Companies have often found it difficult to integrate the rapidly-evolving artificial intelligence tools into their workflows, as these technologies are developing much faster than traditional corporate software procurement methods can keep pace. Brex, a leading corporate credit card startup, faced challenges similar to those encountered by much larger enterprises. In response, Brex decided to completely revamp its approach to adopting AI-based software tools, ensuring it could remain agile and competitive amid the rapid pace of innovation.
Speaking at an industry event, Brex’s Chief Technology Officer James Reggio described how the company’s original procurement procedure involved a lengthy and meticulous pilot phase, typically lasting several months. However, once innovative AI products surged onto the market following the release of ChatGPT, the startup found this lengthy process inadequate. Teams that had requested specific software tools often lost interest by the time approval processes were finalized, making the traditional procurement pipeline effectively obsolete.
As a result, Brex developed a new, streamlined framework for accelerating its internal legal validations and data processing agreements. This allowed the company to more rapidly explore and evaluate a wide array of emerging AI technologies. To further discriminate among the hundreds of available tools, Brex implemented what Reggio calls a “superhuman product-market-fit test,” which relies heavily on employee feedback. Teams now play a more significant role in determining the value of each tool—guiding acquisition strategies based on their firsthand experience of how effectively these products address their evolving needs and workflows.
Reggio explained that deep engagement with the users who find the greatest immediate value in new software solutions has become central to Brex’s assessment strategy. Two years into this initiative, he noted that out of roughly a thousand AI tools tested internally, Brex opted not to renew approximately five to ten major deployments.
To further empower employee-led software adoption, Brex grants each engineer a $50 monthly allowance to license and test out approved AI tools of their choice. According to Reggio, this tactic decentralizes decision-making, letting individual engineers select and experiment with the tools they find most beneficial to their personal and collective productivity. He remarked how the diversity of chosen applications demonstrated that employees valued flexibility and showed no overwhelming preference for any single, market-leading solution.
This experimentation approach has also helped Brex identify precisely which tools merit company-wide licensing deals, based on solid data about user adoption and satisfaction.
Ultimately, Reggio advised other enterprises grappling with the AI-driven technology landscape to “embrace the messiness” of the process. He reminded industry peers that making imperfect or short-term decisions is inevitable, but that rapidly experimenting and adapting is vital. Overthinking and prolonged analysis could delay implementation beyond usefulness—given how swiftly the AI landscape continues to change, six to nine months of careful deliberation may risk irrelevance by the time decisions are finalized.
