From the filing
PDF page 227
No printed folio visible · Image transcription
Read the page text
[Embedded graphic 1 — OCR text] mercado libre Customer since 2023 Transforming e-commerce in Latin America Results >100% Merged code up year over year Code submissions reviewed autonomously over the past year 500,000+ Non-tech teams active in Cowork weekly 78% Background Mercado Libre is the leading e-commerce and financial technology company in Latin America, with operations in 18 countries. It made its developers more productive by rolling out Claude Code across its engineering organization. The goal was to help the team ship faster by shifting the upkeep of its infrastructure—version migrations, dependency patches, and test coverage across 40,000 microservices— onto Claude. Impact Mercado Libre evaluated multiple AI providers to find one that could autonomously handle more of its operational burden and make its engineering team more effective. It uses Claude Code for its most demanding workloads—codebase migrations, automated code review, and test generation—and for individual developer productivity. Mercado Libre became one of the earliest adopters in January 2025; today 23,000+ developers across the organization use it. Claude also runs the company’s wide-scale autonomous changes: over the past year, agents with human oversight have reviewed more than halfa million code submissions and migrated roughly 9,000 services to newer platforms. The productivity gains are already significant: in Q2 2026, code submissions rose 110% year over year, merged code more than 100%, and deployments nearly 75%. Quality improved at the same time, with rollbacks down year over year. In 2025, about 40% of the code Mercado Libre shipped to production was written by Claude and reviewed by a human engineer before release. By 2026, human-written code has become the exception. The change reshaped the company’s development team, with 30% of managerial roles “Productivity gains with AI have put our most senior engineers back to building code, not just reviewing it. Claude Cowork now reaches 31,000 employees—one of the largest enterprise deployments globally.” Mercado Libre now contributing code again. Mercado Libre’s near-term goal is for 90% of its coding to run autonomously, including the full pull-request review process. It also plans to fold its separate backend, frontend, and database roles into a single builder role, with agents filling the gaps. Mercado Libre then extended Claude to non-engineering teams. Finance went first: within a month, most of the team had adopted Claude Cowork, using shared skills to generate branded presentations in minutes—work that used to take several people hours a week. Cowork also shortened the monthly financial close by a third and improved internal reporting. Finance can now preview the close on demand instead of waiting for month end, so the team spots shifts like rising costs sooner and reacts faster. Because Claude is grounded in Mercado Libre’s business context, it also drafts the first explanations for why results moved, which the business and finance teams then validate. From there it spread across marketing, HR, legal, compliance, and the business and commercial teams, reaching more than 26,500 people across these teams—78% of them now active every week. Some procurement teams even built their own Claude-powered agents to field supplier questions, without involving IT. “I'd almost say we no longer use spreadsheets —everything’s automated, and we have dashboards built on Claude all over the place. Now we're redesigning core processes from scratch, AI-first—things people always dreamed of but thought were impossible,” said Maria Pia Garcia, Strategic Transformation Director at Mercado Libre.
This page includes text transcribed from an image. Check the original for tables, images, and exact formatting.