While Boost’s CTO Rinku Sharma doesn’t deny that Artificial Intelligence has been a boon to payments, he cautions that a continuous mix of programming, Artificial Intelligence, and humans-in-the-loop is the best path to design success.
Boost provides B2B payment solutions. It facilitates commercial card payments and serves as a bridge between buyers and suppliers, while also providing card-to-account services for companies that do not accept cards. Boost only accepts the funds flow, not the funds themselves; it structures payment data from many forms into usable information that assists with authorization and settlement.
Sharma has experience with both the business and technology sides. That allows him to liaise between two sides who can speak different languages and have different expectations.
Artificial Intelligence can distort those expectations further. Sharma said agentic development has completely changed the game; folks no longer need extensive training to design solutions.
People thinking they’re developers when they’re not is risky for companies. Product teams wonder why tech teams take so long to validate their creations. If something can be created so quickly, can’t it be validated just as fast?
Not if you want a quality product. Not even close if you want to avoid monitoring, compliance, and explainability issues.
“It’s a key engineering facet that gets overlooked in the AI age,” Sharma said.
He conceded that the days of entry-level designers banging out lines of code are over. That doesn’t mean that designers are headed for extinction. A hot industry topic is how to stay ahead of the bottleneck of design reviews needed to validate Artificial Intelligence-produced work.
“You must still invest a lot of time into the design,” Sharma said. “If you feed the wrong assumptions into the model, especially in development models, you can get a very nice product but it will still miss the mark.”
Sharma sees this driving the need for more senior software developers. Those with intimate knowledge of system architecture are increasingly valuable.
It’s crucial to work with other experienced staff to get their perspectives. That helps avoid the rush to market that is a hallmark of every fintech era, but especially the Artificial Intelligence one.
Did the AI take shortcuts? Did it use compromised libraries that leave the company vulnerable to infiltration? Only detailed reviews will detect that.
Sharma said Boost prides itself on clean data. It first sanitizes the data before tracking the end-to-end data flow across the organization. Its cloud-based system provides a single source of truth.
“That’s the data we feel confident feeding into the model,” Sharma said. “We can then derive more insights from that data.”
Boost’s blend of human and Artificial Intelligence contributions is multi-layered, with both aspects frequently checking the other. Teams create products that are refined by Artificial Intelligence agents. The tech team then reviews it before explaining it to the production side.
When both sides are comfortable, it goes to the agent for completion. Then an independent model conducts a review as a check against system bias. Developers then have the final says
“There’s a high likelihood that AI generates slop and takes you in a different direction,” Sharma said. “That’s why having that human touch and control in different stages of the life cycle is so important.”

