Coding and financial services are two of the most common domains that people cite when I ask about the most promising applications of artificial intelligence. Now we have an example that combines both of them.
Wednesday morning, corporate card issuer Brex announced a series of new “agentic” products, powered by Anthropic’s Claude models, that aim to help finance teams handle employee expense reimbursements and travel expenses. More interesting than the products themselves, though, is the process used to build them.
Brex leaders were initially inspired by the in-house agents Anthropic’s own finance team (which is also a customer of Brex) had cobbled together using Claude models. These included agents that could analyze employees’ expenses to highlight duplicates...
Coding and financial services are two of the most common domains that people cite when I ask about the most promising applications of artificial intelligence. Now we have an example that combines both of them.
Wednesday morning, corporate card issuer Brex announced a series of new “agentic” products, powered by Anthropic’s Claude models, that aim to help finance teams handle employee expense reimbursements and travel expenses. More interesting than the products themselves, though, is the process used to build them.
Brex leaders were initially inspired by the in-house agents Anthropic’s own finance team (which is also a customer of Brex) had cobbled together using Claude models. These included agents that could analyze employees’ expenses to highlight duplicates or out-of-policy expenses; answer employee questions about expenses, benefits and payroll; and track metrics around vendor spending like how long a purchase order takes to get approved and which vendors Anthropic is spending the most money with, said Adam Dix, Anthropic's head of finance operations.