aceModerate evidence
Published recordBusinessGood Space
More
AI as Shared Business Infrastructure: Firm Growth and the Distribution of Gains
- Truvace record
- TRV-2026-2482raw data (JSON)
- Version
- v1Published
- First published
- Evidence type
- Observational studyPreprint, not yet peer reviewed
- Sources
- 1 sourcePeer-reviewed
- sha256
- 7a0a62160ca2…check it
Generative AI improves performance on individual tasks, but whether those gains translate into firm growth, and for which firms, remains an open question. We study a leading Chinese e-commerce platform that in 2024 embedded a generative AI system in the operating software firms use to run their businesses, drawing on centrally developed models and platform-wide data to deliver ready-to-use applications to every firm. Complete records of each firm’s AI use, linked to its monthly operations and sales, let us trace…
The gain
In a study of 148,770 small sellers on a Chinese e-commerce platform, firms that adopted the built-in generative AI system saw monthly sales rise 51.6% and orders rise 47.0% compared with matched non-adopters, with the biggest proportional gains for the smallest firms.
What this doesn’t fix
Baseline estimates apply only to continuing physical-goods sellers with annual sales of at most RMB 5 million on one anonymized platform, and firms chose whether to adopt, so the instrumental-variable estimates apply to firms whose adoption responds to the instrument.
Sources
1 on file- NBER Working Paperdoi:10.3386/w36003
Cite this record
v1Truvace Impact Record TRV-2026-2482, v1: “AI as Shared Business Infrastructure: Firm Growth and the Distribution of Gains.” Truvace, 2026-11-12. /record/TRV-2026-2482 (accessed on the day you copy it). sha256 7a0a62160ca2d34d…
Change history
1 versionEvery change made to this record since it was first published, and why. Nothing has changed since it went up.
Certified into the record
Check this record hasn’t been altered
How to check it yourself, without trusting this page
Download any version from /api/record/TRV-2026-2482 and hash its canonical field yourself, for example with shasum -a 256. You should get its content_hash. Each version’s text also ends with prev:and the hash of the version before it (version 1 points to 64 zeros). If anyone had changed a single character of any version after it was saved, those hashes wouldn’t line up.