How Brex is keeping up with AI by embracing the ‘messiness’

As companies grapple with the rapid evolution of AI technology, Brex, a corporate credit card startup, is no exception to the challenges faced in adopting effective AI tools. Acknowledging the lag in traditional sales cycles, Brex has redefined its software procurement approach to stay competitive. During the HumanX AI conference in March, Brex's CTO, James Reggio, shared insights on the company's initial struggles with a lengthy piloting process that stifled their ability to onboard new AI tools. "In the first year following the rise of ChatGPT, our existing procurement strategy was simply too slow. By the time we completed our internal controls, the teams requesting the tools had already lost interest," Reggio explained. This realization prompted a complete overhaul of Brex's procurement process. The company developed a new framework for data processing agreements and legal validations tailored specifically for AI tools. This allowed for a more agile vetting process, enabling faster access for testing and implementation. Reggio described their innovative approach as utilizing a “superhuman product-market-fit test” to determine which tools warrant investment beyond initial pilot trials. This method empowers employees to play a significant role in selecting tools that provide real value. "We collaborate closely with those who derive the most benefit from the tools to assess their uniqueness and longevity," he noted. With around 1,000 AI tools currently utilized in the company, Brex has canceled or opted not to renew five to ten larger deployments, refining their software portfolio in the process. To further enhance decision-making, Brex allocates a monthly budget of $50 to each engineer, allowing them to license any approved software tool they deem beneficial. By decentralizing spending authority, Brex enables engineers to make informed choices that optimize their workflows. Reggio remarked on the intriguing nature of this strategy, noting that they haven’t witnessed a rush towards any specific tool, indicating a diverse range of preferences among employees. This new procurement approach has also informed the company about the need for broader software licensing deals, leading to a more accurate assessment of how many engineers use specific tools. In conclusion, Reggio emphasized the importance of embracing the complexities of the current AI landscape, suggesting that enterprises should accept that navigating the adoption of new tools may be messy. "Understanding that not every decision will be perfect is crucial to ensuring we remain competitive. The risk lies in overthinking and delaying deployment while the technology landscape continues to shift rapidly," he advised.