Case Study: From AI Copilots to Agents
Carlos García Ottati, CEO of used car marketplace Kavak, revealed a year-long transition from ineffective AI "copilots" to deeply integrated AI agents. The company replaced its assistance tools with agents that now handle over 90% of customer interactions. Ottati stated that while the transition resulted in a year of flat growth, it was a necessary step to unlock massive operational scale and efficiency in a high-fraud market.
- The transition to AI agents that now handle 90-95% of customer interactions was followed by a period of 4x growth, after an initial year of flat performance during the changeover. - To further invest in AI and technology, Kavak recently secured $300 million in a Series F funding round, which was led by the a16z Growth fund from Andreessen Horowitz. - The company's focus on AI is a direct response to the challenges of the used-car market in Latin America, which has historically been characterized by high levels of fraud and informality. - Beyond customer service, Kavak applies machine learning for critical operations such as determining vehicle prices, conducting credit scoring analyses for financing, and automating internal processes. - One of Kavak's specific AI implementations involves a partnership with Autonoma AI, which uses intelligent agents to simulate user journeys on Kavak's platform, proactively identifying and addressing potential issues before they affect customers. - Kavak's CTO, Fernando Scasserra, brings extensive experience from his 15-year tenure at Mercado Libre, where he was part of the technical team's growth from 40 to 8,000 members. - The operational improvements, including the AI transition, led to Kavak achieving its first full month of consolidated global profitability in December 2025. - The quality and reliability improvements are reflected in their metrics, with only 5% of vehicles requiring post-sale services and just 1% of customers choosing to return or exchange their vehicle.
