“Agent coding has significantly reduced the half-life of code. It’s actually very easy for other people to try different technologies for size.” This observation from James Reggio, CTO at Brex, sums up the fundamental challenges and opportunities facing mature fintech companies today. It’s about how to integrate rapidly evolving AI technologies without sacrificing the stability and compliance required of regulated financial institutions. Reggio recently sat down with Latent Space’s Swyx and Alessio Fanelli to detail Brex’s disciplined three-pronged strategy to navigate this transformation, with an emphasis on its in-house agent platform and counterintuitive cultural approach to talent acquisition.
Brex’s strategy is based on three pillars: Corporate AI, Operational AI, and Product AI. Enterprise AI focuses on internal workflows and aims to increase employee productivity 10x across all departments by leveraging off-the-shelf and custom tools. Operational AI targets high-cost, high-risk areas unique to finance, such as KYC, underwriting, fraud detection, and dispute resolution, with the clear goal of reducing operational costs while maintaining strict regulatory standards. Finally, Product AI ensures that Brex remains essential to customers by introducing new capabilities to the boardroom that help them justify Brex as an integral part of their enterprise AI strategy.
The most profound changes have occurred in the operational realm, where Brex has discovered that simplicity trumps complexity. Rather than pursuing over-engineered reinforcement learning models, the company prioritized building agents based on standard operating procedures (SOPs). “We found that SOP-driven agents beat over-engineered reinforcement learning in financial operations,” Reggio said, highlighting that breaking down complex workflows like KYC and underwriting into auditable and repeatable steps enables rapid automation and ensures the necessary compliance. This approach ensures that output remains fully explainable and accountable, a non-negotiable requirement in fintech, even as automation increases. These operational agents, such as KYC agents and underwriting agents, run on the Brex Agent Platform, a centralized internal structure that serves as the “secret fourth pillar” that enables scalability and consistency across the organization.
This base platform includes an LLM gateway to route requests, a prompt manager to manage SOPs, a centralized knowledge base for deep business understanding, and an evaluation framework for rigorous testing. This internal tool acts as a force multiplier, insulating core application logic from the rapidly changing landscape of larger language models and frameworks. By providing this layer of abstraction, Brex allows engineers to focus on developing specialized agents that deliver tangible business value through internal cost savings and improved customer experience, rather than getting bogged down in maintaining bespoke integrations.
The most obvious example of the Product AI pillar is Brex Assistant. This is a feature designed to replace traditional, cumbersome dashboards with an Executive Assistant (EA) model. Reggio explained that the end goal for employees using Brex is for the product to “completely disappear,” leaving only the corporate card itself and automated assistance to handle expense documentation, travel reservations, and procurement policies. Rather than a monolithic AI, this assistant is an orchestrator in a multi-agent network that coordinates specialized subagents such as audit agents, procurement agents, and reimbursement agents to complete multi-turn conversations and tasks. This decentralized, hierarchical structure allows Brex to maintain professionalism and accuracy across diverse financial workflows while providing a unified and intuitive user interface.
A key element of this strategy is Brex’s unique approach to talent, which is distilled into a “quiver-welcome” philosophy. Recognizing that great builders often have a founder’s mindset, Brex intentionally seeks out people who have started or are planning to start a company. This approach is attractive to talented builders. That’s because Brex offers “an interesting problem to solve, but instant delivery” to instantly deploy new financial AI applications to more than 40,000 companies. Reggio explained that they like to hire people with a strong sense of agency and product acumen, giving them the resources to solve tough problems and build solutions that have immediate, large-scale impact. This cultural decision ensures that the AI team (which Reggio describes as a small, close-knit group of about 10 people, a mix of young AI-native engineers and experienced staff) is lean, ambitious, and focused on driving production-grade agents.
Additionally, the focus on agent tools has subtly changed the role of engineers. As AI tools handle more routine and repetitive coding tasks, value metrics move away from vanity metrics such as the percentage of code generated by AI. Instead, the focus is increasingly on second-order effects such as maintaining a healthy codebase, managing the “slop” and “drift” introduced by generative models, and maintaining deep code ownership. Reggio emphasized the importance of moving beyond traditional metrics, noting that the role of an engineer is now similar to that of a supervisor or mentor who guides and structures the work that AI agents do. This paradigm elevates the human role to one of architectural design and strategic oversight, ensuring that human attention, an ultimately scarce resource, is directed to the most important business and technical challenges.
The discussion highlighted that for Brex, AI transformation is not just about adopting new tools, but about fundamentally re-engineering its operations and product philosophy around agent systems, ensuring that speed and innovation remain paramount, even in a high-stakes, regulated environment.
