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

Companies have difficulty adopting the right AI tools as the technology evolves at a much faster pace than their slow sales cycles.

Business credit company Brex is no different. The startup stood for the same problem as its Enterprise opposite hits. The result: Brex has completely changed its approach to software -purchasing to ensure that it is not lagging behind.

During the Humanx AI conference in March, Brex CTO James Reggio WAN told that the company initially tried to assess these software tools through its usual purchasing strategy. The startup quickly discovered that his months of pilot process simply would not work.

“In the first year after Chatgpt, when all these new tools came on scene, the purchasing process itself would actually take so long that the teams asked to obtain a tool that lost interest in the tool by the time we actually passed all the necessary internal controls,” Reggio said.

Then Brex realized that it had to fully reconsider his purchasing process.

The company started to come up with a new framework for data processing agreements and legal validations for setting AI tools, Reggio said. This allowed Brex potential AI tools to check faster and get them faster in the hands of testers.

Reggio said that the company uses a “superhuman product market-fit test” to find out which tools are worth investing outside the pilot program. This approach gives employees a much greater role in determining which tools the company should assume based on where to find value, he added.

“We are going deep with the people who get the most value out of the tool to find out if it is actually unique enough to keep,” Reggio said. “We are actually, I would say, about two years in this new era where there are 1000 AI tools within our company. And we have certainly canceled and not renewed, maybe five to 10 different larger implementations.”

Brex gives his engineers a monthly budget of $ 50 to license which software tools they want from an approved list.

“By delegating that expenditure authority to the people who are going to use this, they make the optimum decisions for optimizing their workflows,” Reggio said. “It is actually very interesting and we have not seen any convergence. I think that has also validated the decision to make it easy to try a lot of different tools, is that we have not seen everyone hurry and say,” I want a cursor. “

This approach helped the company to find out where the wider license deals needs for software also based on a more accurate workforce of how many engineers use something.

In general, Reggio said that the best way for companies to approach the current AI innovation cycle is, in his opinion, to “embrace messiness” and accept that finding out which tools will be a bumpy process and that is fine.

“Knowing that you are not always going to make the right decision from the gate, just like the utmost importance to ensure that you are not lagging behind,” Reggio said. “I think the only mistake we could make is to think about this and spend six to nine months on evaluating everything very carefully before we implement it. And you don’t know what the world will look like like nine months.”