While working for his family’s dental supply business, Benjamin Cohen would often travel with company reps on sales calls. He was impressed watching reps build strong customer relationships, but as he learned more about the world of business-to-business (B2B) sales, he saw a problem.

While sales reps were taking customer orders, they weren’t making orders happen.

“In distribution businesses, there are more customer product combinations than stars in the sky, by quite a bit. What that means for reps is that it’s hard to find the right products for every single customer exactly when they need it.” said Cohen, A.B. ’19, an applied math concentrator at the Harvard John A. Paulson School of Engineering and Applied Sciences. “It turns out delivering this type of personalization is something A.I. is excellent at.”

He launched a startup, proton.ai, to apply some of the personalization tools pioneered by e-commerce giants like Amazon to the business-to-business wholesaling market. Proton.ai was named the McKinley Commercial Gold Medal Winner at this year’s i3 Innovation Challenge, sponsored by the Technology and Entrepreneurship Center at Harvard.

The proton.ai software offers six modules to help sales reps work more effectively and efficiently, as well as increase e-commerce revenue. The algorithm-driven software suggests products a customer has never bought, but is likely to buy based on past purchases.

The software also identifies substitute and complementary items that help an online customer ‘complete the shopping cart;’ predicts when a customer is likely to stop buying; identifies when items are ready for re-order; and tracks all customer interactions on- and offline.

Deployment data has shown promising results, Cohen said.

“We’ve increased revenue-per-customer touch over the phone by nine times. We’ve got individual customer service reps who previously didn’t upsell now adding $10,000 per week to orders. We’ve increased the revenue generated by previously used online recommendation engines tenfold,” Cohen said. “At full deployment, we’ll see a 10 percent or more lift in revenue at our $100M+ customers.”

Proton.ai also gets deployed on a customer’s e-commerce platform to increase revenue through up-selling, beating existing recommendation engines by 10 times in A/B tests, Cohen said.

He and his team spent the summer building their product from scratch. They carefully developed and tested algorithms, but spent even more time with sales reps seeking to understand workflows and how their software could help.

The biggest challenges they faced had nothing to do with the technology.

“Our sales rep application only works when reps trust the recommendations enough to actually suggest them to a customer,” Cohen said. “Building trust between our technology and a human on the other side has been the most difficult challenge.”

Taking the SEAS course Startup R&D (ES 95R), taught by Paul Bottino, Executive Director of Innovation Education, has been a huge help to Cohen as he and his team iteratively build a product for their market.

They drew on advice from Bottino and their student peers about how to incorporate, organize their development process, collect feedback, and focus on design thinking.

The most significant lesson Cohen learned is that, for all the wonders of technology, it will never be the only answer to a complex challenge like B2B selling.

“I thought, how hard can it be? We’ll just build an app, deploy it, and we’ll have 10 customers in no time,” he said. “What I didn’t realize is that dealing with people is the biggest challenge. How do we get sales reps to understand, how do we learn from them and build something together? How do we get customers to see the value before we deploy? I didn’t realize how much harder that would be than the actual building.”

Cohen is working to answer those questions as he and his team continue refining their product. They have several customers already, with plans to continue adding customers.

Looking down the road, Cohen is excited for the potential to expand his software into an AI-first enterprise platform that could incorporate business functions like automatically placing inventory orders or forecasting total sales demand.

Cohen finds AI technology exciting, but like the sales reps he has worked with so closely, the human element is what really inspires him.

“The most rewarding part of this journey has been working with our amazing team and being able to work directly with users,” he said. “Seeing how reps work and how much we can help them do their jobs better, for me, is the most exciting part.”