AI Moves Fast. Brex Moved Faster—By Letting Go of Control

 How Brex Is Surfing the AI Wave by Embracing the Messiness

For many companies, the promise of artificial intelligence has come with a problem: speed. While AI innovation moves at lightning pace, enterprise procurement processes tend to move like molasses. Brex, the corporate credit card startup, found itself stuck in that very trap—trying to evaluate new AI tools using traditional methods designed for a much slower world.

It didn’t work.

In the aftermath of ChatGPT’s breakout, Brex began seeing an influx of AI tools on the market, each promising to revolutionize workflows. But every time a team asked to trial a tool, the internal procurement process would take so long that interest in the tool faded before it ever got approved.

That moment was the wake-up call.

Brex realized that keeping up with AI required rethinking not just what tools it used, but how it evaluated and adopted them in the first place.

A New Playbook for a New Era

Rather than overhaul procurement piece by piece, Brex tore up the script entirely. The company created a fast-track legal and data validation framework specifically for AI tools. That alone helped reduce the approval process from months to days. But perhaps more importantly, they shifted the decision-making power to the people who actually use the tools: the engineers.

Every engineer at Brex now receives a monthly $50 budget to license whatever AI software they believe enhances their workflow—from an approved list, of course. It’s a decentralized model that trades control for agility and insight.

Instead of trying to predict which tools would scale, Brex let the experiment play out in real time.

Why Controlled Chaos Works

According to Brex CTO James Reggio, this approach helped the company separate novelty from utility much faster. If a tool truly drives productivity, engineers continue to use it. If not, it naturally falls off. This ongoing, ground-up feedback loop gives Brex a clearer picture of which tools are worth scaling company-wide—and which aren’t worth renewing.

There’s no expectation that every decision will be right the first time. In fact, Reggio believes that assumption is part of the problem for most companies. The old mindset of “evaluate everything perfectly before making a move” just doesn’t hold up in today’s AI environment. By the time you finish your internal review, the tool you were vetting may already be obsolete.

So Brex leans into the messiness. It encourages trying things, learning quickly, and not being afraid to let go of tools that don’t deliver. As Reggio puts it, the only real mistake is overthinking the process and moving too slowly.

Lessons for the Rest of Us

There’s something quietly radical about Brex’s model. It’s not just faster—it’s more honest. Rather than pretend there's a single "best" tool for everyone, it acknowledges that different teams and roles need different solutions. And instead of forcing consensus before action, it empowers small experiments that lead to clearer decisions.

Brex isn’t chasing perfection. It’s chasing momentum.

And in the world of AI, that might be the only way to stay ahead.