Artificial intelligence (AI) has a high application potential in many areas of the economy, and its use is expected to accelerate strongly in the coming years. This is linked with changes in working conditions that may be substantial and entail serious health risks for employees. With our paper we are the first to conduct an empirical analysis of employers' increasing flexibility requirements in the course of advancing digitalization, based on a representative business survey, the IAB Job Vacancy Survey. We combine establishment-level data from the survey and occupation-specific characteristics from other sources and apply non-linear random effects estimations. According to employers' assessments, office and secretarial occupations are undergoing the largest changes in terms of flexibility requirements, followed by other occupations that are highly relevant in the context of AI: occupations in company organization and strategy, vehicle/aerospace/shipbuilding technicians and occupations in insurance and financial services. The increasing requirements we observe most frequently are those concerning demands on employees' self-organization, although short-term working-time flexibility and workplace flexibility also play an important role. The estimation results show that the occupational characteristics, independently of the individual employer, play a major role for increasing flexibility requirements. For example, occupations with a larger share of routine cognitive activities (which in the literature are usually more closely associated with artificial intelligence than others) reveal a significantly higher probability of increasing flexibility demands, specifically with regard to the employees' self-organization. This supports the argument that AI changes above all work content and work processes. For the average age of the workforce and the unemployment rate in an occupation we find significantly negative effects. At the establishment level the share of female employees plays a significant negative role. Our findings provide clear indications for targeted action in labor market and education policy in order to minimize the risks and to strengthen the chances of an increasing application of AI technologies.