Prioritize investing in AI-native startups built from scratch over legacy companies merely layering basic chatbots onto existing workflows.

Focus your capital on businesses utilizing autonomous agents to execute complex, high-value transactions like loan underwriting and auto sales rather than simple customer support.

Target private companies and upcoming IPOs successfully deploying hundreds of thousands of daily per-customer agents to maximize long-term lifetime value, similar to pioneers like Kavak.

Allocate research toward companies developing proprietary evaluation (evals) infrastructure, as rigorous testing frameworks are the primary bottleneck and key to scaling agentic systems.

Avoid traditional enterprises achieving only marginal efficiency gains through superficial AI adoption, as they face imminent disruption by fully autonomous competitors.

AI-Native Enterprise Infrastructure and Autonomous Agents (Unlisted/Private Sector)

• The discussion centers heavily on AI agents, specifically shifting organizational architectures from traditional human-managed workflows to fully AI-native companies powered by autonomous, long-running agents.
  • Companies like Kavak (a used car marketplace and fintech) have transitioned to an agent-first model where roughly 96% of customer interactions and 95% of transactions are handled entirely by AI agents rather than human employees.
  • Rather than using simple task-based workflows or multi-agent graphs, the guest highlights a shift toward instantiating hundreds of thousands of independent per-customer agents daily, each operating within its own virtual machine and tasked with maximizing customer lifetime value over long horizons.
  • The discussion emphasizes evals (evaluations) as the critical bottleneck and accelerator; successful implementation requires spending equal engineering time and money on rigorous evaluation frameworks as on building the agents themselves.
  • A major investment and operational thesis is presented: incumbent corporations struggling to adopt AI superficially (getting marginal efficiency gains) face disruption by AI-native greenfield startups designed from scratch around artificial superintelligence, mirroring historical industrial revolutions (such as the transition from steam-shaft factories to flat electric-powered factories).

Takeaways

• For investors evaluating enterprise software and AI applications, prioritize companies and startups that are building AI-native business models from scratch rather than legacy enterprises simply layering third-party chatbots (like ChatGPT or Claude) onto existing workflows.
  • Look for companies successfully implementing autonomous agents that handle high-value, complex transactions (such as auto sales, loan underwriting, and financing) rather than basic customer support ticketing.
  • Monitor the strategic importance of evaluation (evals) infrastructure; companies that build proprietary, rigorous evaluation loops to measure true business outcomes (like customer conversion and lifetime value) are significantly more likely to succeed in scaling agentic systems.