SlovakiaPublic sector2024 – ongoing

National Bank of Slovakia

GPT-4o field experiment on employee workplace tasks

A National Bank of Slovakia experiment found higher-quality answers and 21% less task time with GPT-4o

WorkingCompany's own claim

Problem

The bank provided a setting to test how generative AI affected the quality and speed of employee work. Researchers examined whether benefits differed across workplace tasks.

Approach

Researchers ran a field experiment with central-bank employees in 2024. Participants were randomly assigned GPT-4o access for a subset of workplace tasks. The study compared response quality and completion time with and without access.

Results

33–44%average improvement in response quality with GPT-4o access, depending on the study specificationCompany's own claim[1]

21%less average task-completion time with generative AI access in the field experimentCompany's own claim[1]

Takeaway

The measured gains came from an employee experiment. They did not establish bank-wide production savings.

Sources

A Task-Based Approach to Generative AI: Evidence from a Field Experiment in Central Banking
National Bank of Slovakia8 Jul 2025Research paper or report

Source summaries

[1]National Bank of Slovakia8 Jul 2025

Aleš Maršál and Patryk Perkowski reported a randomized employee experiment at the bank. GPT-4o access improved average response quality by 33–44% and reduced average completion time by 21%.

Full details are in the original sources.

Added 1 Nov 2026 · Updated 1 Nov 2026