a16z

This episode explores why AI agents are harder to deploy in enterprises than in startups. The speakers discuss integration limits, legacy systems, access control, headless SaaS, human-in-the-loop workflows, and why AI may reshape jobs rather than simply replace them.

In this episode, the conversation explores what it really means to build software and companies in an AI-native world. Rather than treating AI as a simple replacement for existing tools, the discussion looks at the deeper shift happening underneath: how agents may change software interfaces, reshape workflows, pressure companies to expose better systems and APIs, and create entirely new economic models around computation, usage, and automation. At the same time, the speakers argue that AI adoption will likely be far slower and messier inside large enterprises than many people in Silicon Valley expect, because real-world organizations still have to deal with legacy systems, security, permissions, and operational risk. Overall, the episode is a thoughtful debate about where AI agents will truly create leverage, what kinds of businesses may benefit first, and why the biggest opportunities may come not from replacing old systems overnight, but from gradually rebuilding how software, work, and digital infrastructure operate.