Companies everywhere are racing to keep up with the explosion of AI tools. The challenge? These technologies are evolving much faster than most businesses’ slow, traditional sales and procurement cycles can handle.
Brex, the corporate credit card startup, has been no exception. Like many of its peers, Brex quickly realized that its old way of buying and testing software was too sluggish to keep pace with the AI boom.
At the HumanX AI conference in March, Brex CTO James Reggio shared how the company had to rethink its entire approach. Initially, Brex tried to evaluate new AI products using its standard procurement process, which can take months to complete. But by the time approvals came through, the teams who requested the tools had usually lost interest or moved on to something else.
“In the first year after ChatGPT, when all these tools started popping up, our procurement process would take so long that by the end, no one even cared about the tool anymore,” Reggio explained.
To fix this, Brex overhauled how it brings new software into the company. First, they created a new framework for reviewing data processing agreements and handling legal checks—making it much faster to get tools in front of employees who could try them in real workflows.
The company also introduced what Reggio calls a “superhuman product-market-fit test.” Instead of relying on long pilots and top-down decisions, Brex now lets employees play a bigger role in deciding what tools are worth keeping. The goal is to invest in products that teams genuinely find valuable day to day.
“We go deep with the folks who are getting the most value out of the tool to figure out whether it’s actually unique enough to retain,” Reggio said. “We’re about two years into this new era where we have 1,000 AI tools in the company. And we’ve definitely canceled and not renewed maybe five to 10 larger deployments.”
To empower engineers, Brex gives them a monthly budget of $50 to spend on any approved software they want to try. This decentralized approach helps employees optimize their own workflows without waiting for leadership to pick winners.
“By delegating that spending authority to the individuals who are actually using the tools, they make better decisions for their day-to-day work,” Reggio said. “It’s interesting—we haven’t seen everyone rush to adopt the same exact tool. That’s validated our choice to make it easy to experiment.”
This strategy has also given Brex clearer data about which products deserve company-wide licenses based on real usage instead of guesses.
Looking back, Reggio believes the most important lesson for any company trying to navigate this AI wave is to accept the uncertainty.
“Knowing that you’re not going to always make the right decision out of the gate is paramount to making sure you don’t get left behind,” he said. “The one mistake would be to overthink everything and spend six to nine months evaluating each option carefully before trying it. You don’t know what the world will look like nine months from now.”
Also Read : Y Combinator Alum Launches New $34M Fund Dedicated to YC Startups, Backed by Garry Tan
