Kavak Runs Entire Region on AI Agents, Treating Humans as Tool Calls
astrange · x · 2026-08-10
Kavak, the largest used car marketplace in Latin America, is operating an entire region using AI agents, including functions typically handled by a CEO. These agents autonomously negotiate and sell financial products. Alejandro Maza, Chief Product & AI Officer, has built one of the most advanced enterprise agent implementations.
Key architectural highlights include:
- Persistent Omnichannel Memory: Every customer receives a dedicated agent that maintains memory continuity across voice, WhatsApp, and web platforms.
- Humans as Tool Calls: When an agent gets stuck, it doesn’t transfer the user to a human queue. Instead, it "calls" a human for the answer and integrates that data back into the loop for continuous system learning.
- Goal-Oriented Execution: Instead of relying on narrow task evals, agents are given a single top-level directive: maximize customer Lifetime Value (LTV).
The team also completely re-architects its tech stack whenever AI infrastructure advances to maintain its edge.
Related event: Used Car Platform Kavak Transforms Operations with AI Agents(4 posts)→
More from coding & agent
- swyx: Delete Your Agent Skills to Avoid Nasty Context Pollution — pvncher · 2026-08-11
- Trigger.dev Launches Durable Chat Agent Surviving Refreshes and Crashes — addyosmani · 2026-08-11
- DCAS: Decoupling CLI Agent Scaffolding to Internalize Planning — centre-for-swe · 2026-08-11
- Building Effective Coding Agents: Basic Tools Are All You Need — deliprao · 2026-08-11
- Edit Banana: Open-Source Framework Turns Static Images into Editable DrawIO Files via SAM 3 — tom_doerr · 2026-08-11
- Needle 2: A 14MB Agentic LLM Hitting 500 Tokens/sec on Raspberry Pi 5 — Henrie_the_dreamer · 2026-08-11