on these findings, Section 3.2.5 further discusses the analytically derived potential of LLM-supported automation and augmentation with regard to the current KM practices identified across both survey rounds. 3.1. Comparison of the Three Closed Question Results of Survey 1 and 2. 3.1.1. Question 1—Assessment of the Importance of KM. All participants were asked to rate how important they consider KM will be for their company over the next five years. The results show that the vast majority regard KM as highly important. As illustrated in Figure 1, all respondents rated the importance of KM in their organization between seven and ten. This distribution of responses confirms that there is a strong awareness of the topic’s significance within the industry. The descriptive comparison suggests that the individual responses and company averages in Survey 2 were slightly more positive overall, with ratings ranging from eight to ten, whereas in Survey 1 the perceived importance was rated between seven and nine. 3.1.2. Question 2—Change in the Importance of KM. The second question asked participants to assess how they think the importance of KM is expected to change within their respective companies over the next five years. Once again, the overall sentiment is clear. Figure 2 illustrates that all participants anticipate a significant increase in the importance of KM in their organizations, with ratings ranging from seven to ten. This highlights the growing relevance of addressing KM from both a scientific and a business perspective within the manufacturing industry. Once again, the descriptive pattern suggests that participants in Survey 2 assessed the future importance of KM somewhat higher than those in Survey 1. 3.1.3. Question 3—Current Efforts in KM. In question three, participants were asked to assess their company’s current efforts regarding KM, particularly in light of the imminent retirement of numerous knowledge holders. As shown in Figure 3, the answers varied between one to seven, but the distribution of all responses still indicate a strong potential for improvement in dealing with KM. Five participants rated their company’s efforts as average (rating five), while ten assessed them as insufficient, with ratings ranging from one to four, and four as rather sufficient from six to seven. These results suggest that the implementation of KM practices differs between the companies surveyed. However, there is a strong potential for improvement in all participating companies. The responses to question three indicate that the assessments in the first survey were more widely distributed and, overall, more positive than those in Survey 2. Participants in the first survey tended to believe more strongly that the manufacturing industry is currently making sufficient efforts in KM, whereas participants in the second survey evaluated the current situation slightly more critically. This pattern may help explain why participants in Survey 2 also rated the present and future importance of KM somewhat higher than those in Survey 1. 3.1.4. Descriptive Comparison of Survey 1 and Survey 2. To complement the graphical comparison in Figures 1–3, the three closed questions were additionally summarized using mean and median values for the total individual responses, the total company-weighted averages, and both survey rounds separately. Table 3 shows the calculated values, which will be discussed below. For Question 1, both survey rounds indicate a very high perceived importance of knowledge management. At the same time, Survey 2 showed higher central tendency values than Survey 1, both on the individual level (mean 9.50 vs. 8.33; median 10 vs. 8) and on the company-weighted level (mean 9.30 vs. 8.50; both medians 9). This suggests that the already high relevance of KM was assessed even more strongly in the second survey round. For Question 2, the descriptive values were also high in both rounds and showed only minor differences. On the individual level, the mean increased from 8.11 in Survey 1 to 8.50 in Survey 2, while the median increased from 8 to 9. On the company-weighted level, the mean values were nearly identical (8.17 vs. 8.25), with the median remaining 8 in both rounds. This indicates a broadly similar assessment that the importance of internal knowledge management is expected to increase further over the next five years. For Question 3, the descriptive values were considerably lower than for Questions 1 and 2 in both survey rounds, pointing to a more critical assessment of current efforts in knowledge management. The individual means were 4.44 in Survey 1 and 3.90 in Survey 2, while the medians were 4 and 4.5, respectively. The company-weighted means were likewise close to one another (4.33 vs. 4.10). Overall, this suggests that both survey rounds show a similar critical view of the current state of KM implementation. Taken together, the descriptive comparison indicates a broadly consistent response pattern across both survey rounds: knowledge management is perceived as highly relevant, expected to become even more important in the future, and at the same time assessed as not yet sufficiently implemented in current practice. Survey 2 did not fundamentally change this overall picture, but rather reinforced the perceived relevance of KM and refined the empirical pattern observed in Survey 1. 3.2. Question 4—Current KM Practices in German Manufacturing Companies. In the open-ended fourth question, participants were asked to describe how KM is currently practiced within their companies. The responses were discussed collectively afterwards and revealed a wide range of methods, encompassing both analogue and digital solutions. While the companies are clearly pursuing multiple approaches to knowledge retention, the answers and the discussion afterwards indicated that these strategies are often fragmented or informal in practice. Across both survey rounds, 63 mentions of tools and methods were collected and grouped into the following four overarching cate-gories: documentation and databases, digital platforms and tools, mentoring and (personal) knowledge transfer, and exchange and rotation programs. Figure 4 displays the number of mentions for these tools and methods within each category. Each individual response was assigned to only one category. If a method was mentioned multiple times by respon-dents from the same company, it was counted only once per category. While some of these solutions are implemented company-wide, others represent isolated initiatives within different departments. While some solutions could plausibly fit more than one category, each item was assigned to the category that best reflected its primary organizational use in the survey context. The four categories reflect different but complementary modes of organizational KM in manufacturing. Documentation and databases primarily capture and preserve explicit knowledge in a structured and retrievable form, for example through process descriptions, instructions, logs, presentations, and formal repositories. Digital platforms and tools provide technical infrastructures for storing, sharing, and accessing knowledge across teams and departments, often with a stronger focus on collaboration, retrieval, and day-to-day usability. Mentoring and knowledge transfer represent interpersonal and experience-based forms of KM, which are particularly relevant for the transfer of tacit knowledge, practical know-how, and contextual understanding. Exchange and rotation programs, in turn, support knowledge transfer across organizational, departmental, or geographic boundaries by enabling employees to gain insights into other functions, teams, or sites. Together, these categories illustrate that KM in manufacturing does not rely on a single instrument, but on a combination of formalized documentation, digital infrastructures, and socially embedded knowledge transfer mechanisms. 3.2.1. Documentation & Databases. Documentation and databases were the most frequently mentioned knowledge re-tention methods in both surveys (27 references). For the first survey (U1–U6) more than ten references were made to technical documentation (U1, U3, U4, U6), work instructions (U2, U6), process descriptions (U1, U4, U5), or error logs (U1, U2, U3). These are used both across departments and within specific teams. In addition to traditional paper-based documentation, all companies also utilize digital databases. Both local storage systems (U1, U2, U3, U4) and central, company-wide platforms (U5, U6) were mentioned. Alongside these formal documentation practices, all companies use specific IT systems to support knowledge management at the team level. Some organizations, for example, use Enterprise Resource Planning-Systems (ERP) and Manufacturing Execution Systems (MES) not only for production control but also for storing and organizing explicit knowledge (U1, U5). The companies in Survey 2 (C1–C5) rely on similar diverse documentation practices such as project diaries (C1), problem-solution logs (C1), and project folders in print and digital form (C2) with Word, Excel or PDF-documents. Both local and centralized digital systems are used, including SharePoint-solutions and CRM databases (C3), as well as SAP-based management systems (C2, C4). Several participants highlighted the importance of structured guides and templates, such as PDF-presentations, quick guides for machine setup and maintenance (C4), and certification-related manuals combining text, images, and videos (C5). SQL-based analyses using machine data (C1) further support the integration of explicit knowledge into everyday operations. Across all surveyed companies, the results indicate a consistent emphasis on struc-tured, formalized documentation as a foundation for organizational knowledge retention. However, some of these practices are implemented in separate departments and not consis-tently across the entire organization. 3.2.2. Digital Platforms & Tools. Nineteen references described the use of digital platforms and tools for knowledge organization and sharing while six were mentioned in Survey 1 and thirteen in Survey 2. Around one third of the responses in the first survey referred to the use of digital platforms for knowledge retention. These range from internal company wikis (U3, U4) and quality-focused wiki systems (U3) to learning platforms (U3, U5) and web-based tools such as Microsoft Teams (U1, U5), or other internal collaboration solutions (U2). While some companies use structured training sessions and e-learning platforms, there are also cases where digital tools have been implemented but are not widely adopted. Several respondents reported challenges with regard to user acceptance and engagement (U1, U3). A lack of training and insufficient structuring of content were frequently cited as barriers. However, individual cases of successful implementation were also mentioned—for example, actively using information systems platforms specifically designed for production staff for the purpose of developing skills and generating ideas (U3). Examples mentioned in Survey 2 include dedicated systems such as Meusburger’s WBI software and internal database WIVIO WMS 2024 (C1), as well as more general platforms like OneNote (C1, C2) and QWiki (C2) as a structured, web-based knowledge platform that allows collaborative content creation and retrieval. Learning management systems with internally developed content (C3) and collaboration tools such as Microsoft Teams (C4) were also cited. The results show a diverse landscape of digital solutions, ranging from general-purpose office tools to specialized systems, each contributing to the accessibility and dissemination of explicit knowledge within teams. 3.2.3. Mentoring & (Personal) Knowledge Transfer. Mentoring and informal knowledge exchange were counted fourteen times. It is the second most mentioned category of the first survey and the third most mentioned of the second survey. Nearly half of all responses in Survey 1 mentioned established practices, such as mentoring programs (U4, U6), tandem models (U1, U4), or buddy systems (U2). In these approaches, new or younger employees are paired with experienced colleagues to accelerate onboarding and facilitate the exchange of personal knowledge. Such methods enable newcomers to learn from seasoned staff and ease the integration process. These programs also promote lifelong learning among experienced employees by facilitating direct exchange with recently graduated engineers, allowing them to engage with the latest theoretical—and in some cases practical—developments in their respective domains. While some of these programs are formally structured (U2, U4), others are implemented more informally within individual teams (U1, U5). Direct peer-to-peer exchange (U2) was also mentioned as a relevant method, often depending on personal relationships and interpersonal affinity. Several responses indicated that knowledge is more likely to be shared within groups where emotional or collegial bonds exist—that is, where mutual empathy is present (U1, U2, U5, U6). One company in Survey 2 described structured mentoring systems, such as workshop-based mentoring systems pairing younger employees with experienced practitioners (C1). Similarly, mentoring models linking junior and senior engineers or toolmakers (C1) were reported as an established practice to accelerate learning and preserve practical know-how. Another company (C3) described an internal expert network that connects experts from several fields within the organization and was perceived as an effective way to support the continuity of tacit knowledge. 3.2.4. Exchange and Rotation Programs. Other mentioned tools or methods in Survey 1 that could be summarized in a fourth category involved formalized programs for knowledge transfer by means of job rotation or international employee exchange (U1, U6). One company reported that employees rotate between departments or even international sites for defined periods, with the goal of sharing knowledge across different processes and gaining experience in other areas of production (U6). These programs, in some cases, span several years, enable employees to exchange their expertise across different locations, and facilitate lifelong learning by engineers. No comparable practices were mentioned in Survey 2, suggesting that exchange and rotation programs may represent complementary rather than core KM practices in the investigated domain. 3.2.5. LLM-Supported Automation Potential of Current KM Practices. Based on the current KM practices identified across both survey rounds and the subse-quent joint discussions, several preliminary areas of potential LLM-supported automation and augmentation can be analytically derived. These potentials do not yet replace existing KM mechanisms, but rather indicate where LLMs may help improve the accessibility, structuring, and practical usability of organizational knowledge in manufacturing contexts. In the category of documentation and databases, LLMs appear particularly promising for semantic search, automated summarization, question answering over technical docu-ments, and the extraction of relevant information from heterogeneous sources such as process descriptions, error logs, project folders, presentations, checklists, work instructions, and certification-related materials. In addition, LLMs may support the standardization and contextualization of existing documentation by transforming static content into more accessible, task-oriented formats. This appears particularly relevant in the present study, as documentation practices were the most frequently mentioned topic, and at the same time, the subsequent joint discussions showed that the companies surveyed often apply these practices across different departments, formats, and systems. A suitable architectural foun-dation for such LLM-supported KM systems is Retrieval-Augmented Generation (RAG), in which private company documents are indexed locally and only contextually relevant excerpts are passed to the LLM at inference time. This keeps sensitive operational data (e.g., MES/ERP records, technical documentation) within the company boundary, as no proprietary content is required to train or fine-tune the underlying model. A recent specific extension is the Document GraphRAG approach that was introduced by Knollmeyer et al. in a manufacturing context, which augments a standard RAG pipeline with a knowledge graph derived from the intrinsic structure of technical documents. This combination is particularly relevant for the fragmented documentation landscapes identified above, because it preserves cross-document relationships during retrieval and thereby improves the traceability and robustness of answers in domain-specific question answering. For digital platforms and tools, LLMs could serve as an interface layer across frag-mented knowledge systems such as SharePoint, SAP, QWiki, learning platforms, Microsoft Teams, internal databases, and specialized software environments. This may allow em- ployees to retrieve knowledge through natural-language interaction rather than navigating multiple tools separately. Moreover, LLMs may support automated tagging, classification, clustering, and the generation of user-specific knowledge views, thereby improving re-trieval performance and reducing search effort. The findings suggest that such support may be particularly useful where digital tools already exist, but are not yet fully integrated into daily work routines or are perceived as difficult to structure and use. Within mentoring and interpersonal knowledge transfer, LLMs seem especially rel-evant for supporting the elicitation and externalization of tacit knowledge. For example, guided dialogue systems may help experienced employees articulate experiential knowl-edge, recurring decision rationales, and practical problem-solving strategies in a more structured form. Such approaches could also support onboarding processes by translating expert knowledge into reusable question–answer formats, interactive learning materials, or context-sensitive assistance systems. In this respect, LLMs may complement, but not replace, interpersonal formats such as mentoring, tandem models or buddy systems, which remain central to the transfer of tacit knowledge in manufacturing. At the same time, language-based interaction and the structured elicitation of knowledge from experts may help reduce the strong dependence on personal relationships that currently characterizes many tacit knowledge transfer practices. Compared with one-to-one tandem models, such approaches may offer a more scalable alternative, as knowledge can be captured once and made available to multiple users across different teams, sites, or organizational contexts. In addition, LLM-supported systems may reduce temporal and organizational dependen-cies in knowledge transfer, as they do not necessarily require that a direct successor is already present in the company while the experienced knowledge holder is still available. Similarly, such systems may support a more location-independent form of knowledge preservation and transfer, which is particularly relevant for distributed production settings and companies with multiple sites. In relation to exchange and rotation programs, LLMs may support the transfer of knowledge across departments, roles, and sites by helping summarize local experiences, identify transferable practices, and document lessons learned in a more structured and ac-cessible way. While these programs are inherently social and experiential, LLM-supported tools could complement them by making the resulting knowledge less dependent on in-dividual memory and available to other organizational actors. Although this category appeared only in Survey 1 and therefore seems less central than the other recurring pat-terns, it still points to a relevant field of application for cross-site and cross-functional knowledge transfer. However, these potentials also imply important design requirements. For such systems to be accepted and used in practice, employees must be willing to interact with them in the course of their daily work. This suggests that low-threshold usability, intuitive interaction, and the possibility of using the system in the user’s native language may play an important role. In addition, perceived empathy in system behavior, trust in the system itself, and trust in the organizational use of AI are likely to influence adoption. The system should therefore be integrated into employees’ everyday work routines in a way that does not create substantial additional effort. At the organizational level, successful use will also depend on strategic alignment, managerial support, and possibly incentive structures that encourage knowledge sharing through the system. It should be noted, however, that these points do not constitute a complete set of requirements. Rather, they reflect preliminary considerations derived from the authors’ interpretation of the findings and from the joint discussions with the survey participants and should therefore be understood as preliminary design considerations derived from the interpretation of the findings and the participant discussions, rather than as directly measured variables of the present survey. Further research is needed to refine, validate, and prioritize these requirements in more detail. To provide a more concise overview of the results and their practical relevance, Table 4 summarizes the principal empirical findings and the corresponding implications for LLM-supported KM systems. The table shows the implications for LLM-supported KM derived from the recurring empirical patterns observed across the two survey rounds. Overall, the findings suggest that the strongest LLM potential in manufacturing KM lies not merely in automating isolated documentation tasks, but in connecting fragmented knowledge sources, supporting context-sensitive retrieval, and helping organizations preserve and operationalize tacit knowledge more systematically. At the same time, such support would need to be designed carefully in order to ensure trust, usability, contextual accuracy, and compatibility with organizational requirements regarding confidentiality and human oversight. 4. Discussion. The comparison of Survey 1 and Survey 2 reveals a consistent pattern across both empirical rounds. In both surveys, KM was rated as highly important for the participating companies and was expected to become even more relevant in the future. At the same time, the current efforts of the manufacturing industry in KM were assessed considerably more critically, indicating a clear gap between the perceived strategic relevance of KM and its actual implementation in practice. This awareness-implementation gap was evident in both surveys and was slightly more pronounced in Survey 2, in which participants assessed the present and future importance of KM more strongly while evaluating current KM efforts more critically. This pattern is consistent with previous research. A global survey conducted by Harvard Business Review Analytic Services reported that 97% of respondents considered KM to be critical, whereas only 44% believed that their organization managed knowledge effectively. The present findings suggest that this discrepancy is also visible in German manufacturing companies and may even intensify under the conditions of demographic change, increasing process complexity, and the growing need for flexibility in production. The results are also in line with earlier work indicating that KM is widely recognized as strategically important, but is still insufficiently embedded in organizational practice. In this respect, the present study confirms that the challenge is not a lack of awareness, but rather the limited translation of this awareness into structured and sustainable KM routines. The findings further show that companies apply a broad range of approaches to retain and share knowledge, combining traditional documentation with digital and interpersonal methods. Documentation and databases remain the dominant instruments for managing explicit knowledge, whereas digital platforms and tools vary considerably in their maturity, accessibility, and integration into day-to-day practice. This unevenness suggests that the digital transformation of KM is still at different stages across the surveyed companies. At the same time, tacit knowledge continues to be transferred primarily through interpersonal formats such as mentoring, tandem models, buddy systems, and expert networks. While these practices are clearly valued, they often remain dependent on personal relationships and local routines rather than being embedded in systematic organizational structures. A key contribution of the present study lies in the comparison of two consecutive sur-vey rounds. Across both surveys, the same principal patterns reappeared: a high perceived relevance of KM, insufficient current implementation, a strong reliance on documentation-based approaches, and continued dependence on interpersonal transfer for tacit knowledge. Survey 2 did not fundamentally alter the thematic structure identified in Survey 1, but rather refined it by providing additional examples of digital tools and by reflecting a somewhat more critical assessment of current KM efforts. Likewise, exchange and rotation programs appeared only in Survey 1 and may therefore be interpreted as complementary rather than central practices in this domain. Overall, the recurring thematic patterns across two empirical rounds suggest an exploratory indication of thematic convergence in the principal KM patterns observed in German manufacturing companies. Taken together, the findings provide clear answers to the study’s research questions. First (RQ1), KM is perceived as highly important in German manufacturing companies and its importance is expected to increase further in the coming years, indicating not only strategic relevance, but also a growing need for more systematic, scalable, and context-sensitive KM in production-related settings. Second (RQ2), organizational knowledge is currently managed primarily through documentation-based approaches, digital platforms, and interpersonal knowledge transfer formats, with tacit knowledge still being handled largely through personal exchange and experience-based interaction. Third (RQ3), the findings reveal several barriers to the systematic transfer and documentation of tacit knowl-edge, including fragmentation of KM practices, dependence on individual relationships, the limited scalability of one-to-one transfer formats, incomplete integration of digital tools into daily routines, and varying levels of user acceptance and structure. Fourth (RQ4), the comparison of the two consecutive survey rounds reveals a convergence of findings, suggesting that the principal patterns identified are sufficiently stable to inform the design of LLM-supported KM applications in this domain. From the perspective of AI-enabled knowledge management, these findings are par-ticularly relevant (RQ3). They suggest that the most promising role of LLM-supported systems in manufacturing is not merely the storage of explicit knowledge, but the con-textual retrieval, structuring, and accessibility of knowledge that is currently dispersed across documents, systems, and human experts. In particular, LLM-based applications may help bridge the gap between formal documentation and the practical use of organizational knowledge by supporting semantic search, interactive question answering, onboarding, and the elicitation of tacit knowledge through guided dialogue. At the same time, the results underline that successful implementation will depend on technical feasibility, but also on trust, usability, and organizational acceptance. In this sense, LLM-supported KM in manufacturing should be understood as a socio-technical design challenge rather than merely a software deployment task. The more detailed analysis of current KM practices further indicates that several established tools and methods may be meaningfully supported or augmented through LLM-based functionalities. In particular, documentation repositories, error logs, work instructions, learning platforms, internal wikis, and mentoring-supported onboarding processes appear to offer considerable potential for semantic retrieval, contextualization, summarization, and more structured access to organizational knowledge. However, the results suggest that the value of LLMs in this context lies less in fully automating isolated documentation tasks than in connecting fragmented knowledge sources, improving acces-sibility, and supporting the systematic elicitation and operationalization of tacit knowledge. This is particularly relevant for tacit knowledge transfer. Language-based interaction and the structured elicitation of knowledge from experts may help reduce the strong depen-dence on personal relationships that currently characterizes many tacit knowledge transfer practices. Compared with one-to-one tandem or mentoring models, such approaches may offer a more scalable alternative, as knowledge can be captured once and subsequently made available to multiple users across different contexts. In addition, LLM-supported systems may reduce temporal and organizational dependencies in knowledge transfer. Similarly, such systems may support a more location-independent form of knowledge preservation and transfer, which is particularly relevant for distributed production settings and companies with multiple sites. At the same time, these considerations should not be interpreted as a complete or final set of requirements for LLM-supported KM in manufacturing. Rather, they reflect preliminary design implications derived from the authors’ interpretation of the findings and the joint discussions with survey participants. Further research is therefore needed to refine, validate, and prioritize these requirements in more detail. In this regard, the present study can be understood as an empirical problem-analysis and requirement-identification step for the future design of LLM-supported KM solutions in manufacturing. From a broader methodological perspective, future work may benefit from a research perspective, in which the empirically identified needs, barriers, and LLM-related opportunities are translated step by step into the design, demonstration, and evaluation of practical KM artefacts. Despite these contributions, the study has limitations. The empirical basis is limited to two survey rounds with a comparatively small number of overall participants, which restricts the generalizability of the findings. In addition, the study relies on self-reported assessments and descriptions of KM practices, which may be influenced by individual perceptions, selective recall, and differing interpretations of what constitutes effective KM. At the same time, this limitation should be interpreted in light of the study design and access strategy. Although the number of individual participants was objectively small, the surveys covered more than ten different manufacturing companies and included experts from higher hierarchical levels, in some cases from C-level or top management positions. This constellation is particularly relevant, as managerial support is essential for ensuring organizational acceptance of KM initiatives and for enabling subsequent company-internal studies with broader employee samples. In this sense, the current study provides not only exploratory empirical insights, but also an important strategic foundation for future research and implementation. Nevertheless, the anonymization of company information limits the possibility of linking specific practices to more detailed organizational contexts. Future research should therefore build on this managerial-level access by extending the empirical base to larger samples within the participating companies and by translating the present findings into specific design requirements for LLM-supported KM tools in manufacturing. In particular, further work is needed to examine which organizational, technical, and human-centered conditions enable the successful adoption of such systems. Relevant questions include how tacit knowledge can be elicited and represented more systematically, how industrial knowledge sources can be integrated into reliable LLM-based applications, and which governance mechanisms are required to address issues of transparency, privacy, and trust. 5. Conclusions. By combining two consecutive survey rounds, this study provides a broader empirical basis for assessing current KM practices in German manufacturing companies. Across both surveys, the findings consistently point to a high perceived importance of KM, a strong and growing need for more systematic knowledge preservation in production-related settings, a persistent gap between strategic relevance and practical implementation, and a strong reliance on documentation-based approaches for preserving knowledge. At the same time, tacit knowledge is still transferred primarily through interpersonal formats such as mentoring, tandem models, and expert exchange, which are valuable but difficult to scale systematically. The comparison of both survey rounds further indicates a broad convergence of results. The principal categories of current KM practice remained stable across both datasets, while Survey 2 mainly refined rather than fundamentally changed the findings of Survey 1. This suggests a stability of principal patterns within the investigated domain and provides a sufficiently consistent empirical basis for subsequent work on AI-enabled KM. The findings also provide clear answers to the study’s research questions. KM seems to be perceived as highly important and increasingly relevant in German manufacturing companies, while current organizational KM still relies heavily on fragmented documen-tation, digital platforms of varying maturity, and interpersonal forms of tacit knowledge transfer. The results further show that barriers to the systematic preservation of tacit knowl-edge include dependence on individual relationships, the limited scalability of one-to-one transfer formats, incomplete integration of digital tools into daily routines, and the need for stronger organizational support. From the perspective of AI-enabled KM, the results indicate clear implications for future system design. LLM-supported solutions appear particularly promising where companies need contextual access to fragmented organizational knowledge, improved retrieval across heterogeneous sources, support for onboarding and mentoring, and more systematic ways of preserving tacit engineering knowledge. At the same time, the findings suggest that the most valuable contribution of LLMs lies not in replacing human expertise, but in augmenting existing KM practices through context-sensitive retrieval, structured elicitation, and more accessible interaction with organizational knowledge. The next step in research should therefore be the development of an overall concept for LLM-supported knowledge management in industrial companies. Given the complexity of such systems, this concept will likely need to be broken down into manageable sub- functions or application modules, which can then be designed, implemented, and evaluated step by step under real-world conditions and with larger samples. In this way, individual components of LLM-supported KM can be iteratively refined and optimized in practice. Beyond this development path, future research should also examine the organizational and human-centered conditions of successful adoption, including trust, managerial support, usability, and organizational acceptance. In this sense, the present study provides an empirical foundation for a subsequent design-oriented research phase toward intelligent and systematic KM in manufacturing, which is intended to contribute to the long-term competitiveness of the manufacturing industry. formal analysis, P.F.; investigation, P.F. and P.W.; resources, P.W.; data curation, P.F.; writing—original draft preparation, P.F.; writing—review and editing, P.F. and P.W.; visualization, P.F.; supervision, P.W.; project administration, P.W. All authors have read and agreed to the published version of the manuscript. Funding: This research received no external funding. Data Availability Statement: The data presented in this study are available on request from the corresponding author due to privacy reasons and specific non-disclosure agreements. Conflicts of Interest: The authors declare no conflicts of interest. Abbreviations The following abbreviations are used in this manuscript: KM Knowledge Management LLM Large Language Models AI Artificial Intelligence ERP Enterprise Resource Planning MES Manufacturing Execution Systems RAG Retrieval-Augmented Generation 1. Finkel, P.; Wurster, P.; Radler, R. Large language models (LLM) in production environments. Ind. 4.0 Sci. 2024, 40, 48–55. [CrossRef] Bettiol, M.; Di Maria, E.; Micelli, S. Industry 4.0 and knowledge management: An introduction. In Knowledge Management and 2. Industry 4.0; Bettiol, M., Di Maria, E., Micelli, S., Eds.; Springer: Cham, Switzerland, 2020; pp. 1–18. [CrossRef] 3. Deschermeier, P.; Schäfer, H. Die Babyboomer Gehen in Rente; IW-Kurzbericht 78/2024. Available online: the linked source (accessed on 9 April 2026). 4. Statistisches Bundesamt. Press Release No. 077: Employment Slightly Declined in January 2025. Available online: the linked source (accessed on 9 April 2026). 5. Schiedermair, I.; Kick, E.; Baumgartner, M.; Kopp, T.; Kinkel, S. Wissensmanagement in KMU: Kriterien zur Identifikation interner Schlüsselpersonen. ZWF 2023, 118, 395–399. [CrossRef] Rülicke, S. Prozessintegriertes Wissensmanagement—Ein Lösungsansatz auf den demografischen Wandel. In Praxis der Wirtschaft-6. spsychologie III: Themen und Fallbeispiele für Studium und Anwendung; Mehlich, P., Brandenburg, T., Thielsch, M.T., Eds.; Monsenstein und Vannerdat: Münster, Germany, 2014; pp. 249–263. 7. Balzer, V. Strategische Planung von Produktionskompetenzen. Ph.D. Thesis, Universität Stuttgart, Stuttgart, Germany, 2024. [CrossRef] Ackermann, B. Strukturierte Wissenstransfers—enormes Potenzial für Wirtschaft und Verwaltung. In Erfolgreicher Wissenstransfer 8. in Agilen Organisationen: Hintergrund—Methodik—Praxisbeispiele, 2nd ed.; Springer Gabler: Wiesbaden, Germany, 2020; pp. 29–44. [CrossRef] 9. Alavi, M.; Leidner, D.E. Review: Knowledge management and knowledge management systems: Conceptual foundations and research issues. MIS Q. 2001, 25, 107–136. [CrossRef] Author Contributions: Conceptualization, P.F. and P.W.; methodology, P.F. and P.W.; validation, P.F.; 10. Ackoff, R.L. From data to wisdom. J. Appl. Syst. Anal. 1989, 16, 3–9. 11. zur Heiden, P.; Kaltenpoth, S. Wissensmanagement für Wartung und Instandhaltung im Verteilnetz—Konzeption eines Assisten-zsystems basierend auf einem Large Language Model. HMD 2024, 61, 911–926. [CrossRef] 12. Finkel, P.; Wurster, P. Unlocking Tacit Knowledge in Industrial Production: Exploring Barriers, Practices, and LLM-Driven Potentials for Knowledge Management. In Proceedings of the 59th Hawaii International Conference on System Sciences (HICSS), Maui, HI, USA, 6–9 January 2026. Available online: the linked source (accessed on 10 April 2026). 13. O’Leary, D.E. Large language models and applications: The rebirth of enterprise knowledge management and the rise of prompt libraries. IEEE Intell. Syst. 2024, 39, 72–75. [CrossRef] 14. Storey, V.C. Knowledge management in a world of generative AI: Impact and implications. ACM Trans. Manag. Inf. Syst. 2025, 16, 26. [CrossRef] 15. Hadi, M.U.; Al Tashi, Q.; Qureshi, R.; Shah, A.; Muneer, A.; Irfan, M.; Zafar, A.; Shaikh, M.B.; Akhtar, N.; Wu, J.; et al. Large language models: A comprehensive survey of its applications, challenges, limitations, and future prospects. TechRxiv 2024, TechRxiv:23589741.v3. [CrossRef] 16. Bendel, O. Prompt. Gabler Wirtschaftslexikon. Available online: the linked source (accessed on 9 April 2026). 17. Yoon, A. Why large Language Models Are the Future of Manufacturing. World Economic Forum. Available on-line: the linked source (accessed on 9 April 2026). 18. Finkel, P.; Wurster, P. Analysis of the Current State and Best Practices of KM Applications in the Manufacturing Industry. In Proceedings of the 19th International Conference Inter-Eng 2025: Interdisciplinarity in Engineering; Moldovan, L., Gligor, A., Eds.; Lecture Notes in Networks and Systems; Springer: Cham, Switzerland, 2026; Volume 1813. 19. Rank, Y.; Streloke, L.; Bründl, P.; Bodendorf, F.; Franke, J. Large Language Models for Tacit Knowledge Elicitation in Industry 5.0: A Literature Review. In The Human Side of Service Engineering; AHFE Open Access: New York, NY, USA, 2025; Volume 182, pp. 36–46. [CrossRef] 20. Statistisches Bundesamt (Destatis). Frauen in Führungspositionen; Destatis: Wiesbaden, Germany, 2025. Avail-able online: the linked source (accessed on 10 April 2026). 21. Statistisches Bundesamt (Destatis). Erwerbsbeteiligung von Frauen Nach Berufen; Destatis: Wiesbaden, Germany, 2026. Avail-able online: the linked source (accessed on 10 April 2026). 22. Knollmeyer, S.; Caymazer, O.; Grossmann, D. Document GraphRAG: Knowledge Graph Enhanced Retrieval Augmented Generation for Document Question Answering Within the Manufacturing Domain. Electronics 2025, 14, 2102. [CrossRef] 23. Harvard Business Review Analytic Services. Unlocking Gen AI’s Potential by Strengthening Data and Organizational Readiness; White Paper. Harvard Business Review Analytic Services, 2024. Available online: the linked source (accessed on 9 April 2026). 24. Fischer, V. Zeit, zu Handeln: Wissensmanagement Ist Kein Luxus, Sondern eine Notwendigkeit. Modell Aachen Insights. Available online: the linked source (accessed on 9 April 2026). Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.