Skip to content
Agents tracked: 442 Downloads (7d): 202M down 19.3% GitHub stars: 10.4M Releases (7d): 693 Agent status: 1 with issuesSince last limit reset: Claude 28d · Codex 1d Updated Oct 21, 2026
Case study Mercado Libre

How Mercado Libre replaced its voice IVR with Decagon AI agents across LATAM

Mercado Libre, a Latin American ecommerce and fintech platform, replaced its rigid voice IVR with Decagon's voice agents. The team gained natural conversations, no-code control over agent behavior and visibility into why customers call.

The problem

Mercado Libre's voice support ran on an IVR with fixed menu options. It could not personalize responses or tell urgent requests from non-urgent ones. The team also had no visibility into why customers called, so the channel was hard to scale and improve.

How they did it

  1. Pick a voice AI platform that handles escalation requirements and natural multilingual conversation.
  2. Define agent behavior as Agent Operating Procedures written in natural language, so changes need little engineering help.
  3. Set guardrails that spell out when and how the agent escalates, which matters in a regulated environment.
  4. Tune Brazilian Portuguese quality with a native-speaking QA reviewer, covering rhythm and feel as well as translation accuracy.
  5. Roll out gradually, raising the share of interactions the agent handles each week as confidence grows.
  6. Use Watchtower analytics to monitor production, check changes in real time and analyze customer intents.

Results

  • The team can now see why customers reach out, an insight the article says did not exist before.
  • Watchtower is used daily for technical monitoring, production checks during product changes, intent analysis and finding issues the agents could resolve alone.
  • As reported by Decagon, the progressive rollout let Mercado Libre raise agent-handled volume week over week. The article gives no resolution or CSAT figures.

As reported by the source (Decagon customer story); AgentGid did not measure these figures.

Takeaway. Replacing a black-box IVR with a voice agent plus conversation analytics, rolled out in stages, gives support teams both better conversations and data on customer intent.
AgentGid's take

This suits a large support organization with voice volume, regulatory escalation rules and native-language QA staff, since the setup is rated advanced. Decagon has no public pricing (Custom, contact sales), and the only rollout result is vendor-reported, with no resolution or CSAT figures. Rasa (Free) is an open-source alternative for engineering-heavy teams, though its free tier covers just one bot and 1,000 external conversations a month.

The agent used here

Similar use cases

Guides