Ten minutes. That's the time it takes for a company to properly document a malfunction, write a customer email in an understandable way, or create a shift log that doesn't turn into a novel. It's also the time when AI is often truly useful: briefly and directly. This reality can now be quantified for the East Westphalia-Lippe region. A company survey (2021 vs. 2025) and a resulting working paper from the it's OWL project Arbeitswelt.Plus show that the median daily AI usage time has increased from zero minutes to ten minutes. At the same time, many companies are spending more time on the introduction and piloting of AI.

The ten minutes figure is not an isolated finding. It is supported by a second metric: the percentage of respondents who "never" use AI drops significantly from 29 percent (2021) to 5 percent (2025). At the same time, the working paper interprets this development as an indication of greater integration of AI tools into everyday work processes and speaks of the "increasing operational relevance" of AI in companies.

This is a sobering but important insight. When people talk about AI, they quickly resort to big words: disruption, productivity leaps, new business models. The survey from OWL, however, shows that AI is being used more frequently, but in small doses. This fits the industrial work environment, where effects often accumulate over many small steps in processes.

Key findings from OWL at a glance

• Penetration: The proportion of "never used AI" decreases from 29% (2021) to 5% (2025) (significant).

• Daily routine: Median daily AI usage time increases from 0 to 10 minutes.

• Less debate: Planning/discussion drops from 40% to 23% (significant).

• More pilot operations: Introduction/piloting increases from 26% to 37% (significant).

• Usage is growing slowly: Application/use increases from 21% to 25%.

• Laggards remain stable: Share without AI activities remains at 15%.

Source: Working Paper (Arbeitswelt.Plus / it's OWL), panel surveys 2021 and 2025.

From discussion to testing: Pilot projects are growing, but deployment is slow.

When asked at which stage companies use AI, the picture of company responses shifts significantly between 2021 and 2025: Planning and discussion decrease from 40 to 23 percent, introduction and piloting increase significantly from 26 to 37 percent, and application and use grow from 21 to 25 percent.

The results suggest that while companies make progress after discussion and planning, they remain in the pilot phase for an extended period. Furthermore, the proportion without AI activities remains constant at 15 percent.

In other words, getting started is easier more often. Scaling remains the bottleneck. That's precisely what makes these ten minutes so interesting: they show that usage is taking hold in everyday life. But they don't guarantee a rollout, standards, or reliability.

Text instead of automation: Why generative AI is delaying its entry into the market.

The working paper also offers a clue as to why AI is becoming so visibly integrated into everyday life right now: Since the emergence of generative AI applications like ChatGPT, automatic text generation, in particular, has gained in importance. ChatGPT is described as a watershed moment, with automatic text generation and natural language processing gaining in significance, while several features associated with traditional automation have declined statistically significantly since 2021.

Increased usage, increased friction: AI implementation remains complex

Anyone expecting that increased use will automatically lead to fewer problems will be tempered by this paper. Feedback on challenges shows that dealing with AI remains complex. Compared to 2021, a greater number of factors are perceived as problematic. These problems encompass not only technical and infrastructural issues, but also legal, ethical, and organizational aspects. The paper describes a persistent ambivalence: more application, but demanding requirements.

This also fits the piloting logic: those who test AI quickly encounter questions that cannot be solved with a tool update. Governance, responsibilities, data security, training, and team acceptance are all key issues. The working paper explicitly cites long implementation phases and complexity as key factors and links this to the particular complexity of AI applications.

What decision-makers can glean from this: progress yes, scaling remains a challenge.

• Getting started is becoming easier because AI is becoming part of everyday life. There's less "never," more regular use, and the data points to integration into work processes.

• The transition from testing to widespread use is the real challenge. Pilot projects are growing rapidly, deployment only moderately, and the paper explicitly mentions a longer stay in the pilot phase.

• The new wave often begins with language, not automation. Text generation and language processing are gaining in importance, while classic automation features are declining.

• Increased usage does not necessarily mean less friction. Challenges are viewed more broadly, including legal, ethical, and organizational issues.

Note on the survey

The sample sizes differ: in 2021, 318 people from 89 companies participated, while in 2025, 240 people from 50 companies took part. The paper explains the decline primarily by lower participation from top management, while the proportion of employees increases.

The article  "Ten minutes of AI a day: Why OWL companies use artificial intelligence more often, but rarely longer than necessary"  first appeared on  it's OWL  .