Kavak's AI Agents Drive 2.1x Human Conversion, 'AI CEO' Lifts Profits 50%
a16z · youtube · 2026-08-10
In an a16z podcast, Kavak's Chief Product & AI Officer detailed how the Latin American used-car marketplace completely rebuilt its organization around AI agents.
Key Practices & Data:
- Full Agent Integration: Agents now handle 96% of customer interactions and 95% of transactions. AI sellers achieve 2.1x the conversion rate of human teams, and car loans are approved in 3 minutes.
- The 'AI CEO' Experiment: An AI system managing one city increased local profits by 50% in its first month.
- Architecture Overhaul: After spending two years building complex multi-agent graphs, the team destroyed that work and pivoted to a 'One Agent Per Customer' architecture.
- Engineering Focus: Kavak dedicates as much engineering effort to agent evals as it does to building the agents themselves.
Organizational Shift:
- Simply giving employees AI tools is insufficient; the massive opportunity lies in redesigning the organization itself.
- Through a 'Jedi Academy' program, everyone from executives to mechanics is trained to build with AI.
- Traditional organizational structures and metrics must evolve when agents do the majority of the work.
Related event: Kavak Restructures Business Around AI Agents(3 posts)→
More from coding & agent
- Prime-agent Harness Tested: GLM 5.2 Shows Strong Results on FutureSim Q2 — a1zhang · 2026-08-10
- Ante 0.2 Ships a 15MB Local Coding Agent Managing llama.cpp Offline — Exciting-Camera3226 · 2026-08-10
- Why You Should Know About SSRF Vulnerabilities in the Agentic AI Era — andersonbcdefg · 2026-08-10
- Surviving the AI Era: From Task Executor to System Architect — mdancho84 · 2026-08-10
- Karpathy's Open-Source Project Sparks Debate: Will AI Agents Replace Data Scientists? — mdancho84 · 2026-08-10
- Rust-based agentic framework Steffi AI for local models, focusing on long-running tasks and control — ahstanin · 2026-08-10