Pedro Franceschi, CEO of Brex, a company valued at $5 billion, argues that the dominant framework for building AI agents is structurally flawed. The fix: stop thinking in terms of agents and start thinking in terms of employees, each with a defined job, specific skills, a manager, and a budget.
The interview includes a live demo of OpenClaw, the tool Pedro uses personally to manage both his company and his daily life. That demo alone is worth the runtime. Watching a sitting CEO operate an AI system in real time, on real workflows, reveals more about practical implementation than any framework post ever will.
The broader argument here is not about tooling. It is about mental model. If your agents are failing, it is probably because you gave them a task instead of a role. Pedro's employee model imposes accountability structures that tasks do not have. Read the full breakdown to understand where most builders go wrong before they write a single line of code.
[WATCH ON YOUTUBE →]