AI productivity in the ECB's data: perception versus real impact

According to economists António Dias da Silva, Laura Lebastard and David Sondermann, writing in an analysis published on 26 August 2026 on the European Central Bank's official blog, the share of euro area workers using artificial intelligence has doubled in two years, rising from 26% in 2024 to 52% in 2026. Those who use these tools do so on average about three days a week. The study draws on the ECB's Consumer Expectations Survey, which polls roughly 20,000 people across 11 countries every month, and offers a snapshot of real-world adoption three weeks after the AI Act's transparency rules became applicable (2 August 2026).

The authors report that the median user claims to save three hours a week, around 7.7% of median working hours. The same economists point out, however, that this benefit is distributed very unevenly: most workers see moderate savings, while only a minority experience very large efficiency gains. Moreover, since only 48.8% of respondents say both that they use AI and that they save time, the authors calculate that the impact on the efficiency of the economy as a whole falls to about 3.8% of hours worked. It is worth stressing that these are figures entirely self-reported by workers, not objective measurements of additional output — a methodological limit that tends to capture subjective and potentially optimistic perceptions.

The biggest savings come from specialist tasks: generating or fixing code is worth almost eight hours a week, though it concerns a small share of workers, around 8%. More common activities such as research and writing yield smaller benefits. On adoption, the authors report a rate of 61% among graduates against 37% among those with lower qualifications, while younger workers are around 20 percentage points more likely to use these tools than their older colleagues. Among non-users, 41% say they are simply not interested and 33% consider AI irrelevant to their role, alongside doubts about accuracy and employers failing to provide tools. About half of all workers say they would use AI more with proper training — an investment that around 50% of firms say they plan to make over the next twelve months.

Despite widespread fears, the economists note that no negative effects on employment have emerged so far at company level, recalling an earlier blog post of 4 March 2026 showing that AI-intensive firms had in fact stepped up hiring. Related ECB research estimates an AI-driven productivity gain of around 0.35 percentage points a year for the euro area — an order of magnitude the post cites but does not demonstrate. Like every text on the blog, the analysis carries the standard disclaimer: the views are the authors' own and do not commit the ECB or the Eurosystem. Saving time does not automatically mean producing more, or better; often it just redraws the boundaries of our daily distraction. If about half of all workers say they would use AI more with proper training, perhaps the real bottleneck is not the algorithm but the organisational capacity of those meant to lead its integration. — Olya