Peter Smith
● Rohan Paul recently highlighted a groundbreaking paper called "Generative AI and Firm Productivity: Field Experiments in Online Retail" by Lu Fang, Zhe Yuan, Kaifu Zhang, Dante Donati, and Miklos Sarvary. This research offers some of the first concrete evidence that generative AI actually boosts business productivity, delivering real economic benefits beyond just cutting costs or improving efficiency.
● The team ran large-scale experiments on a major cross-border retail platform throughout 2023–2024, testing AI tools across seven customer-facing areas. These included multilingual chatbots, search optimization, chargeback handling, and better product descriptions — all evaluated across millions of users. The results? AI adoption drove sales increases ranging from 0% to 16.3%, depending on the specific tool. Crucially, since labor, capital, and pricing stayed constant, these gains represent genuine productivity growth, not inflated numbers.
● The most impressive results came from a pre-sale chatbot that handled customer questions in multiple languages, smoothed out product searches, and improved conversion rates without changing cart sizes. Other features — like real-time translation for support and enhanced product data — also delivered measurable results. On average, the added value came to about $5 per customer per year, which is significant at scale. Interestingly, smaller or newer sellers gained the most, as AI helped make up for their limited marketing and service capabilities.
● The researchers do note that smaller companies might struggle to implement these systems effectively, which could widen the AI advantage gap in e-commerce.
Peter Smith
