Companies are struggling to adopt the right AI tools as technology evolves at a much faster pace than slower sales cycles.
The same goes for Brex, a corporate credit card company. We found that startups face the same problems as their enterprise counterparts. Results: Brex has completely changed its approach to software procurement to ensure that they are not left behind.
Brex CTO James Reggio spoke to TechCrunch at the Humanx AI conference in March. The company initially sought to evaluate these software tools through its usual procurement strategy. The startup quickly discovered that the months-long pilot process didn’t work.
“When all these new tools came on the ground in the first year of CHATGPT, the process itself actually ran so long, and the teams that wanted to source the tools lost interest in the tools by the time they actually passed through all of the internal controls they needed,” Reggio said.
That’s when Brex realized that he had to completely rethink its procurement process.
The company began by coming up with legal verification to implement new frameworks and AI tools for data processing agreements, Reggio said. This allowed Brex to explore potential AI tools more quickly and get into the hands of testers.
Reggio said the company uses “superhuman product market fit tests” to understand the tools that are worth investing beyond the pilot program. This approach, he added, will have a much bigger role in determining the tools the company should employ based on where employees are finding value.
“We go deeper with those who get the most value from our tools and know if they’re unique enough to actually hold them,” Ledgio said. “Essentially, we’ve been cancelled in this new era with 1,000 AI tools in the company, almost two years later, and we’re not updating with 5-10 big unfoldings.”
Brex offers engineers a monthly budget of $50 to license the software tools they need from an approved list.
“By delegating that spending authority to individuals seeking to leverage this, they make the best decisions to optimize their workflow,” Regio said. “It’s actually really interesting and I’ve never seen convergence. I think it also tested the decision to make it easier to try out different tools. It means that no one has ever seen rushing and say, ‘I want a cursor.’ ”
This approach helped the company know where a wider licensing transaction for the software would be needed.
Overall, Reggio said that the best way for companies to approach the current AI innovation cycle is, in his opinion, to “accept messiness,” and to understand which tools to adopt is a bumpy process, and know that it’s okay.
“Knowing you’re not going to make the right decisions from the gate all the time is the most important thing to ensure you’re not left behind,” Reggio said. “I think one of the mistakes we can make is to spend six or nine months thinking this again and assessing everything very carefully before unfolding.
