Use of artificial intelligence by companies in Germany

Prevalence and enabling factors

Christian Gerhards, Myriam Baum

Against the background of current breakthroughs in machine learning and in light of the reignited debate regarding the substitutability of human work, data from the BIBB Establishment Panel on Training and Competence Development (BIBB Training Panel) is being used to investigate the extent to which the use of artificial intelligence (AI) by companies in Germany has increased over recent years and to examine in which companies AI is used particularly frequently.

AI in the world of work

Everyone is now talking about AI, especially following the launch of ChatGPT-3.5 at the end of 2022.* Machines and computers, which are capable of independent work and learning and to process huge quantities of data, have considerable potential to change the world of work (cf. e.g.  Sevindik 2022; Acemoglu/Restrepo 2019 and 2020) as well as to possibly replace human labour. The debate surrounding the substitutability of human work is by no means new. However, previous analyses do not give any reason to assume widespread displacement of human work in Germany in the near future (cf. e.g. Bonin et al. 2015; Helmrich et al. 2016; Dengler/Matthes 2021; Schneemann et al. 2021). There has, nevertheless, now been a shift in the discussion with regard to the nature of tasks that are replaceable. AI is increasingly able to take over tasks which older research studies viewed as being less suited to substitution (e.g. certain cognitive tasks; cf. e.g. Acemog-Lu/Restrepo 2019 and 2020). In order to be able to evaluate the consequences for the human world of work, the current extent of AI use in the workplace or at companies in Germany must first be clarified.

Representative results for all companies in Germany are thus far hardly available. Analyses undertaken with the Mannheim Innovation Panel in 2019 conclude that use amongst the companies surveyed – primarily from manufacturing and company-related services– is around six percent (cf. BMWi 2020; Sevindik 2022). Sevindik (2022) initial analyses with the BIBB Training Panel show that approximately three to four percent of all companies in Germany were using AI in 2019 resp. 2020. Evaluations from an employee survey carried out in 2019 (DiWaBe) show that still nearly 90 percent of employees have little or nothing to do with AI, but the prevalence of its use in the workplace is increasing (cf. ibid. 2022).

AI applications have undergone significant further development since these assessments. In order to continue these analyses for the latest developments in a rapidly changing field, we use the most current data from the BIBB Training Panel to investigate in the following the prevalence of AI at companies in Germany between 2020 and 2022. We then use regression analysis to observe which companies are making greater use of AI or are at least planning to use AI.

Database and methodological approach

The analyses are based on data from the BIBB Establishment Panel on Training and Competence Development (referred to as BIBB Training Panel, cf. Information Box). In order to measure the use of AI, two separate items were included from the survey waves 2020 to 2022 onward in which companies were asked whether they use the following digital technologies.

  • Artificial intelligence and machine-based learning for physical work processes, e.g. deep learning and pattern recognition in production, maintenance, buildings management or upkeep
  • Artificial intelligence and machine-based learning for non-physical work processes, e.g. deep learning and pattern recognition in marketing, procurement or human resources

Three possible responses are differentiated, weather these technologies are being used (1), acquisition is planned (2) or no use is being planned (3).

For the analysis, the answers of the two items were combined as: no AI is being used; use of AI is at least planned in at least one instance; AI is being used in at least one instance.

BIBB Training Panel

The BIBB Establishment Panel on Training and Competence Development (BIBB Training Panel) is a regular annual representative survey of all companies in Germany which has been conducted since 2011. Between 3,500 and 4,000 companies are included each time (cf. Friedrich/Gerhards 2023). The main emphases of the survey are on vocational education and training (VET) and continuing training. Moreover, changing focus modules are surveyed.  Digitalisation processes at companies have been a more detailed object of consideration since 2016. 

Further information on the survey is available at www.bibb.de/qp.

Prevalence of AI by companies in Germany

Figure 1: Deployment and planning of the use of AI by companies 2020 to 2022 Foto-Download (Bild, 203 KB)

Figure 1 shows that use of AI by companies in Germany continuously increased from about three percent in 2020 to around five percent in 2022. The additional enquiry regarding whether acquisition of AI was being planned in future was positively answered by approximately five percent of companies each year.

Company structural characteristics and company use of AI

Figure 2: Factors which influence the planning and use of artificial intelligence by a company (2022)
Foto-Download (Bild, 513 KB)

Even though the number of companies using AI is still very low, the question arises, which company characteristics correlate with the use of AI. For this reason, an ordinal regression model was calculated based on the data from the 2022 wave. With the model we investigate which factors exhibit a positive correlation with the use or at least planned use of AI as opposed to non-use (cf. Figure 2 and electronic supplement). The advantage of this ordinal logistic regression model is that it facilitates investigation of planning and use of AI versus non-use in a two-stage model, while controlling for several other influencing factors. The characteristics considered were: company structural characteristics such as company size; sector and chamber membership; location in eastern or western Germany; company involvement in VET and continuing training (proportion of continuing training participants, provision of VET); company qualification level (proportion of employees performing low-skilled tasks); status of digitalisation (simple summation of technology use) –without the use of AI. The selected variables are aligned to the studies made by Sevindik (2022) and by the BMWi (2020). 

One finding shown in Figure 2 is that planning (green) and actual use (blue) frequently lie close together in terms of likelihood. Thus, companies using AI and companies planning its use are quite similar. A clear correlation is as well revealed with regard to company size. The larger the company, the more likely it will use or plan to use AI. Differentiation according to sector shows that AI is less prevalent in the sectors trade and repairs, medical services and public sector and education compared to manufacturing industry (reference category). A further positive correlation is revealed in the level of use of other digital technologies at the company (highly digitalised companies are more likely to use AI). Overall qualification level at the company, measured by the proportion of employees performing low-skilled tasks, also exhibits a clear correlation. The lower the average qualification level of tasks carried out by employees, the less likely is the use of AI. No clearly aligned or significant correlations emerge for the further variables (cf. regression table in the electronic supplement).

Growth potential for use of AI by companies and further research needed

The results over the course of time since 2020 show that the use of AI by companies in Germany is currently increasing. Although the use of AI in absolute terms is still at a low level, considerable potential is revealed for short-term further growth to at least double the amount of use.

A comparison between companies in Germany in 2022 indicates that the particular current pioneers in use of AI are bigger companies, companies which use a large amount of technology, and companies in the non-public and non-medical sector. In addition, the use of AI appears to either require the employment of people with higher-skilled jobs or, conversely, employees with higher-skilled jobs enable the use of AI, as the employment of people with simple jobs has a negative impact on the use of AI. Further analyses will be needed to explore which precise correlations exist between AI and training at the company, e.g. how the use of AI correlates with VET and continuing training requirements or how the use of AI and the company’s task or qualification structures are mutually dependent.


Acemoglu, D.; Restrepo, P.: The wrong kind of AI? Artificial intelligence and the future of labour demand. In: Cambridge Journal of Regions, Economy and Society, 13 (2020) 1, pp. 25-35. URL: https://doi.org/10.1093/cjres/rsz022

Acemoglu, D.; Restrepo, P.: Artificial intelligence, automation, and work. In: The economics of artificial intelligence: An agenda. Chicago, 2019. pp. 197-236

BMWi (Bundesministerium für Wirtschaft und Energie): Einsatz von Künstlicher Intelligenz in der Deutschen Wirtschaft – Stand der KI-Nutzung im Jahr 2019. Berlin 2020

Bonin, H.; Gregory, T.; Zierahn, U.: Übertragung der Studie von Frey/Osborne (2013) auf Deutschland. ZEW Kurzexpertise (2015)

Dengler, K.; Matthes, B.: Folgen des technologischen Wandels für den Arbeitsmarkt: Auch komplexere Tätigkeiten könnten zunehmend automatisiert werden. IAB-Kurzbericht 13 (2021). URL: https://doku.iab.de/kurzber/2021/kb2021-13.pdf

Friedrich, A.; Gerhards, Ch.: BIBB-Qualifizierungspanel 2021. Version 1.0. Bonn 2023. URL: www.bibb.de/dienst/publikationen/de/download/19194

Helmrich, R.; Tiemann, M.; Troltsch, K.; Lukowski, F.; Neuber-Pohl, C.; Lewalder, A. Ch.; Güntürk-Kuhl, B.: Digitalisierung der Arbeitslandschaft en. Keine Polarisierung der Arbeitswelt, aber beschleunigter Strukturwandel und Arbeitsplatzwechsel. Bonn 2016. URL: www.bibb.de/dienst/publikationen/de/download/8169

Sevindik, U.: Verbreitung und Einsatz von Künstlicher Intelligenz in Deutschland: Auswirkungen auf berufliche Anforderungen und Strukturen. Version 1.0 Bonn 2022. URL: https://res.bibb.de/vet-repository_780476

Schneemann, Ch.; Zika, G.; Kalinowski, M.; Maier, T.; Krebs, B.; Steeg, St. u. a.: Aktualisierte BMAS-Prognose »Digitalisierte Arbeitswelt«. Forschungsbericht: 526/3 (2021)


Electronic Supplement

Regression table available as an electronic supplement at www.bwp-zeitschrift.de/dienst/publikationen/en/material/12201


(All links: status 02/04/2024)

Dr. Christian Gerhards
Researcher at BIBB 

Myriam Baum
Researcher at BIBB 


Translation from the German original (published in BWP 1/2024): Martin Kelsey, GlobalSprachTeam, Berlin