Botsitting: The Surprising Workload That’s Undermining Germany’s AI Push
02.07.2026 - 06:25:10 | boerse-global.de
Office workers in Germany now spend an average of 6.4 hours per week correcting and supervising the work of artificial intelligence systems, a burden that undermines the promised productivity gains of the technology.
Data from the software company Glean reveals that this “botsitting,” as it is called, has become a significant time sink. The finding was echoed by AI strategist Sol Rashidi, who recently reported deactivating several AI agents because they required constant human oversight instead of delivering time savings.
Jobs Grow, Not Shrink, in High-AI Firms
Adding to the paradox is a long-term study by Ramp and Revelio Labs, which tracked roughly 22,000 U.S. companies from January 2021 through February 2026. Companies that invested most heavily in AI expanded their workforces by an average of 10.2 percent within two years. At entry-level positions, the increase was even steeper at 12 percent. The authors expressed skepticism toward executives who cite AI as a reason for layoffs, though the growth is concentrated in the information sector, signaling shifting skill demands.
Adoption Spreads, but Depth Remains Shallow
The push into AI is widespread but shallow. According to the “AI Potential Germany 2026” study by Amazon Web Services and Strand Partners, 63 percent of German companies now use artificial intelligence – up from 53 percent the previous year. Yet only 15 percent of businesses are fundamentally redesigning their processes, a drop from 21 percent a year earlier. The majority rely on simple applications.
This gap is particularly visible in agentic AI – systems that act autonomously. Only 4 percent of companies have fully deployed such agents, with 12 percent running pilots.
Tech Giants Unleash New Agent Ecosystems
Major software vendors have dramatically expanded their AI offerings since July 1. Microsoft made Copilot Cowork and the Agent for Copilot generally available, introducing consumption-based billing via a credit system: each task costs a fixed cent amount. The system uses models from Anthropic and can manage emails, documents, and team posts largely autonomously.
Zoom simultaneously launched its own platform for AI-powered virtual office spaces. Google released a beta of Gemini Spark for macOS, which automates desktop workflows, and AWS is rolling out specialized work environments for AI agents worldwide.
Measurable Wins in Specialized Fields
Despite the general hiccups, AI is yielding clear results in niche areas. The Fraunhofer IWU developed a forecasting tool for textile manufacturer frottana that explains over 80 percent of sales fluctuations, sharpening demand planning.
In aviation, the DLR Loki project tested an AI assistant for air traffic controllers that boosted capacity by 25 percent in simulations. Practical deployment, however, is still an estimated 10 to 15 years away.
Accounting, by contrast, is already highly automated: receipts are processed in seconds, and booking suggestions are virtually error-free. The role in tax firms is shifting from data entry to strategic consulting.
Consolidation Phase Expected
Gartner predicts that a shakeout is coming. By 2027, roughly 40 percent of current projects in agentic AI may be scrapped as companies confront the botsitting reality.
