Most German firms pilot AI, but few use it in core

Zoi has published a study showing that 76% of large German companies are piloting AI agents, while 19% have integrated them into core business processes. The findings point to a wide gap between experimentation and operational use.

The survey covered 500 IT decision-makers at German companies with more than 2,000 employees and examined how far businesses have progressed in adopting generative AI. It found that the main obstacles to wider use are operational rather than financial. IT infrastructure complexity, legacy-system integration and skills shortages were cited more often than budget concerns or uncertainty over returns.

The results reflect a broader pattern across large organisations: companies have moved beyond initial trials but are struggling to embed AI in day-to-day operations. The study argues that the issue is not whether businesses are willing to test AI tools, but whether they can connect them to existing platforms, processes and management structures in ways that deliver measurable results.

The consultancy grouped respondents into four categories: AI Champions, AI Mainstream, Structural Pioneers and Use Case Drivers. Most companies, 62%, fell into the AI Mainstream category, meaning they are still in the process of systematic development. AI Champions accounted for 21% and were identified as performing strongly in both organisational readiness and measurable business impact.

Use Case Drivers made up 12% of the sample and were described as seeing early quantifiable effects without yet having a strong organisational base in place. Structural Pioneers represented 5% and were described as having laid the foundations for AI deployment without yet translating that into significant business impact.

Strategy gap

One of the clearest divides in the findings was between companies with formal AI plans and those that have linked those plans to performance measures. While 74% of respondents said they had a documented AI strategy, only 34% said it was tied to key performance indicators.

A similar pattern appeared in governance. Rules, compliance frameworks and general oversight are now common in many large companies, but more formal structures such as an AI board or a central leadership role remain far less widespread.

Successful implementation depends on several factors working together rather than on a single decisive change. The study assessed eight dimensions, including AI strategy, data governance, cloud infrastructure, standardised processes, change management, budget stability and data-driven decision-making, and concluded that companies make progress when these elements advance together.

The research also found that companies already measuring the return on investment from generative AI generally reported positive outcomes. At the same time, it noted that failed projects are often excluded from ROI calculations, which may make the overall picture appear stronger than it is.

That matters because the survey suggests many businesses remain active in AI without yet seeing broad operational gains. The gap between technical deployment and business use remains one of the central issues in large-company AI adoption, especially as businesses try to connect new systems with older technology estates.

Danilo Kirschner, Managing Director of Zoi North America, said legacy technology is proving a serious constraint on progress. "If these organizations are hitting a wall transitioning AI from pilot projects to core operations due to legacy infrastructure, it serves as a major bellwether for Fortune 500 companies everywhere," Kirschner said.

Benjamin Hermann, Chief Executive Officer of Zoi, drew a distinction between launching AI projects and turning them into productive business tools. "The study results demonstrate that there is a world of difference between technological and operational excellence, particularly in AI transformation. Starting with AI is easier than being productive with AI-and this is ultimately reflected in the number of automated processes, not the number of API calls. We see that many companies underestimate this discrepancy, causing numerous AI projects to stall at the stage of technological adoption. A successful AI transformation requires a cultural shift driven by internal AI champions. Identifying and empowering them are the most critical steps toward an AI revolution," Hermann said.

The survey analysis was led by Prof. Dr. Jan Kirenz and his team at Stuttgart Media University. He said the figures showed broad interest in AI, but also a failure so far to turn that interest into widespread value creation in core operations.

"The data paints a clear picture: 76 percent of large German enterprises are already actively piloting AI agents, proving a widespread interest in innovation. However, since only 19 percent deploy these technologies within their core processes, broad value creation has yet to materialize. The barriers to scaling are rarely financial. Therefore, the decisive lever lies in professionalizing strategy, governance, and operations in lockstep, turning AI from a pilot project into an integral part of the core business," Kirenz said.

The study adds another data point to the debate over whether large companies can move generative AI beyond isolated tests and into routine use. For many of the businesses surveyed, the next stage will depend less on enthusiasm for the technology than on their ability to reshape systems, management structures and internal expertise around it.