Incentive Structures for AI Technology Adoption: A Cluster-Randomized Trial in Engineering Organizations

How to cite

Berndt, Johannes; Sadun, Raffaella; Shukla, Soumitra; Englmaier, Florian; Hesler, Nikolaus von; Tamayo, Jorge; Kayser, Simon. 2026. "Incentive Structures for AI Technology Adoption: A Cluster-Randomized Trial in Engineering Organizations." AEA RCT Registry. AEARCTR-0017768. https://doi.org/10.1257/rct.17768-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Johannes Berndt* , Raffaella Sadun , Soumitra Shukla , Florian Englmaier , Nikolaus von Hesler , Jorge Tamayo , Simon Kayser

* Primary investigator

Registry ID
AEARCTR-0017768
DOI
https://doi.org/10.1257/rct.17768-1.0
First registered
Last updated
Status
In development
Trial dates
2026-01-12 – 2027-12-31
Countries
Private
JEL codes
M54, O33, J24, D23
Registry record
http://www.socialscienceregistry.org/trials/17768

Abstract

We study how organizational incentive structures affect the adoption of AI-powered productivity tools among engineers. Despite growing evidence that AI assistants can enhance individual productivity, firms report substantial variation in adoption rates during technology rollouts. This gap between potential and realized adoption may stem from misaligned incentives: employees bear short-term learning costs while long-term benefits accrue to the organization.

We conduct a cluster-randomized controlled trial with engineering teams at a large multinational manufacturing firm during the rollout of an AI-powered engineering assistant. The tool assists engineers with code generation, interface design, and hardware configuration tasks. During an internal rollout of a AI-powered productivity tools, we regularly offer incentives to encourage experimentation with the new tools and encourage knowledge diffusion within the team. The managers of these teams will be encouraged to support the experimentation and knowledge diffusion.

We measure adoption through platform telemetry (usage frequency, feature utilization), productivity metrics (task completion rates, efficiency), and survey measures (perceived usefulness, intention to continue use). The trial runs for three months with data collection at baseline and monthly intervals. We will also continue data collection 3 months after the end of the treatment period.

This study contributes to our understanding of how firms can accelerate technology adoption by addressing incentive misalignments at both individual and team levels. Results will inform organizational strategies for deploying AI tools at scale.

Intervention

During an internal rollout of a AI-powered productivity tools, we regularly offer incentives to encourage experimentation with the new tools and encourage knowledge diffusion within the team. The managers of these teams will be encouraged to support the experimentation and knowledge diffusion.

Experimental design

This is a three-arm cluster-randomized controlled trial comparing incentive structures for AI technology adoption. Engineering teams are the unit of randomization, with all members of a team assigned to the same treatment arm.

Design details

Randomization method
Computer-generated random assignment using pre-specified random seeds for reproducibility. Randomization is conducted by the research team using a randomization tool that assigns teams to arms based on random number generation.
Randomization unit
Team (cluster): All members of a team are assigned to the same treatment condition.
Clusters
Anticipated total: 1500 teams (approx.: 2000-4500 engineers) Approached total: 4700 teams (represents all teams who were informed about release of new AI-powered productivity tool but likely majority not affected) Average cluster size: 10 Range of cluster: 2-121
Observations
Anticipated total: 2000-4500 engineers Approached total: 38000 engineers (represents all individuals who were informed about release of new AI-powered productivity tool but likely majority not affected)
Arms
Arm 1 (Control): ~500 teams (1000-1500 individuals) Arm 2 (Individual Incentives): ~500 teams (1000-1500 individuals) Arm 3 (Manager + Team Incentives): ~500 teams (1000-1500 individuals)
Primary outcomes
Primary Outcome 1: AI Tool Adoption Primary Outcome 2: Feature Breadth Primary Outcome 3: Sustained Adoption
Secondary outcomes
Secondary Outcome 1: Self-Reported Productivity Impact Secondary Outcome 2: Telemetry-tracked Productivity Secondary Outcome 3: Knowledge Sharing Secondary Outcome 4: Intention to Continue Use

Cited by

No citation data available for this registration yet.

Keywords