What are the key takeaways from this a16z Podcast episode?

Key insights include: **Copilot tools fail at adoption:** KAVAK built AI copilot tools for employees in late 2022 and found staff simply did not use them. The fix was replacing copilots entirely with agents deployed directly into customer-facing funnels — starting with the hardest problems first, such as loan underwriting and post-breakdown support, not generic customer service.; **Expect one year of flat growth during AI transition:** KAVAK went from 300% annual growth to zero growth in 2023 while restructuring around agents. Output KPIs — sales, purchases, financing — deteriorated before recovering. The strategy required committing to a single funnel with no Plan B until agents reached parity, then moving to the next funnel.; **Build for the next model, not the current one:** KAVAK's engineering philosophy targets future model capabilities rather than optimizing for today's. Once an agent reached 1.5x human performance in a funnel, the team stopped refining it and moved on, trusting that model improvements would compound gains without additional engineering effort.

What did Carlos Garcia Otati and Angela Strange discuss on a16z Podcast?

Carlos Garcia Otati, founder and CEO of KAVAK, details how his used car marketplace serving Latin America and the Middle East transitioned from 10,000 employees and 300% growth to a leaner AI-first operation, where agents now handle 90–95% of customer interactions across four vertically integrated business lines. Key topics include: **Copilot tools fail at adoption:** KAVAK built AI copilot tools for employees in late 2022 and found staff simply did not use them. The fix was replacing copilots entirely with agents deployed directly into customer-facing funnels — starting with the hardest problems first, such as loan underwriting and post-breakdown support, not generic customer service.; **Expect one year of flat growth during AI transition:** KAVAK went from 300% annual growth to zero growth in 2023 while restructuring around agents. Output KPIs — sales, purchases, financing — deteriorated before recovering. The strategy required committing to a single funnel with no Plan B until agents reached parity, then moving to the next funnel..

How long is this episode of a16z Podcast?

This episode is 59 minutes long. SignalCast provides an AI-generated summary so you can get the key insights in about 3 minutes.

a16z Podcast

From Copilots to Agents: Rebuilding the Company Around AI

February 18, 2026

59 min episode · 2 min read

Carlos Garcia Otati,Angela Strange,Gabriel Vasquez

Episode

59 min

Read time

2 min

Topics

Health & Wellness, Relationships, Investing

AI-Generated Summary

Published Feb 18, 2026

Key Takeaways

✓Copilot tools fail at adoption: KAVAK built AI copilot tools for employees in late 2022 and found staff simply did not use them. The fix was replacing copilots entirely with agents deployed directly into customer-facing funnels — starting with the hardest problems first, such as loan underwriting and post-breakdown support, not generic customer service.
✓Expect one year of flat growth during AI transition: KAVAK went from 300% annual growth to zero growth in 2023 while restructuring around agents. Output KPIs — sales, purchases, financing — deteriorated before recovering. The strategy required committing to a single funnel with no Plan B until agents reached parity, then moving to the next funnel.
✓Build for the next model, not the current one: KAVAK's engineering philosophy targets future model capabilities rather than optimizing for today's. Once an agent reached 1.5x human performance in a funnel, the team stopped refining it and moved on, trusting that model improvements would compound gains without additional engineering effort.
✓97% of company value is created after year 15: The most durable businesses compound value through daily 1% friction reductions for users, not breakthrough moments. KAVAK tracks this by noting that 40% of its buyers are purchasing their first car ever — a direct result of solving financing penetration, which sits at 5% in Mexico versus 90% in the US.
✓Annual CEO self-firing exercise: Garcia Otati formally fires himself each year, writes a job description for the ideal CEO of the next phase, then evaluates whether to rehire himself. The process forces identification of what to stop doing, what new skills to build, and what the company's stakeholders — employees, investors, family — actually need from leadership at that stage.

What It Covers

Carlos Garcia Otati, founder and CEO of KAVAK, details how his used car marketplace serving Latin America and the Middle East transitioned from 10,000 employees and 300% growth to a leaner AI-first operation, where agents now handle 90–95% of customer interactions across four vertically integrated business lines.

Key Questions Answered

•Copilot tools fail at adoption: KAVAK built AI copilot tools for employees in late 2022 and found staff simply did not use them. The fix was replacing copilots entirely with agents deployed directly into customer-facing funnels — starting with the hardest problems first, such as loan underwriting and post-breakdown support, not generic customer service.
•Expect one year of flat growth during AI transition: KAVAK went from 300% annual growth to zero growth in 2023 while restructuring around agents. Output KPIs — sales, purchases, financing — deteriorated before recovering. The strategy required committing to a single funnel with no Plan B until agents reached parity, then moving to the next funnel.
•Build for the next model, not the current one: KAVAK's engineering philosophy targets future model capabilities rather than optimizing for today's. Once an agent reached 1.5x human performance in a funnel, the team stopped refining it and moved on, trusting that model improvements would compound gains without additional engineering effort.
•97% of company value is created after year 15: The most durable businesses compound value through daily 1% friction reductions for users, not breakthrough moments. KAVAK tracks this by noting that 40% of its buyers are purchasing their first car ever — a direct result of solving financing penetration, which sits at 5% in Mexico versus 90% in the US.
•Annual CEO self-firing exercise: Garcia Otati formally fires himself each year, writes a job description for the ideal CEO of the next phase, then evaluates whether to rehire himself. The process forces identification of what to stop doing, what new skills to build, and what the company's stakeholders — employees, investors, family — actually need from leadership at that stage.

Notable Moment

During the AI transition, KAVAK sustained a full calendar year of flat revenue — down from 300% growth — while agents underperformed humans across critical funnels. Garcia Otati describes resisting the urge to hire humans back, treating each funnel as a no-fallback commitment until agent performance recovered.

Know someone who'd find this useful?

Episode Transcript

When you're building AI, the first thing that you need to build is the brakes of the system. It's the understanding on on where you're gonna deploy it and how you're gonna deploy it. We had thousands and millions of information lines about our users, and the first thing that we did was we just built this system that allowed us to understand everything about every user. We built these Copilot tools, and we realized very quickly they didn't adopt them. So we went funnel by funnel, putting agents in front of, like, critical aspects of the business. We put the agents in front of our users really quickly. It it was very painful. If you see the best companies out there, 97% of their value gets created after year fifteen, To year one five, 97% of their value. 40% of used car transactions in Latin America end in fraud, not the lemon problem you find in The United States. Kidnappings, forged documents, stolen vehicles. And in Mexico, only 5% of the used car market is financed, compared to 90% in The US. Carlos Garcia Otati founded KAVACA after being defrauded twice trying to buy and sell cars in the region. The company he built handles everything, buying, reconditioning, selling, financing, warranties, and logistics. Four separate businesses underneath one consumer experience, operating across Latin America and The Middle East. By 2021, Kovac had 10,000 employees and was growing at 300. Then capital disappeared and the company had to rethink everything. Instead of retreating, Carlos bet on AI, replacing co pilot tools that employees wouldn't adopt with agents that now handle more than 90% of customer interactions. The transition took a full year of flat growth. A sixteen z's Angela Strange and Gabriel Vasquez speak with Carlos Garcia Tati, founder and CEO of Kavac. Carlos, for those who aren't familiar with KAVAK, can you introduce the company and tell us a little bit about the scale of where you're at? Yeah. Sure. Well, first of all, thanks for having me here. It's amazing. KAVAK is an all in one marketplace to do anything related to your car. So we work out of Latin America. We work in Mexico, Brazil, Chile, Argentina. We also have operations in The Middle East. And what we essentially do is we manage the lifetime of our users from the moment that they wanna sell their car to the moment that they wanna buy a car. If their car breaks down, we're there as well. So we manage the end to end process of a user's journey while they own their car. And we do everything. We buy these cars, we recondition them, we then sell them online, we provide the financing to our users as well, We manage their warranties, their services throughout the process. We pay their tickets for them if they want us to do that and we just make sure that we're just building that relationship with our user as they go along. …

Get the full transcript (11,260 words) + summary by email — free

One-time email with the complete transcript and AI summary of this episode. No account needed.

One email, no spam. We’ll also show you what SignalCast does.

Browse all a16z Podcast transcripts →