The artificial intelligence-productivity relationship: a systematic literature review and research agenda

Andreas Cebulla, Ehsan Abedin, Janice Jones · Journal of Productivity Analysis · 9 Jul 2026

Where this sits in the debate

Will AI actually boost productivity and growth? · its finding: it depends (why studies differ)

Direction of the finding as read from the abstract by an AI model (AI-written).

Key takeaway AI-written

A systematic review synthesizing 76 peer-reviewed studies (2015-2025) on AI and productivity finds a broadly positive but highly inconsistent relationship, with effect size and direction depending heavily on how AI and productivity are measured and on organizational, sectoral, and institutional context. It maps a fragmented empirical landscape rather than producing a new estimate, identifying workforce skills, complementary innovation, and digital infrastructure as recurring enablers and data limitations, cost, and regulation as recurring barriers.

The paper in three steps AI-written

Written by an AI agent from the paper's abstract, which is the primary source and is quoted below.

The question

What does the empirical literature say about the relationship between AI and productivity in firms, industries and economies?

The approach

A systematic review of 76 peer-reviewed studies published from 2015 to 2025, synthesised with thematic inductive analysis.

The findings

Most studies point in a positive direction, but the size and consistency of effects vary with how AI and productivity are measured and with context. Enablers include organisational capabilities, skills, complementary innovation and digital infrastructure; barriers include data limits, cultural resistance, regulation and adoption costs.

How to read it: review. A synthesis of existing studies — breadth over novelty; quality depends on what it includes and excludes.

Limitations and counterpoints AI-written

Many reviewed studies bundle AI with robotics, broadband and wider digital transformation, so the review cannot isolate AI's own effect.

Abstract

Abstract Artificial Intelligence (AI) is widely recognised as a transformative general-purpose technology with the potential to reshape productivity dynamics across firms, industries, and economies. Despite growing empirical research, the evidence remains fragmented due to variation in definitions, measurement approaches, and contexts, making it difficult to draw consistent conclusions about productivity outcomes. This study addresses this gap by systematically synthesising 76 peer-reviewed studies published between 2015 and 2025. The review suggests a positive directional pattern between AI-related technologies and productivity outcomes across much of the literature, although the magnitude and consistency of effects vary considerably depending on how AI and productivity are operationalised, as well as across organisational, sectoral, and institutional contexts. Many reviewed studies examine AI as part of broader digital transformation initiatives, including robotics, automation, digital infrastructure, broadband expansion, and related technological systems. Through thematic inductive analysis, the study identifies key enablers of productivity, including organisational capabilities, workforce skills, complementary innovation pathways, and digital infrastructure. It also highlights recurring barriers, such as data limitations, cultural resistance, regulatory constraints, and high adoption costs, demonstrating that gains are neither uniform nor automatic. By synthesising heterogeneous empirical evidence and linking mechanisms to outcomes, the review provides a nuanced understanding of AI’s role in shaping productivity and offers a foundation for future research and policy aimed at maximising AI’s productivity potential.

About this publication

Academic journal. Published in an academic journal — typically peer-reviewed before publication, though indexing alone doesn't guarantee review depth.

Published: 2026 (3 months ago)
Citations: 0 (OpenAlex)