Employees in Germany are adopting artificial intelligence at work faster than many employers are providing approved tools, training or rules for its use, according to a University of Konstanz survey. Of the workers who used AI, 55% said the tool they used most often had been officially introduced by their employer.
The Konstanz AI Study 2026, which surveyed 1,105 employed people in May, found that 38% used AI at work, up from 35% a year earlier. Only 24% of all respondents said their organisation provided internal or company-owned AI tools.
The figures point to a mixed model of adoption. Staff are using readily available systems to draft, research, analyse information and support routine tasks, while many employers are still deciding which services they will approve and how they should be used.
Training and rules remain scarce in small firms
The gap was widest in smaller organisations. Just 11% of workers at firms with up to 49 employees said their employer offered AI training, and 10% reported formal rules governing its use. The comparable figures were 29% and 31% at corporations with more than 1,000 employees.
Large firms were also more likely to offer internal tools. The study found that 36% of employees at corporations reported access to employer-owned or internal AI systems, against 11% in small organisations.
For employers, informal use carries two competing effects. Staff may find time-saving uses before a formal technology programme is ready. Yet an organisation may not know whether commercially sensitive material is being entered into public systems, who checks generated output, or whether workers have been trained to spot errors.
The researchers recommend clearer policies on permitted tools, information that should not be entered into them, human oversight and checks on generated material. Those measures are likely to matter most where workers are already using AI without a company rollout.
Access remains uneven across the workforce
Use differed sharply by job and education. Forty-nine per cent of staff in office and knowledge work reported using AI, compared with 25% in production-related and physical jobs. The rate was 56% among workers with a high level of education and 21% among those in the study’s lowest education category.
The survey records workers’ reported use and perceptions, rather than measured gains in productivity or job losses. Only 23% said AI was already making a noticeable contribution to their employer’s value creation. At the same time, 40% expected negative consequences for the wider labour market over the next decade, while 17% expected to lose their own job.
Those responses suggest that workplace use is spreading without a shared view of its commercial results or employment effects. The risk for smaller employers is not simply slower access to software. It is falling behind on the training and governance that determine whether employee experimentation becomes secure, repeatable work.
The study also sits alongside questions about how staff use AI once it is available. Our earlier coverage of research on AI shortcuts and managerial judgement examined why accountability and human checking can matter when technology enters routine decisions.
