Choosing a business partner may be one of the highest-stakes decisions an entrepreneur makes, and Daniela Amodei believes the best vetting process starts far from the office. The cofounder and president of Anthropic says prospective founders should travel together before committing to anything—if the trip leaves one of them exhausted rather than energized, the partnership is probably a poor fit.

“Instead of starting a company together, go on vacation together,” Amodei said during a talk at the Stanford Graduate School of Business earlier this year. “Share a room with them. Be like, ‘How did that go?’ And if you’re like, ‘Man, all I wanna do is spend more time with you,’ great.”

“If you’re like, ‘Really, I’m gonna need a vacation to recover from my vacation,’ it might be the wrong choice,” she continued.

Few people understand the value of founder chemistry better than Amodei. In 2021, she launched Anthropic—an AI safety and research company that is one of the most prominent names in artificial intelligence—alongside her brother, CEO Dario Amodei, and five other cofounders. She describes herself as “extremely lucky” with that founding group, but her advice rests on more than fortune: strong interpersonal relationships, she argues, are the real foundation of a durable partnership. The stakes are more than anecdotal: research by Harvard Business School professor Noam Wasserman, drawing on data from thousands of founding teams, found that 65% of high-potential startups fail as a result of conflict among cofounders.

Amodei has known her brother her entire life, and several of the other cofounders have been part of her circle for 15 years. The majority had also already reported to her or to Dario while they were all working at OpenAI. That shared history gave the group an established framework for giving each other feedback—and, importantly, she said they knew “who we were as people.”

The other essential ingredient was consensus on the company’s purpose. Anthropic’s founders were united in wanting to build something of their own; the core group had all previously worked at OpenAI but broke off for various reasons, including differing views on the company’s vision. By the time they launched, the creators of the Claude chatbot were, in Amodei’s words, “in the same zone already.”

She illustrated the point with a thought experiment. “If you locked yourself and your cofounder in another room, and you wrote down or drew a picture of what it is you’re trying to build, you’re not gonna walk out and one has drawn a unicorn and the other has drawn a platypus,” she said. “That’s the type of situation where you think you’re doing the same thing, but I think it just doesn’t end well.”

How to vet cofounders—from stress tests to writing exercises

Amodei is not alone in urging founders to scrutinize their partners. Paul Graham, the cofounder of Silicon Valley startup accelerator Y Combinator—which has backed thousands of startups since its 2005 founding—has long cautioned entrepreneurs to be careful about who they choose, agreeing that good character is immensely important in a partnership because it is tested “severely” when times get tough.

“What people wished they’d paid more attention to when choosing cofounders was character and commitment, not ability. This was particularly true with startups that failed,” Graham wrote in a 2009 essay. “The lesson: Don’t pick cofounders who will flake.”

Andy Dunn, the cofounder and former CEO of Bonobos, went further, laying out five specific tests entrepreneurs should complete with a prospective partner before entering “the intense commitment of starting a company together.” First comes the stress test, practiced through difficult conversations. Next is the time test—spending substantial time together to bond as “business parents.” Partners should also complete the role exercise, ironing out who will do what and how decision-making changes at scale, and the “when things go south” test, which sets up a gameplan for their inevitable departures. Finally, Dunn recommended examining their differences to make sure they are not too alike: “very divergent skill sets” are key.

“You need to spend enough time with each other socially, getting to know each other’s friends and loved ones, and treating it like a marriage,” Dunn wrote for Fortune in 2024.

Written exercises can serve a similar purpose. Before creating CIV, an investment platform that invests in and scales tech companies, VC cofounder Jeff Rosenthal and his business partner, Patrick Maloney, put each other through an exam of sorts. The duo each outlined their company vision in writing—from office schedules to their outlook for the business—exchanging notes “like pen pals” to ensure “nothing was lost in translation.”

“We wanted to make sure that we were aligned in our vision for the type of business we would build, the type of culture we wanted, our expectations of each other,” Rosenthal told Business Insider last year. “And we didn’t wanna leave anything unsaid.”

Across these approaches, the common thread is testing the partnership—its chemistry, its character, and its shared vision—before the day-to-day pressures of running a company put it to the test for real.

A version of this story was published on Fortune.com on May 12, 2026.