Aug 10, 2026 · 37 min
Kavak rebuilds its marketplace around persistent AI agents
The Self-Improving Company | Kavak's AI Playbook
Kavak’s experiment shows how AI can reshape not just workflows, but the structure, leadership, and learning systems of an entire company.
- 1Kavak gives each customer a persistent agent that manages interactions, transactions, financing, and other complex decisions.
- 2The company measures agents in production and uses difficult human cases to improve systems rather than merely resolve individual escalations.
- 3Kavak argues that AI-native companies can outcompete incumbents by redesigning organizations instead of layering tools onto existing processes.
Don't miss
Alejandro describes testing an agent in a CEO-like role for a Mexican city, illustrating Kavak’s willingness to probe AI’s capacity for senior responsibility.
The brief
Alejandro Maza Ayala explains how Kavak moved beyond adding AI tools to existing workflows, rebuilding its used-car marketplace around persistent agents with customer-specific context.
Kavak’s agents handle most customer interactions and transactions, guiding vehicle selection, financing, insurance, trade-ins, and loan decisions rather than serving as simple support bots.
The company evaluates agents in production at scale, with hundreds of thousands instantiated daily, and tests whether AI can take on increasingly senior responsibilities.
Kavak’s human feedback loop turns difficult cases into system knowledge: people resolve problems, then capture what happened so agents and the broader organization can improve.
The larger argument is organizational: AI-native companies may gain an advantage by flattening teams, training everyone to build agents, and designing for continuous improvement from the start.
Featuring
ChatGPTListen to the full episode and explore every guest, topic, and moment on PodLume.