Why did Kavak build four pillars (e-commerce, reconditioning, financing, logistics) underneath the main business?

Founder Kavak saw that success in used cars required more than a single sales channel. They built four foundational pillars to handle end-to-end operations at scale: an e-commerce platform, a reconditioning/warranty machine, a financing engine, and a logistics network. This infrastructure allowed them to handle edge cases and scale. Timestamp: 1258

Why start in Mexico City when expanding, rather than going broad from the start?

The team chose Mexico City because it is the largest city in North America, with a large and dense market. They aimed to solve the core problem deeply in one city first, build the necessary data capabilities, and only later scale to other markets. Timestamp: 1410

What does the Spotify/Amazon/Toyota/City Bank analogy mean for Kavak?

The analogy describes Kavak's ambition to build a comprehensive, integrated platform that combines marketplace reach (Spotify/Amazon) with product reliability (Toyota) and financial services (City Bank). It reflects creating a safe, financed, and trusted car ownership path rather than a simple buy/sell. Timestamp: 1062

When did Kavak start thinking about AI and what was the 2017 memo about?

Kavak began planning for AI in 2017, anticipating high-edge-case complexity as the company scaled. The memo outlined a shift toward a 'highway of information' and an ontology to coordinate multiple business units and AI-enabled roles, long before consumer AI tools became mainstream. Timestamp: 1683

How did Kavak implement AI in the org, and what was the 'funnel-by-funnel' approach?

Kavak initially built tools for many teams but found adoption lacking. They then deployed AI agents funnel by funnel, starting with the hardest edge cases (like underwriting and post-sale service), with humans and agents working together and an orchestrator guiding who solves what. This gradual approach yielded parity, then improvement, and finally scale. Timestamp: 2210

What is Kavak's 'GPT7' mindset about AI models?

Carlos explains a philosophy of not building for the current generation (e.g., GPT-4) but preparing architectures to leverage future models (GPT-7+). They prioritize robust data, scalable infrastructure, and forward-compatibility so that advances in AI can be absorbed as soon as they appear. Timestamp: 2543

How did Kavak survive the 2020–2021 downturn and shift toward AI streaming?

The market collapse forced Kavak to rethink growth, lean operations, and profitability. They slowed down to reach sustainable scale, expanded AI capabilities, and ultimately aimed to become significantly profitable while maintaining impact for users. Timestamp: 2720

What advice does Kavak's Carlos have for founders in emerging markets?

Find a real, impactful problem to solve, not just something that feels nice. Be willing to think big and to relocate or adapt if needed, but ensure the problem serves a meaningful portion of the population. Stay purpose-driven through the highs and lows. Timestamp: 3082

What daily habit helps Carlos stay motivated during tough times?

Carlos emphasizes zooming in on user frictions, removing one friction per week, and keeping a long-term perspective by zooming out to see the bigger trajectory. He also notes that doing the daily 'friction removal' is how durable growth is built. Timestamp: 3228

Key Moments

AI Copilots Are a Dead End. Here's What Actually Works | Kavak CEO

a16z

Gaming4 min read72 min video

Mar 5, 2026|1,672 views|44|3

Save to Pod

Key Moments

TL;DR

LATAM car market fix through four pillars and AI-driven scale.

Key Insights

Build a four-pillar infrastructure (e-commerce, reconditioning/warranty, financing, logistics) to scale a complex marketplace in LATAM.

Adopt a data-first ontology; capture edge-case dynamics to train AI, enabling reliable, scalable operations.

Pivot from generic co-pilot tools to a full AI agent ecosystem with an orchestration layer for end-to-end processes.

Expand globally by solving the problem first, not chasing markets; chosen hubs based on market size, fraud, financing, and ticket size.

Lead with resilience and long-term vision, balancing lean growth with profitability during market downturns, while transforming culture.

ORIGINS, MISSION, AND THE LATAM MARKET PROBLEM

Carlos Kavak's journey starts in Venezuela and is marked by constant movement—14 schools and a pattern of reinventing himself. This upbringing taught him to embrace change and see opportunity in disruption. He opened his eyes to a hard truth: in Latin America, the used-car market was largely informal, with roughly 90% of transactions taking place outside formal channels and up to 40% ending in fraud. That reality limited social mobility and kept financing out of reach for most people. Kavak's mission emerged from this: create an end-to-end platform that manages the car's lifecycle—from selling to buying, financing, warranties, and ongoing services—so everyday people can access safer, more affordable mobility. The personal narrative—seeing how ownership transformed life, and witnessing his family’s resilience—shaped a belief that a platform, not just a marketplace, could change lives. The result was a purpose-driven drive to reduce friction, increase trust, and unlock middle-class potential through car ownership, turning a chaotic informal market into a trusted, scalable ecosystem.

BUILDING INFRASTRUCTURE AROUND A HARD PROBLEM

Kavak's founders recognized that solving the LATAM used-car problem required more than a marketplace. They built four interconnected pillars to create a robust, scalable engine. First, an e-commerce platform to enable end-to-end car transactions; second, an in-house reconditioning and warranty machine capable of scaling up to hundreds of thousands of vehicles; third, a financing engine to underwrite customers and provide access to credit; and fourth, a logistics network to move cars securely and efficiently. LATAM presented unique obstacles: limited payment rails, scarce reliable logistics, and little formal financing penetration. The team invested in foundational capabilities—software, hardware (for maintenance operations), and data systems—so the entire lifecycle could be managed seamlessly. This multi-layered approach was essential to move from a fragile, broker-driven market to a trust-based, repeatable process that could scale across regions.

FROM CO-PILOTS TO A GLOBAL ORCHESTRATED AI ARCHITECTURE

A turning point came in 2017 with Kavak’s emphasis on a 'robot pipeline'—an idea to embed AI across the organization to handle edge cases and unlock scale. The company started with a strong data ontology and an operating system to capture every interaction, failure, and outcome. Early attempts at deploying generic AI copilots failed to yield adoption, revealing that a company-wide AI strategy requires more than tools; it needs an architecture and a clear allocation of where humans and AI should act. Kavak designed a layered approach: build an orchestrator and deploy task-specific AI agents that handle niches such as underwriting, warranty processing, and post-sale service, initially in high-friction areas and gradually expanding. The aim was parity with human performance, then surpass it as models improve. This method created a symbiotic workflow where AI agents augment human teams, delivering faster, more consistent service while capturing the data necessary to continuously improve.

GLOBAL EXPANSION: A PROBLEM-FIRST APPROACH AND MARKET SELECTION

Kavak’s global strategy has been about solving a universal problem, not simply chasing geographic growth. Mexico City was chosen as the initial launchpad due to its sheer market size and concentrated opportunity, allowing the team to deeply understand the problem and build the required data-driven infrastructure before scaling. When expanding to other cities, Kavak applied strict criteria: target markets with substantial total addressable value, significant fraud and financing gaps, and enough average car ticket size (roughly above a threshold) to ensure unit economics were sustainable. The company mapped expansion by the density and depth of the problem, not merely population. This disciplined approach meant that each new market required years of groundwork to replicate Mexico City’s successful infrastructure. Kavak’s global footprint—including ventures in the Middle East—reflects a measured, problem-driven globalization rather than a speed-first strategy.

CULTURE, PROOF OF REDEMPTION, AND THE PATH TO PROFITABILITY

The journey through rapid growth and market downturns tested Kavak's culture and resilience. The industry’s swift shifts demanded a durable, long-term plan: slow down to go faster later, invest in AI-enabled scalability, and maintain a steadfast commitment to the mission of helping people access cars and improve their lives. The Netflix versus Blockbuster analogy framed the leadership’s view: they needed to shift from a traditional, physical-inventory model to a streaming-like, AI-enhanced capability that could meet demand at scale. The downturns—where capital slowed and markets contracted—made this shift urgent. Kavak leaned into lean growth while continuing AI investments, ultimately steering toward profitability and sustainable growth. The impact is tangible: about 40% of customers buy their car for the first time, illustrating real social change. The culture emphasized patience with experimentation, clear communication, and a belief that AI could redefine operations, not merely automate them. As the company continues to evolve, it remains focused on long-term value creation for customers, employees, and society.

Mentioned in This Episode

●Software & Apps

●People Referenced

Kavak AI & Growth Cheat Sheet

Practical takeaways from this episode

Do This

Start with an ontology and data highway before heavy AI rollout.

Roll out AI funnel-by-funnel, focusing on edge-case handling first.

Build an operating system to orchestrate humans and agents.

Aim for parity to begin with, then pursue continual improvement as models advance.

Relentlessly remove friction for users on a weekly basis.

Avoid This

Don't deploy co-pilot tools organization-wide before adoption and real-world use.

Don't replace humans everywhere at once; establish where AI adds most value first.

Don't neglect unit economics and core pricing/underwriting pillars while scaling AI.

Don't confuse ambition with speed—ensure data infrastructure scales with expansion.

Common Questions

Kavak operates as an all-in-one car marketplace that covers the lifecycle from selling to buying, financing, warranties, services, and even ticket payment. It claims a leading e-commerce auto presence in Latin America with around 3,500 people and roughly 10,000 transactions per year, across Mexico, Brazil, Chile, Argentina, and the Middle East. Timestamp: 41