This site is undergoing review. Some annotations were human-generated, some AI-generated — all are being verified.

Algorithmic Management in the Workplace: New Evidence from an OECD Employer Survey

Milanez, Lemmens, Ruggiu

2025OECD Artificial Intelligence Papers23 citations

Adoption / usageInterdisciplinary

AI (General)Algorithmic managementHuman-AI collaborationDecision-making

Abstract

Algorithmic management – the use of software, which may include artificial intelligence (AI), to fully or partially automate tasks traditionally carried out by human managers – has received increased attention in recent years. On the one hand, it has the potential to deliver productivity and efficiency gains as well as greater consistency and objectivity of managerial decisions within firms. On the other hand, there is growing evidence from other studies of its potential detrimental impacts on workers. As policymakers grapple with how to respond to the challenges that algorithmic management presents, additional evidence is needed. Towards this aim, this study draws on a unique survey of over 6 000 firms in six countries: France, Germany, Italy, Japan, Spain and the United States. The survey offers unprecedented insights into the prevalence of algorithmic management, its perceived impacts and firm-level measures to govern its use. The findings show that algorithmic management tools are already commonly used in most countries studied. While managers perceive that algorithmic management often improves the quality of their decisions as well as their own job satisfaction, they also perceive certain trustworthiness concerns with the use of such tools. They cite concerns of unclear accountability, inability to easily follow the tools’ logic, and inadequate protection of workers’ health. It is urgent to examine policy gaps to ensure the trustworthy use of algorithmic management tools.

Summary

Milanez, Lemmens, and Ruggiu conduct a representative survey of 6,047 mid-level managers across six OECD countries (France, Germany, Italy, Japan, Spain, and the United States) to document the prevalence of algorithmic management tools, their perceived impacts on managerial decision-making and job quality, trustworthiness concerns, and firm-level governance measures.

Main Finding

Algorithmic management is widespread in the United States (90% adoption) and European countries surveyed (79% average), but less prevalent in Japan (40%). While 60% of managers report improved decision-making quality, 64% express at least one trustworthiness concern, most commonly unclear accountability (28%), lack of explainability (27%), and inadequate protection of workers' physical and mental health (27%).

Primary Datasets

OECD Employer Survey on Algorithmic Management

Key Methods: Representative employer survey with telephone interviews (CATI) of 6,047 mid-level managers across six OECD countries, with stratified sampling by firm size and sector, weighted to ensure representativeness
Sample Period: 2024
Geographic Coverage: OECD
Sample Size: 6,047 firms with 20 or more employees
Level of Analysis: Firm, Individual
Occupation Classification: None
Industry Classification: NACE Rev. 2

Notes

OECD Publishing [Claude classification]: This OECD working paper provides the first representative estimates of algorithmic management adoption for the United States and Japan. For European countries, it improves upon existing estimates by gathering adoption data at the granular level of specific tool types (instruction, monitoring, evaluation). The study is unique in surveying managers directly about their firsthand perceptions of impacts, trustworthiness concerns, and governance measures. Data collection was conducted by Ipsos NV via telephone interviews (CATI) between June and August 2024. [Claude classification]: This OECD working paper provides the first representative estimates of algorithmic management adoption for the United States and Japan. For European countries, it improves upon existing estimates by gathering adoption data at the granular level of specific tool types (instruction, monitoring, evaluation). The study is unique in surveying managers directly about their firsthand perceptions of impacts, trustworthiness concerns, and governance measures. Data collection was conducted by Ipsos NV via telephone interviews (CATI) between June and August 2024. [Claude classification]: This OECD working paper provides the first representative estimates of algorithmic management adoption for the United States and Japan. For European countries, it improves upon existing estimates by gathering adoption data at the granular level of specific tool types (instruction, monitoring, evaluation). The study is unique in surveying managers directly about their firsthand perceptions of impacts, trustworthiness concerns, and governance measures. Data collection was conducted by Ipsos NV via telephone interviews (CATI) between June and August 2024. [Claude classification]: This OECD working paper provides the first representative estimates of algorithmic management adoption for the United States and Japan. For European countries, it improves upon existing estimates by gathering adoption data at the granular level of specific tool types (instruction, monitoring, evaluation). The study is unique in surveying managers directly about their firsthand perceptions of impacts, trustworthiness concerns, and governance measures. Data collection was conducted by Ipsos NV via telephone interviews (CATI) between June and August 2024. [Claude classification]: This OECD working paper provides the first representative estimates of algorithmic management adoption for the United States and Japan. For European countries, it improves upon existing estimates by gathering adoption data at the granular level of specific tool types (instruction, monitoring, evaluation). The study is unique in surveying managers directly about their firsthand perceptions of impacts, trustworthiness concerns, and governance measures. Data collection was conducted by Ipsos NV via telephone interviews (CATI) between June and August 2024. [Claude classification]: This OECD working paper provides the first representative estimates of algorithmic management adoption for the United States and Japan. For European countries, it improves upon existing estimates by gathering adoption data at the granular level of specific tool types (instruction, monitoring, evaluation). The study is unique in surveying managers directly about their firsthand perceptions of impacts, trustworthiness concerns, and governance measures. Data collection was conducted by Ipsos NV via telephone interviews (CATI) between June and August 2024.