Brex is shifting its AI strategy from traditional agent orchestration to a decentralized "Agent Mesh" for autonomous finance, according to James Reggio, the company's CTO. The fintech firm believes this new approach, which emphasizes independent, role-specific agents communicating in plain language, will enable greater automation and flexibility compared to relying on a central coordinator.

Reggio told VentureBeat that Brex aims to "effectively disappear" through total automation powered by AI. The company found that agents operating in narrow, specific roles are more modular, flexible, and auditable for its purposes. The architectural goal is to provide every manager within an enterprise a single point of contact within Brex, handling all their responsibilities, including spend management, travel requests, and spend limit approvals.

The financial services industry has increasingly adopted AI and machine learning. Brex's move reflects a broader trend of generative AI agents evolving from copilots to autonomous systems. However, Brex is betting against the conventional wisdom of centralized agent orchestration, arguing that it can become a constraint.

The Agent Mesh is designed as a network of independent agents with full visibility, allowing them to communicate and collaborate without a rigid, pre-defined workflow. This approach contrasts with traditional orchestration frameworks that rely on a central coordinator to manage agent interactions.

Brex's development of the Agent Mesh follows its earlier work with Brex Assistant, suggesting a progression towards greater autonomy in its AI-powered financial services. The company has not yet released specific metrics on the performance of the Agent Mesh or its impact on operational efficiency. However, Reggio indicated that the company views this as a key step toward achieving its goal of total automation and a more seamless user experience. The company plans to continue developing and refining the Agent Mesh based on user feedback and performance data.