The exact text we saved

This is the version exactly as it was saved, character for character. The SHA-256 hash below is a 64-character code worked out from that text. Run the same text through SHA-256 and you get the same code; change a single letter and the code comes out completely different. Each version also stores the previous version’s hash, which chains them together so no version can be swapped out quietly.

TRUVACE RECORD VERSION
record: TRV-2026-2482
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-11-12T07:40:33.706202Z
status: published
lens: g_space
sector: business
headline: AI as Shared Business Infrastructure: Firm Growth and the Distribution of Gains
dek: 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…
gain_title: 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.
problem_title: (none)
trace_subject: (none)
gain_reading: 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.
gain_evidence: adoption raises monthly sales by 51.6%, primarily through a 47.0% increase in the number of orders | monthly sales rise by 61.5% among firms in the bottom quartile of pre-launch sales, compared with 10.3% in the top quartile
problem_reading: (none)
problem_evidence: (none)
quick_read: A working paper from economists at the University of Wisconsin-Madison and Zhejiang universities studied a leading Chinese e-commerce platform that built a generative AI system into its seller software in 2024. Comparing 148,770 adopting firms with 77,347 matched never-adopters, the authors found that adoption raised monthly sales by 51.6% and orders by 47.0%, with little change in average order price.

The gains were largest for the smallest sellers. Monthly sales rose 61.5% among firms in the bottom quartile of pre-launch sales, against 10.3% in the top quartile. The authors say this suggests that when a platform supplies AI as shared infrastructure, the benefits need not go mainly to large firms.

The authors also looked at whether adopters simply took business from competitors. Product categories with more pre-existing information problems saw more total sales, more new firms and more new products, and tests on firm exit and competitive exposure found little evidence of displacement. The study covers firms with annual sales below RMB 5 million on a single anonymized platform, so the results may not carry over to larger firms or other markets.
limitation: 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.
tag: Evidence-backed gain
key_points: Sellers paid RMB 99 (about $14) a month after one free month; the AI covered content generation (images, videos, titles) and business operation (rules guidance, performance reports, competitor benchmarks, customer service). | Adopters reached more customers: page views and new visitors rose roughly 45%, and sales to new customers rose 75.6%. | Adopters introduced 66.2% more new products and earned 50.5% more sales from newly introduced products. | Quality measures improved: negative review rates fell 21.9%, firm-liability disputes fell 20.6% and delivery time fell 1.5%, though the overall refund rate rose 4.2%, driven by consumer-attributed refunds such as change of mind. | The AI made routine work cheaper: product images rose 62.6% and listing updates per product rose 124.8%, while staff time per order fell 5.6%. Producing equivalent content at market prices would cost the median bottom-quartile firm 22.7% of annual sales, against 0.4% in the top quartile. | Using other estimators gave different sizes: an imputation estimator implied a 40.1% sales rise, firms adopting only content generation gained 25.2% and only business operation 22.8%. | Instrumental-variable estimates, built from pre-launch reliance on AI-relevant tasks, provincial digital readiness and AI spread in related categories, gave the same conclusions with larger point estimates.
rundown: The platform, anonymized under a data-use agreement, began piloting the system in March 2024, launched content generation in April and business operation in July. By July 2025 more than half of the platform firms observed had used at least one functionality. The authors link usage logs to monthly sales, traffic, listings, quality and staffing records from October 2023 to July 2025, which lets them measure adoption directly rather than from surveys or job postings.

The baseline sample is continuing sellers of physical goods with pre-launch annual sales of at most RMB 5 million, matched on tenure, product category, store type, sales, orders and active products. Treatment required adopting both functionalities within a month of each other. The design is a matched difference-in-differences with an event study that showed no differential pre-trends, plus an instrumental-variable check with a first-stage F-statistic of 207.80.

Two mechanisms are proposed. AI lowers the cost of routine work, and gains were largest for firms whose operating costs were highest relative to sales before adoption. AI also reduces information frictions: adopters' titles lengthened 20.2% and the share of product attributes reported rose about 30%, with the largest gains where refund rates or sales volatility were highest.

On market effects, the AI was available to every firm, so no category went untreated. The authors compared categories by 2023 refund rates, and high-friction categories saw larger increases in total sales, incumbent sales, firm entry, entrant sales and new-product introductions after launch, with parallel pre-trends. Category-level AI adoption was not significantly linked to firm exit, and the gap between adopters and controls did not widen with competitive exposure.
sources:
- peer_reviewed | NBER Working Paper | https://doi.org/10.3386/w36003 | 2026-11-01
prev: 0000000000000000000000000000000000000000000000000000000000000000

SHA-256 hash

7a0a62160ca2d34d9ecbf65611f02f62b63e00922091961820888bc4227ae825

Previous version’s hash

0000000000000000000000000000000000000000000000000000000000000000