Science1 publisher2 min readPublished
Employers introduced only 55 percent of the AI tools their staff use most often
The third wave of the Konstanz AI study surveyed 1,105 employees in May 2026 and put workplace AI use at 38 percent, a three-point rise that sits inside the sample's own error bar. Just over half of the tools those users reach for came in through the employer.
The Scientist · Science desk
What happened
- The third wave of the Konstanz AI study surveyed 1,105 employees in May 2026 and found the share using AI in their work at 38 percent, up from 35 percent a year earlier.
- Just 55 percent of those AI users said the tool they use most frequently had been officially introduced by their employer.
- Use splits by occupation: 49 percent in office and knowledge-based work against 25 percent in production-related and manual jobs, a divide the authors describe as largely unchanged.
- In small organizations, 11 percent of employees reported having received AI training and 10 percent reported binding rules on how AI may be used.
- Forty percent of employees expect AI and automation to hurt the labor market over the next ten years, while 17 percent worry about losing their own job.
Compiled by The ScientistSomething wrong?How this is made
Why it matters
- constraint A policy written against the tools an employer procured governs only a bit more than half of what people actually use most; the remainder sits outside whatever inventory exists. That limits what an audit of sanctioned software can tell anyone.
- decision Anyone budgeting on the premise that workplace adoption is accelerating needs an effect this wave leaves open. Until a reading separates from the error bar, flat is the defensible assumption.
- cost When the employer skips training, the learning still happens and employees pay for it in their own time and with their own choice of tool. Firms usually absorb a new technology the other way round.
- exposure The employees with the least training and the fewest rules are also the ones Gerdiken says risk being left further behind, so the governance gap and the inequality gap fall on the same people.
Take the 55 percent at face value and the other 45 percent of AI users are working with a tool that arrived without any formal introduction from their employer [3][1]. Scaled against the 38 percent who use AI at all, that is about 17 of every 100 employees in the sample [2]. The question asked about official introduction and left open who did the choosing, so that 45 percent mixes tools a team adopted, tools a colleague passed along and tools one person signed up for alone [3].
The year-on-year move deserves less weight than it will get. Three points, from 35 to 38 [2][3]. For 1,105 respondents, the 95 percent interval around a 38 percent proportion is roughly plus or minus 2.9 points under simple random sampling [1][4]. The authors did not publish the earlier wave's sample size, so the difference between waves cannot be tested from what is published. Both a flat line and a slow climb fit.
"In many organizations, the major AI revolution has not unfolded as a planned transformation process so far," said Florian Kunze, head of the study and professor of organizational behavior at the University of Konstanz [11]. The study describes adoption as frequently driven by employees instead of being strategically managed [15]. The practice, in other words, is arriving before the policy, the training and the secure infrastructure [15].
Use by education runs 56 percent among employees with high educational attainment and 21 percent among those with lower attainment, a spread of 35 points [6][6]. "Without targeted support measures, there is a risk that existing inequalities will become further entrenched," said Elena Gerdiken, a doctoral researcher at the Future of Work Lab in Konstanz [12]. Small organizations are also less likely to hand out company-owned AI tools or set formal guidelines [8].
What a survey of employees measures is what employees report and recall. The instrument is self-report, not a tool inventory or a set of network logs. So the unsanctioned share here is the share people described, and the researchers' own remedy list runs to clear rules, job-related training and targeted support for small and medium-sized enterprises [14]. On value, the same respondents are unimpressed so far: 23 percent see AI as already making a significant contribution to their organization's value creation, and 12 percent think it has reduced the number of entry-level positions [10].
What to watch
- A fourth wave using the same instrument, with both wave sizes published, would show whether 38 percent is a plateau or a slow climb.
- Whether the 10 percent of small-organization employees reporting binding rules moves, since the recommended support measures would land there first.
- Any measurement independent of self-report: a tool inventory or traffic logs set against what employees say they use.