James Reggio shares how Brex built a lean, high-impact AI strategy centered on three pillars: enhancing internal workflows (corporate AI), automating critical financial operations like underwriting and fraud detection (operational AI), and delivering AI-powered product value that customers can justify to their boards. Rather than betting on complex reinforcement learning, Brex achieved faster results using SOP-driven agents that break tasks into auditable steps. A small, founder-led AI team developed an internal agent platform enabling multi-agent networks—such as executive assistants coordinating specialist agents for reimbursements or policy checks—while avoiding tool lock-in by letting employees choose their own AI stacks. Rigorous evaluation methods, including multi-turn simulations and LLM-as-judge techniques, ensure reliability. The company fostered AI fluency through a tiered framework and re-interviewed engineers to promote hands-on adoption. Despite rapid growth, engineering headcount remained flat, proving AI’s leverage when paired with sound architecture and cultural alignment.