MercadoLibre 第三季度 AI 投入环比增加 3000 万美元

By PYMNTS  |  November 4, 2026

 | 

E-commerce and FinTech company MercadoLibre increased its investment in artificial intelligence by $30 million quarter over quarter as it builds its work around AI, the company said in a Wednesday (Nov. 4) letter to shareholders about its third quarter results.

“We increased the investment quarter on quarter by $30 million, and we continue to see very, very good results, both consumer-facing as well as internal efficiencies,” MercadoLibre Chief Financial Officer Martín de los Santos said during a Wednesday earnings call.

MercadoLibre’s usage of AI grew “much faster” than its spend, the company said in the letter. The firm has become more efficient in its use of the technology by getting better at choosing the right model for each task and by expanding its use of open-source models, it said.

MercadoLibre has seen returns from its investment in AI in the form of revenue uplift, better user experience, cost avoidance and productivity, according to the letter.

The company provided several examples in the letter. The revenue uplift from deploying AI in search more than offset the cost of the models, the use of AI in customer service has reduced the need for representatives by “a multiple,” the company’s FinTech AI assistant Mago is now holding 19 million conversations per month and improving the user experience, AI has helped double the output of the technology teams over the past year, and the technology has freed up almost 20% of the hours worked by the finance teams.

During Wednesday’s call, de los Santos highlighted Mago, the AI assistant on the company’s FinTech platform.

“That continues to increase engagement with our users,” de los Santos said. “We see [Net Promoter Score] improving because of Mago. We see better engagement with our platform, more of their funds coming to Mercado Pago. So, all positive signs.”

De los Santos also pointed to the use of AI in search as an application that provides a measurable return on investment.

“We included an LLM on the search algorithm, and we can see through the incremental conversion and the click-through rate that we’re seeing in those search results, we can see that the cost of the models are more than compensated by the incremental revenues that we’re generating,” de los Santos said.

“So, all the metrics that we measure are very effective in terms of productivity,” de los Santos said.

For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.