The transition from assistive AI copilots to self-directed software agents is widely considered the next major paradigm shift in enterprise IT. However, an empirical study released by market research firm Lünendonk and Hossenfelder demonstrates a stark discrepancy between executive ambitions and operational reality across Germany, Austria, and Switzerland. While corporate leaders overwhelmingly anticipate significant efficiency and cost benefits, almost no organization has managed to transition autonomous agents into comprehensive production environments.
Published on September 17, 2026, the study titled "Agentic AI: vom Copiloten zum Autopiloten" was conducted in cooperation with Cosmo Consult, Materna, HyPlus, Reply, Sopra Steria, and Woodmark. The authors surveyed 180 chief information officers, IT heads, and business department leaders across the DACH region. The strategic appetite for automation is virtually unanimous: 96 percent of respondents confirmed that they expect tangible cost reductions and process acceleration from deploying agentic artificial intelligence.
Despite this high enthusiasm, the research uncovers a pronounced deployment bottleneck. A total of 80 percent of surveyed enterprises remain stranded in exploratory or testing phases. Specifically, 58 percent are running limited pilot projects, while another 18 percent are conducting isolated test runs without direct operational ties. Only 19 percent of companies currently utilize autonomous agents in productive environments, and even those deployments remain confined to narrowly defined operational niches.
The shortfall is particularly acute when examining holistic process execution. Just one percent of businesses across the DACH region have achieved full end-to-end integration of AI agents across their core enterprise workflows. The overwhelming majority of proof-of-concept projects collapse before reaching general availability. The primary cause of this attrition is not the AI model capability itself, but the fragmented corporate systems and legacy software stacks that prevent reliable system-level execution.
The single greatest roadblock identified in the report is inadequate data infrastructure. IT organizations report that between 60 and 80 percent of their total project expenditure and workforce hours must be spent cleaning up, structuring, and interconnecting enterprise data silos. Autonomous agents require dependable, low-latency access to accurate corporate knowledge to execute multi-step workflows. Without this foundation, agentic models produce errors or stall, prompting IT leaders to halt live rollout plans.
These findings indicate that the road to autonomous enterprise software demands far more groundwork than initially projected. Organizations looking to capitalize on agentic automation must treat data engineering and governance as mandatory prerequisites rather than secondary considerations. Until internal data architectures are modernized and connected, the vision of an autonomous, autopilot-driven enterprise will remain confined to laboratory experiments for 99 percent of regional businesses.