One example of this is LLMs, which have been specially developed to generate human-like speech. These models are trained using huge amounts of data and use techniques such as unsupervised learning to learn the patterns of human language. However, they are unable to draw logical conclusions or fully understand complex causal relationships. In addition, they often struggle to verify facts, comprehend emotions or make ethical decisions because they lack conscious experience or moral judgment [4].<\/p>\n\n\n\n
![\"Classification](\"https:\/\/industry-science.com\/wp-content\/uploads\/2024\/11\/Wurster_I4S-24-6_Bild-1-1024x486.jpg\")
Modern LLMs represent the state of the art in natural language processing, in that they are able to interpret, generate and adapt human-like text. These models are hugely versatile: They can, for example, be used for text summarization and generation or for programming support [4]. Figure 1 shows the classification of LLMs in the context of AI.<\/p>\n<\/div>\n<\/div>\n\n\n\n
LLMs are already being used in an industrial context: For contract analysis during procurement, for the automation of customer service, for software documentation and troubleshooting and for error correction in additive manufacturing. Another example of use is the structuring and establishment of knowledge databases, in which large amounts of text are efficiently analyzed and relevant information extracted [4, 7]. Despite their widespread industrial use, their application in production remains limited. This study is the first to examine the potential of LLMs in the production environment in a targeted and cross-company manner, independent of specific providers, and sheds light on their potential value for the digitalization of production processes<\/a>.<\/p>\n\n\n\n As part of the potential analysis described above, 13 experts from three medium-sized companies were selected from the manufacturing industry in German-speaking countries. The companies were selected on the basis of their geographical location, their commitment to innovation and their willingness to introduce new technologies with the aim of providing a personal and qualitative reflection of manufacturing companies. The following anonymized companies were interviewed in the course of this study:<\/p>\n\n\n\n In order to ensure a general overview of the companies\u2019 perspectives on the potential of LLMs, different hierarchical levels of the same companies were specifically included in the interviews. The selection of the 13 interview partners from their respective companies is therefore based on their operational role and expertise in production processes, covering various functions and hierarchical levels. The following were selected: five department managers, one segment manager, two shift supervisors, three trainees, one trainer and one quality inspector. This includes one woman and twelve men aged between 18 and 50.<\/p>\n\n\n\n The semi-structured guided interviews were conducted at the respective companies. In this way, individual, authentic insights into the experiences of the employees could be gained. The method combined fixed and spontaneous questions within the framework of a guideline, which created space for personal perspectives and flexible reactions to individual responses [8]. The flexibility of this method was crucial for identifying and understanding individual potentials as well as specific challenges that can be addressed through LLMs.<\/p>\n\n\n\n All interviews were conducted on site at the respective companies between April and June 2024. This allowed for the observation of both verbal and non-verbal signals, increasing the quality of the data [9]. Verbal consent was obtained before each interview, following an explanation of the objectives, procedure and the participants\u2019 confidentiality. All interviews were recorded and carefully transcribed to enable detailed data analysis. <\/p>\n\n\n\n The data analysis was carried out using Mayring\u2019s method of qualitative content analysis, which makes it possible to take into account both the specific context of the data generation and the uniqueness of the communication [8]. This method follows a structured procedure with fixed rules for categorizing text passages, which ensures a systematic and comprehensible analysis of the collected data [10]. In order to identify central themes, the original material was converted into central paraphrases via a summarizing content analysis and systematically condensed [8].<\/p>\n\n\n\n This was followed by a structuring content analysis in which deductive categories were applied to the text in order to structure the data [10]. The combination of inductive and deductive approaches utilizes the advantages of both to potentially gain new insights directly from the data as well as to validate existing theories [11]. This dual methodology ensured that the analysis both delivered individual ideas and contributed to the development of detailed hypotheses. For reasons of data protection, the participating companies were anonymized and further artifacts of the study were not published.<\/p>\n\n\n\n The analysis of the interviews provided individual observations into the central aspects of employee satisfaction<\/a>, motivation and loyalty. Five main factors for the motivation (or frustration) of production employees and thus potential main areas of application for LLMs in the production environment were identified and weighted based on the data. Figure 2 shows a correspondingly weighted depiction of these factors, which are explained in more detail below. The weighting was determined based on how frequently that factor or a closely related one was mentioned in the interviews. The factors were sorted in descending order of importance based on the number of mentions.<\/p>\n\n\n\n The interview participants named communication <\/strong>as a central factor contributing to the motivation (or frustration) of production employees.<\/p>\n\n\n\n Statements in this category focused on the quality and efficiency of communication within the company. They included the need for clear and open communication, the handling of communication problems, the role of feedback processes and the overcoming of language barriers.<\/p>\n<\/div>\n<\/div>\n\n\n\n The interviewees opined that inadequate or inaccurate communication results in information deficits, misunderstandings and uncertainties, which have a negative impact on morale and employee satisfaction. One main potential of LLMs could therefore lie in addressing these challenges, particularly with regard to information deficits, process communication, language and cultural barriers, cross-hierarchical communication and systems of feedback and recognition.<\/p>\n\n\n\n Another overarching category that was identified was technology and processes.<\/strong> This category refers to the use and integration of technology and the efficiency of work processes. It includes digitalization and the implementation of modern technologies, approaches for dealing with resistance to these new technologies, frustrations due to inefficient processes and the associated administrative effort. The efficiency and precision of production processes play a central role here.<\/p>\n\n\n\n Challenges associated with manual, time-consuming tasks, inefficient knowledge management systems and difficulties in implementing new technologies<\/a> were highlighted. Therefore, there appears to be potential for optimized knowledge management and the automation of processes to reduce employee frustration and significantly increase productivity.<\/p>\n\n\n\n The third field of application can be summarized under the term further training and induction<\/a><\/strong>. Statements that could be assigned to this category regarded the process of training new employees as well as the opportunities for further training and development within the company. Further training opportunities were cited as crucial for employee motivation, with the analysis suggesting considerable potential for improvement. A standardized and professional training structure could thus significantly improve employee efficiency and satisfaction.<\/p>\n\n\n\nIntroducing the interview partners and applied research methodology<\/h2>\n\n\n\n
\n
Communication as the greatest LLM potential in production<\/h2>\n\n\n\n
![\"Weighting](\"https:\/\/industry-science.com\/wp-content\/uploads\/2024\/12\/Wurster_I4S-EN-24-6_Figure-2-1024x383.jpg\")