Generative AI and Technology Adoption in Organizations
How to cite
Friess, Svenja; Castro, Silvia; Mickeler, Maren Alana; Ho, Hoa; Grinuite, Evelina. 2026. "Generative AI and Technology Adoption in Organizations." AEA RCT Registry. AEARCTR-0018777. https://doi.org/10.1257/rct.18777-1.0
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
Firms are increasingly deploying GenAI tools, however, we lack evidence on effective implementation strategies. We study whether designating a peer facilitator improves the frequency and the quality of employee engagement with a new GenAI tool. Partnering with a commercial bank that introduced a customized GenAI credit-assessment tool, we randomly assign in half of the branches a designated employee who supports colleagues' use of the tool. We measure effects on (i) tool usage patterns and (ii) the quality of loan officers' credit-assessment decisions, and (iii) officer-level heterogeneity in adoption and usage. The design allows us to isolate the causal effect of peer facilitation on GenAI tech adoption, informing organizational strategies for AI implementation beyond simple tool provision.
Intervention
All branches in the study receive a generative AI credit-assessment tool and standard training. The tool is customized and trained on the firm’s internal information base (policies, product databases, procedural manuals, past product to client matches) and employees’ knowledge. The main objective of the tool is to assist employees on the credit analysis of loan applications. In each treatment branch, one or more employees are additionally designated as GenAI facilitator: they take on a support role — encouraging colleagues use the tool, troubleshooting problems, and acting as a knowledge broker (within and across organizational units). Facilitators are identified before deployment targeting employees who are both helpful and technology-affine. Control branches receive the tool and training without a designated facilitator.
Experimental design
The partner bank has 80 branches, with the tool rolled out in two phases. In the first phase, 40 branches receive the tool and enter this study: they are formed into 20 matched pairs, and within each pair one branch is randomly assigned a facilitator and one is not, yielding 20 treatment and 20 control branches. This first-phase sample is confirmed and constitutes our core sample.
Contingent on the partner firm's continued participation, we plan to extend the study to the remaining 40 branches when the tool is rolled out to them in a second phase (i.e., after the intervention concluded) using the same matched-pair randomization to add 20 further treatment and 20 further control branches — for a total of 40 treatment and 40 control (80 branches) if the second phase proceeds.
Design details
Cited by
No citation data available for this registration yet.