What the Konstanz AI Study Reveals About Work
2 Oct 2026
Almost half of people using AI at work use a tool their employer never introduced.
Almost half of people using AI at work use a tool their employer never introduced.
They found it themselves.
That is one finding of the new Konstanz AI Study from the University of Konstanz, just across the Swiss border. Professor Florian Kunze and his team surveyed 1,105 employees in May.
A few more numbers:
→ 38% use AI at work, up from 35% a year ago.
→ In small organisations, only 11% have had any AI training.
→ Only 10% of them work under binding rules for it.
Kunze's reading: people started experimenting on their own, and employers are still catching up on rules, training and secure tools.
My view.
Teaching yourself AI is a good thing. That curiosity is what I hope to see in every team.
The harder part is the one nobody teaches. Which work you hand to AI, and which decisions stay with you.
With a tool you picked yourself, you answer that alone. Often without noticing.
Give away the work. Never the judgment.
If you lead a team, ask which AI tools people already use. Then talk about what they should never hand over.
Which tool did you find?
PS. Here you can find the study → https://www.eurekalert.org/news-releases/1144563
PS 2. Every two weeks I write about AI and leadership in my newsletter: what I test, what works, and what stays human. Sign up here → https://go.thechangerepublic.info/subscribe
What I ask leadership teams after a study like this
Kunze's team surveyed people already at work, not students or early adopters. That is what makes the 38% number worth sitting with. More than a third of employees now use AI at work, up from 35% a year ago, and in small organisations only 11% have had any training for it.
I see the same picture inside the companies I coach. A finance lead found a tool for building forecasts on their own last year. Nobody in IT knew about it until a security review flagged it months later. I wrote more about this exact pattern in the AI finance team that built its own forecast. The person using the tool solved a real problem on their own initiative, with whatever they could find.
This is the part the study cannot measure: once someone picks their own tool, they also make every judgment call about it alone. Which numbers to trust, which draft to send, which client question needs a human answer instead of a generated one. Nobody taught them where that line sits, because nobody knew the tool was in use.
Give away the work. Never the judgment. That line holds whether the tool arrived through a training rollout or through someone's own Saturday afternoon search.
If you lead a team, the practical move is small: ask which AI tools people already use, by name, in the next team meeting. Then agree together which decisions never get handed over.
I go deeper on what changes once someone starts building their own AI judgment in what changes when you build your own AI perspective.
If this resonates and you want a practical guide, I have a free AI Tools and Tips Guide with the tools I use as a coach and leader, and how I decide what to hand to AI and what to keep. Leave your email here and I will send it over: https://www.thechangerepublic.com/free-resources