Enterprises will move slowly on deploying AI agents due to massive security and integration risks with legacy systems. Startups, with less to lose and cleaner stacks, will adopt agent-based workflows rapidly, creating a significant competitive advantage and widening the gap between incumbents and challengers.

New AI coding agents excel at creating fresh applications but struggle with complex, existing codebases. This gives flexible startups a significant advantage over large companies burdened by legacy systems, fundamentally rebalancing power in the tech industry.

Kavak rethinks "human-in-the-loop." When an AI agent gets stuck, it doesn't just escalate the task to a human queue. Instead, the agent makes an API call to a human for help, retains ownership of the problem, and learns from the interaction. This closes feedback loops and makes human teams a resource for the primary agent.

Vinod Khosla argues that data lock-in is a weak moat for personal agents, as users can simply instruct a new agent to handle the migration. The durable competitive advantage will be trust. This gives startups an edge over tech giants like Meta, which suffer from a brand of lack of trust with consumers.

Building effective agents requires intensive, custom work for each client—data cleansing, training, and deployment by skilled engineers. Large incumbents lack the agility and cost structure to provide this bespoke service, creating an opening for focused startups who can afford the human capital.

Assuming technology is increasingly commoditized, a startup's defensibility comes from relentless execution. This includes providing a seamless onboarding experience and superior customer service—areas where large, impersonal frontier labs are unlikely to compete effectively.

Anyone can build a simple "hackathon version" of an AI agent. The real, defensible moat comes from the painstaking engineering work to make the agent reliable enough for mission-critical enterprise use cases. This "schlep" of nailing the edge cases is a barrier that many, including big labs, are unmotivated to cross.

The success of new AI startups is driven by a desire among managers to replace human-led processes with autonomous agents. Customers don't want AI to make their teams slightly better; they want an agent that eliminates the need for the team entirely. This is a demand most incumbent software companies misunderstand and fail to serve.

Asana's CEO argues its key differentiator is a "multiplayer mode" where entire human teams can collaboratively train and correct an AI agent within a project. This contrasts with typical one-on-one chat interactions, creating a unique, compounding learning environment for the agent that Asana believes cannot be easily replicated.