Mercado Libre is Latin America’s leading ecommerce and fintech platform, delivering more than two billion items to 120 million buyers across 18 countries last year alone. But in LATAM, it’s more than a marketplace - it’s a platform where people shop, sell, get paid, and increasingly, manage their financial lives.

Guided by a mission to democratize commerce through technology, Mercado Libre continues to innovate across its ecosystem. So when it came to reimagining customer experience in the agentic commerce era, the company needed a partner that matched that same pace and ambition.

The Challenge: scaling modern customer experiences on outdated infrastructure

Mercado libre's existing voice customer experience infrastructure was built for a different era. The IVR system forced customers through rigid "press 1 for A, press 2 for B" menus, offering no personalized intelligence and no real-world flexibility. For the team, it operated as a black box: no visibility into why customers were calling, no ability to identify patterns, and no path to continuous improvement.

The result was a channel that was both hard to scale and hard to improve. Customers with very different needs — some urgent, some not — were funneled through the same rigid experience, with no way to distinguish intent or tailor the response accordingly.

The time had come to move to a more trusted, agile, and people-centric ecosystem for delivering better customer experiences.

The Solution: deploying cutting-edge AI agents built for the region

When evaluating solutions, Mercado Libre’s team had a specific set of requirements. The platform needed to handle the complexity of the environment, meet fast escalation requirements, and deliver conversations that felt natural across multiple languages.

Decagon stood out in three main areas. First, the flexibility to modify agent behavior without code meant that Mercado Libre’s team could iterate quickly, adjusting Agent Operating Procedures (AOPs) in natural language without always needing to pull in engineering resources. Second, Decagon’s real-time analytics through Watchtower replaced their existing “black box” system and gave the team visibility into conversation quality, intent patterns, and production performance. Third, the conversation quality itself. Rather than the scripted, mechanical experience of IVR, Decagon’s agents delivered natural, flowing dialogue that customers responded to with confidence.

Delivering a high-quality AI experience in Portuguese presented a distinct challenge. Early in the deployment, the team identified language quality nuances that required careful tuning - not just translation accuracy, but the natural rhythm and feel of the conversation. Decagon brought on a Brazilian QA team member specifically to evaluate and improve the Brazilian-Portuguese experience, working with Mercado Libre to refine the agents until the quality met the bar: natural, fluent, and not sounding foreign to native speakers.

The same discipline applied to guardrails. In a regulated environment, fast and reliable escalation is non-negotiable. Decagon’s guardrails framework gave Mercado Libre’s team the controls they needed to define escalation behavior clearly, and the confidence that those rules would be followed consistently at scale.

The Results: delivering speed of iteration at scale

Mercado Libre rolled out Decagon progressively, increasing the volume of interactions handled by the agent week over week as confidence in the system grew. That gradual ramp reflected a disciplined approach: measure, learn, and scale.

Beyond the operational gains, the implementation meaningfully changed how the team accesses and acts on customer data. With Decagon, conversation insights became faster and easier to surface — allowing the team to identify patterns in customer needs and feed those learnings into ongoing improvements across the CX operation far more efficiently than before.

Watchtower has become central to how the team operates day-to-day. The team uses the tool for technical monitoring, real-time production checks during product changes, intent analysis, and identifying opportunities for the agents to resolve issues independently.

“What we have now is the ability to understand why our customers are reaching out. That insight simply didn’t exist before.” 

What’s Next: an intelligent conversation as the new front door

Mercado Libre is moving toward a future where every customer interaction begins with an intelligent conversation, not a generic routing decision.

The work continues — expanding capabilities, improving resolution rates, and letting the data guide what comes next.