From Copilots to Agents: Rebuilding the Company Around AI

Most companies deploying AI are building expensive toys. Carlos García Ottati built a profit machine. The CEO of Kavak, Latin America's largest online used car marketplace, didn't just add chatbots to his customer service. He rebuilt his entire company around autonomous agents that now handle 95% of customer interactions while the business grew 4x and hit profitability during one of the worst funding environments in recent memory.

García Ottati's journey to AI transformation started with a hard lesson about human reliability. As a teenager running a lawn care business in Venezuela, he learned that "most of these kids didn't show up at the time that they needed to mow that lawn. So I was always under delivering to my users, to my customers." The solution wasn't better hiring or incentives. It was changing the entire business model to remove human unreliability from the equation. Two decades later, he applied the same principle to Kavak, but this time the unreliable humans were his own employees, and the solution was AI agents.

The conventional wisdom says you should start AI adoption with copilot tools that assist human workers. García Ottati tried this approach and watched it fail spectacularly. "In late 2022 or early 2023, we spent a lot of time building these tools for everybody in our organization, and we gave them to them. And we realized very quickly they didn't adopt them," he explains. Even with perfect information and well-designed interfaces, employees simply ignored the AI assistance. This forced a fundamental realization: humans don't want to be managed by machines, even helpful ones. They want to either do the job themselves or have the machine do it entirely.

That insight led García Ottati to skip the copilot phase entirely and jump straight to autonomous agents. But the transition required building what he calls "the brakes of the system" first. Kavak had collected millions of data points about customer behavior, chat interactions, and failure points, but "you couldn't extract and automatically know" what it all meant before the ChatGPT era. The breakthrough wasn't just having the data, but being able to process it intelligently to understand every customer's context and intent.

The timing of this AI transformation proved critical. García Ottati faced an impossible business challenge in 2022: "we couldn't just shrink to profitability, we needed to grow, while becoming more lean and efficient at the same time" during a market crash where funding "goes down 99.9%." Traditional cost-cutting would have killed growth. Traditional growth strategies would have burned through cash reserves. AI agents provided a third option: dramatically improve efficiency while scaling customer interactions.

García Ottati frames this transition using a Netflix analogy that reveals how he thinks about competitive advantage. "Incumbents are Blockbuster, Kavak was videotapes sending to your user, and that's probably traditional software. And selling online and underwriting online and everything. And then, you know, moving to AI is sort of like that streaming transition." The key insight is that each technological shift doesn't just improve existing processes, it enables entirely new business models that make the previous generation obsolete.

The results speak for themselves: 90-95% of customer interactions now run without human intervention, the company grew 4x while achieving profitability, and Kavak maintains its position as the dominant player in Latin American used car sales. But García Ottati's approach carries lessons beyond marketplace businesses. His experience suggests that successful AI adoption requires three elements that most companies miss: willingness to abandon incremental improvements, the infrastructure to process unstructured data at scale, and the operational discipline to build constraints before deploying autonomous systems.

The broader implication is that AI adoption follows the classic innovator's dilemma pattern. Companies that try to enhance existing workflows with AI assistance will lose to companies that rebuild their operations around AI capabilities. García Ottati didn't just automate customer service, he reconceived customer service as a data processing problem that machines solve better than humans. The question for other executives isn't whether to adopt AI tools, but whether they're willing to rebuild their companies around what AI actually enables rather than what it replaces.

Watch for this pattern to accelerate across industries. The companies thriving in 2025 won't be the ones with the best AI-assisted employees. They'll be the ones that figured out how to eliminate human bottlenecks entirely while maintaining the quality and judgment that customers expect. García Ottati proved it's possible to thread that needle, but only if you're willing to bet your entire business model on the outcome.