There is a broad consensus on the potential of smart services for production and the added value their use offers. Industrial artificial intelligence (AI) has several advantages. AI technologies, for example, can strengthen resilience, support work processes, increase product quality and thus improve competitiveness. Many companies have recognised these potentials and are developing AI solutions. There are many successful proof-of-concepts (PoC) and pilot projects, but AI technologies successfully implemented in the real environment are scarce. Successful implementation of smart services based on industrial AI in production operations can be understood as its repetitive use and integration into operational business, which is a prerequisite for exploiting the potentials. Currently, little is known about how to achieve successful implementation. In contrast, there is much evidence that the implementation and operation of AI in manufacturing is associated with extensive challenges and barriers. The factors that positively influence the roll-out of AI technologies in manufacturing, however, are little explored. Therefore, this paper focuses on the identification of success factors and barriers for the implementation and operation of AI solutions in manufacturing. Furthermore, it is analysed whether and how the identified success factors and barriers differ from each other in order to subsequently derive initial recommendations for action. The methodology is based on explorative qualitative research. First, 10 semi-structured interviews were conducted with AI experts from a German Original Equipment Manufacturer (OEM). In an expert workshop, the main findings were validated, and possible solution and support options were discussed. Our findings confirm the results found in the literature and complement them with new insights. Success factors and challenges can be found on the technical, organisational, and human side and relate most often to "data", "development and operational processes" and "stakeholder engagement".

No abstract

The development of Human-Centered and Trustworthy AI-based services has recently attracted increased attention in politics and science. Even though that technical advances have received many of the attention lately, ethical considerations are becoming more and more important. One of the most valuable publications in this area is the "Ethics Guidelines for Trustworthy AI" of the European Commission (EC). One approach to assist developers in implementing these requirements during the development process is to provide design guidelines. The aim of this paper is to identify which action-oriented design principles can be applied to satisfy the requirements for Trustworthy AI. For this purpose, the design principles published by Major providers of commercial AI-based services were contrasted with the seven requirements of the EC. The results indicate that some design principles can be used to meet the requirements of Trustworthy AI. At the same time, however, it becomes clear that work on Ethical AI should be extended by aspects related to Human-AI Interaction and service process quality.