The CEO Must Be the Chief AI Officer

Pedro Franceschi argues that AI adoption is not an engineering initiative but a CEO-level responsibility. The technology is akin to electricity just invented—most organizations are still using candles. The key shift is not incremental AI use but reconceiving the entire company's operational fabric. For engineering leaders, this means moving beyond coding assistants to building agentic workflows (like OpenClaw) that act as virtual employees.

Security is a critical enabler. Brex's open-source 'crab trap' proxies agent traffic and uses an LLM judge for real-time policy enforcement, allowing safe agent deployment in production without brittle harnesses. This network-layer approach overcomes traditional security team resistance.

A telling example: KYC process redesign. Instead of automating 20% manual steps, they rebuilt the entire onboarding funnel. Free KYC enabled lead qualification earlier, altering customer targeting. This reframing of problem boundaries is what happens when leaders rethink from scratch rather than tweaking existing processes.

Token consumption should be maxed out—cost sensitivity is shortsighted. Brex tracks token spend with internal tools to attribute cost to products and employees, but the real metric is long-term compounding. The gap between AI-centric companies (within a 10-mile radius of SF/NYC) and the rest is widening rapidly.

Ultimately, the CEO must be the 'chief AI officer' because only they can override organizational antibodies, redesign cross-functional processes, and ensure that AI changes the company's self-identity. The bottleneck isn't technology; it's leadership wisdom to extract unspoken customer signals that models can't capture. The company becomes an architecture of domain-specific agents, not a single AGI.