Kavak's AI Pivot: Replacing Humans with Agents
Carlos Garcia Otati details how Kavak transitioned from a 10,000-employee growth model to an AI-driven operation. The strategy involved abandoning co-pilot tools for autonomous agents, accepting a year of flat growth to achieve 90% automation in customer interactions.
The Strategic Pivot to Autonomous AI
Kavak, a vertically integrated automotive marketplace operating in Latin America and the Middle East, executed a radical operational transformation. Facing a market implosion and capital constraints, CEO Carlos Garcia Otati shifted the company from a high-growth, human-intensive model to an AI-driven operation. The core insight was that traditional "co-pilot" AI tools failed due to low employee adoption. Instead, Kavak deployed autonomous agents directly into critical business funnels, handling over 90% of customer interactions.
Operational Resilience and the Pain of Transition
The transition was not seamless. Kavak experienced a full year of flat growth as it restructured its operations. During this period, key performance indicators deteriorated as agents struggled to match human performance. However, the leadership team maintained discipline, refusing to revert to hiring more humans. They focused on building an "ontology" of data and communication pathways that allowed agents to solve complex edge cases, such as underwriting loans and managing warranty claims, rather than just answering simple queries.
Building for the Future, Not the Present
A key strategic principle was designing systems for future AI capabilities. Rather than optimizing for current model limitations, Kavak built infrastructure that could leverage advancements in AI, such as moving from ChatGPT 4 to hypothetical future versions. This approach allowed the company to scale its AI capabilities rapidly once the technology matured, achieving performance levels 1.5 times better than their best human agents in specific funnels.
Leadership and Market Context
The pivot was driven by the unique challenges of emerging markets, where fraud is rampant and financing penetration is low. By solving these fundamental problems with AI, Kavak created a defensible moat. The CEO emphasizes the importance of "firing yourself" annually to ensure leadership aligns with the company's evolving needs. This self-reflective practice helps leaders adapt to the rapid changes inherent in AI-driven businesses, ensuring they remain effective as the company scales from a startup to a global enterprise.
Key insights
-
AI co-pilot tools often fail due to low employee adoption. Autonomous agents deployed directly into business processes yield higher operational impact.
Impact: Companies can bypass the adoption barrier by integrating AI directly into workflows, leading to faster efficiency gains and reduced reliance on human labor for routine tasks.
-
Transitioning to AI-driven operations requires accepting a period of flat growth and KPI deterioration. This is a necessary phase before achieving superior performance.
Impact: Leaders must build organizational resilience to withstand short-term performance drops, ensuring long-term scalability and cost efficiency.
-
Building AI infrastructure for future model capabilities, rather than current ones, allows companies to leverage rapid technological advancements without rebuilding systems.
Impact: This forward-looking approach ensures that AI investments remain relevant and scalable as models improve, providing a competitive advantage in fast-moving tech landscapes.
-
Deploying AI in complex, high-friction areas like underwriting and warranty claims creates a stronger moat than using it for simple customer service tasks.
Impact: Solving hard problems with AI builds robust data ontologies and skill sets that are difficult for competitors to replicate, enhancing long-term defensibility.
-
Regular self-assessment by CEOs, including "firing themselves" to evaluate fit for the current company phase, is crucial for navigating rapid AI-driven changes.
Impact: This practice ensures that leadership remains aligned with the company's evolving needs, fostering adaptability and strategic clarity in dynamic environments.
Action items
-
Audit current AI tools to identify low-adoption co-pilots. Replace them with autonomous agents in critical business funnels.
Impact: This shift can significantly improve operational efficiency and reduce dependency on human labor for routine tasks, leading to cost savings and faster service delivery.
-
Prepare the organization for a period of flat growth during AI transition. Set clear expectations for KPI deterioration and recovery timelines.
Impact: Managing expectations helps maintain stakeholder confidence and ensures the team remains focused on long-term goals rather than short-term metrics.
-
Design AI infrastructure to be compatible with future model capabilities. Avoid over-optimizing for current model limitations.
Impact: This approach ensures that AI investments remain relevant and scalable as technology advances, providing a competitive edge in fast-moving markets.
-
Identify complex, high-friction areas in your business and deploy AI agents to solve these problems first.
Impact: Solving hard problems with AI builds robust data ontologies and skill sets that are difficult for competitors to replicate, enhancing long-term defensibility.
-
Implement a regular self-assessment process for leadership. Evaluate fit for the current company phase and make necessary persona shifts.
Impact: This practice ensures that leadership remains aligned with the company's evolving needs, fostering adaptability and strategic clarity in dynamic environments.
Quotes
“When you're building AI, the first thing that you need to build is the brakes of the system.”
“We made the typical mistake that everybody's making that is you know, we build these co-pilot tools to give them to our teams so they could just use them and just provide a better customer experience.”
“We don't build for chat GPT 4. We build for Chat GPT 7, right?”