Implementing AI in SMEs - The availability of training data and promotion of open data structures

Artificial intelligence (AI) is finding its way into more and more business areas. More than 30% of the companies in a survey conducted by the industry association Bitkom state that they are currently using AI, are planning to use it, or are at least considering doing so. For just under two-thirds of companies, however, the use of AI is still not on the agenda. Ideas about AI are often characterized by misunderstandings. The requirements and areas of application of various AI applications for companies are as diverse as the tasks and fields of activity of SMEs. Completely new business models and disruptive innovations such as autonomous driving represent only the tip of the broad AI spectrum in practice.

The aim of the study is to establish theoretical foundations and how SMEs can be supported in the implementation and use of AI applications, especially with regard to the necessary data. However, it also shows that the topic of data availability cannot be examined in isolation and that anchoring and raising awareness of this topic in all areas of the company is fundamental. In fact, data is usually available, even if only in unstructured and distributed form, or can be collected with relatively little effort. Here, the specific challenge often lies in preparing existing data and breaking down data silos in order to make the data usable for further exploitation in the context of AI applications. Not least at this stage, however, companies frequently fail to benefit from AI due to an insufficient level of digital maturity and understanding.

In addition to literature research, the study is based on an exploratory online survey of SMEs and IHK consultants with a focus on data availability in the context of AI applications and a case study analysis. Furthermore, a validation of the results took place in interviews with AI experts from the Mittelstand-Digital network.