Kavak's AI Overhaul: An Agent Per Customer

In a significant pivot towards an AI-native future, Kavak, the used car marketplace operating in Latin America, is fundamentally restructuring its operations around AI agents. Ali Masa, Head of AI at Kavak, detailed this ambitious transformation on the ACC podcast, revealing a strategy that bets on building a dedicated AI agent for every customer.

The Vision: superhuman agents

Masa explained that Kavak's vision is to create "superhuman agents" that would outperform the best human employees across all critical dimensions, including conversion rates, lifetime value, and customer experience. This ambitious goal necessitates a complete redesign of the company's architecture and processes.

An Agent Per Customer, Not Per Task

Unlike many companies experimenting with AI for specific tasks, Kavak is committed to a "person-per-customer" agent model. "Every day between 100 and 200,000 agents get instantiated specifically for this customer with its own virtual machine," Masa stated. These agents are designed to remember years of customer interaction, from website visits to past calls, to develop a long-term strategy for maximizing customer lifetime value.

The Transformation Playbook: Three Key Decisions

Masa outlined three critical decisions that underpinned Kavak's transformation:

• Redesigning Around Agents: The initial instinct for many companies is to bolt AI onto existing structures. Masa emphasized that a successful AI transformation requires a complete company redesign, including rebuilding APIs and systems to enable agents to perform tasks.
• Building superhuman agents: The bet is that these agents, by learning from customer interactions and fine-tuning, will outperform human counterparts. "We put them in front of the hardest problems," Masa noted.
• Shifting Measurement Metrics: Kavak moved from transactional KPIs (e.g., cars bought, parts sold) to relational metrics focused on maximizing customer lifetime value across a broad customer base.

The 'Jedi Academy' and Upskilling

Addressing the role of humans in this AI-driven future, Masa highlighted Kavak's internal "Jedi Academy." This program trains everyone, from the CEO to AI engineers and mechanics, in AI capabilities. "We train everyone, and after six weeks, they launch state-of-the-art agents to production," he said. This initiative aims to equip the workforce with the skills to collaborate with and manage AI agents, ensuring they remain relevant in the evolving organizational structure.

AI Sales Agents Outperform Humans

Kavak's agents are not focused on customer support but on sales, a critical and complex function. Masa explained that the agents are designed to be "experts" in multiple areas, such as financing, car advisory, and insurance, effectively creating a "mega expert" for the customer. This has led to significant improvements, including a tripling of NPS and customer satisfaction scores, and a conversion rate that is 2.1 times higher than that of their human sales teams.

Navigating Financial Services with AI

The company is also successfully deploying AI in regulated financial services. Kavak approves car loans in under three minutes, a stark contrast to the usual two-month wait in Mexico and other emerging markets. This is achieved by leveraging extensive customer and car data, enabling personalized interest rates and loan amounts. Masa emphasized that understanding customers deeply throughout the process is key to driving conversion and retention.

The 'AI CEO' and Future of Organizations

In a bold experiment, Kavak placed an AI agent in charge of managing a city in Mexico, Quavaka, as an "AI CEO." In its first month, this AI managed to increase profits by 50%. The agent demonstrated an ability to analyze data, make forecasts, and micromanage execution. This experiment highlights the potential for AI to take on leadership roles, although Masa noted that roles requiring physical dexterity and sensory input, like mechanics, are still reliant on human expertise, albeit augmented by AI sidekicks.

The 'Creative Destruction' of AI

Masa drew parallels to past industrial revolutions, like electricity and computing, where superficial adoption yielded limited gains. He argued that to achieve a "10x improvement," companies must be willing to "destroy and rebuild everything" around AI's core strengths. For founders, this presents a significant opportunity to build net-new companies that are natively designed for the AI era, disrupting existing markets through this fundamental architectural shift.

Advice for Future Founders

For aspiring entrepreneurs, Masa offered a clear message: "This is the most exciting time in human history." He emphasized the democratization of powerful AI tools, making it easier than ever for individuals to build and solve problems. His advice is to "go for it, but go for it deep," encouraging founders to imagine the future of AI and build for that reality, understanding that even linear progress in AI capabilities can lead to exponential value creation.