The Executive Trying To Make Microsoft Smarter About AI

A relative newcomer to the company is behind this bet: Chief Technology Officer Kevin Scott.

Scott joined the tech giant seven years ago as part of an acquisition and has bucked Microsoft’s culture to promote a $10 billion partnership with OpenAI.

Scott said his OpenAI-or-bust strategy was the only way the company could have catapulted itself ahead. His approach has rankled some employees, especially within Microsoft’s research division, who found their own AI projects pushed aside and their resources curtailed. Many have left the company out of frustration.

“It’s just been clear as day that you have to pick the things that you think are going to be successful and give those things the resources to be successful every day," Scott, 52, said in an interview.

Microsoft is about to bet its biggest brands on OpenAI’s technology. It is integrating the tech into PowerPoint, Word, Excel and the rest of its best-selling software in the coming months. It will be the first test of whether customers actually want to pay for new AI features that can generate documents and summarize emails. Microsoft has already made the technology the backbone of upgrades in its Bing search engine aand will add it to its Windows operating system as well.

An amiable and cerebral engineer from rural Virginia, Scott has made a career embracing his outsider status. Within Microsoft, an agreement with Chief Executive Satya Nadella makes him one of the few top leaders allowed to work from the San Francisco Bay Area, rather than Microsoft’s Redmond, Wash., headquarters.

He’s been at Microsoft for less time than many of his C-suite colleagues who came up through the company’s ranks. Nadella joined in 1992. Brad Smith, vice chair and president, started the year after. The chief marketing officer started in 1991 and the chief financial officer in 2002.

Scott—who sports a beard that’s more reminiscent of Colonel Sanders than a tech executive—has become the face of Microsoft’s AI strategy. He’s written a book on AI and hosts a monthly podcast where he interviews tech leaders, Microsoft executives, and celebrities like will.i.am and Neil deGrasse Tyson.

Scott is an avid woodworker and chef who brings the same obsessive energy to his hobbies that he does to technology. He has a fully equipped woodworking shop in the Bay Area and cooks with a tinkerer’s dedication to molecular gastronomy. He once struck up a conversation with the chefs at the famed Napa Valley restaurant French Laundry to share his method of flavoring risotto by extracting corn butter with a centrifuge.

Scott, who lives in the Silicon Valley community of Los Gatos, Calif., with his wife and two teenage children, says in the tech world he often feels like an outsider: “I sometimes still don’t think that I quite fit in."

He grew up far from any tech hub. Scott’s home of Gladys, Va.—a small town at the feet of the Blue Ridge Mountains once known for its tobacco plantations—still shaped his view of technology. In his 2020 book, “Reprogramming the American Dream: From Rural America to Silicon Valley―Making AI Serve Us All," Scott described traveling back to his hometown, seeing the long-abandoned fields and factories and imagining how AI could revitalize rural America.

He graduated with an engineering degree at a private Christian college now called the University of Lynchburg. While working for a Ph.D. in computer science at the University of Virginia, Scott dropped out to join Google in 2003.

In 2011 he moved to help run engineering at LinkedIn, where he developed a reputation for his ability to untangle the intricacies of complex systems. A few years after taking the job, he led the effort to rebuild the structures behind the company’s website from the ground up to handle hundreds of millions of users.

He would often sit in the back of meetings as executives debated strategies, recalls Jeff Weiner, then LinkedIn’s CEO. After everyone had opined, Scott would chime in with what many at the company viewed as the definitive perspective.

Scott proved to be particularly sharp when he was asked to determine whether a technical problem was surmountable. If he said it was, Weiner was confident he’d be right.

“His track record with regard to that kind of assessment was essentially 100%," Weiner said.

Scott joined Microsoft after it acquired LinkedIn in 2016. Nadella had taken over as CEO two years earlier and was reimagining the company.

Microsoft had struggled with a morass of side projects and been late to opportunities like the mobile revolution. Nadella wanted do away with the company’s insular culture that favored in-house ideas and sidelined leaders who came in through companies Microsoft acquired. He wanted to bring more of a fast-growth, risk-taking approach and shift away from what he saw as a conservative corporate culture dominated by Redmond lifers.

At Microsoft, Scott’s casual infusion of colorful vernacular stands out, say current and former colleagues. LinkedIn co-founder Reid Hoffman, now a Microsoft director, recalled that a colleague once pulled Scott aside at a corporate event and asked him to refrain from using four-letter words in his presentation; he complied.

Scott typically swears for emphasis, rather than out of anger. “He says ‘That’s a f—ing great idea!’ " Hoffman said.

In 2017, and less than two months after Microsoft’s acquisition of LinkedIn closed, Nadella picked Scott to become the companywide CTO—a position created specifically for him.It gave him oversight of Microsoft’s research division and also created an Office of the CTO, whose members work with groups across the company to help guide their technical strategy.

Scott soon turned to the company’s AI progress. While Microsoft had been building generative AI models for years, he found that it was behind Google, Meta Platforms’Facebook and others.

Around the same time, OpenAI was looking for an investor with deep pockets and a willingness to build a computing infrastructure to train its AI. It also wanted a backer that understood its mission of building technology that can learn, converse and solve problems like a human, a new level of computing called artificial general intelligence, or AGI.

OpenAI CEO Sam Altman approached Nadella at Allen & Co.’s annual conference in Sun Valley, Idaho, in 2018 to invest in OpenAI and the two companies soon decided to work together. Scott and Microsoft CFO Amy Hood worked out the details of the deal.

It gave Microsoft early access to OpenAI’s products and a guarantee that the tech giant would back the startup with its infrastructure. It also pledged to allow OpenAI to continue to pursue its goal of AGI.

Altman, who knew Scott from his Google days, said he wouldn’t have done the deal without him. Scott, he felt, could balance OpenAI’s ambitions and Microsoft’s need to use the technology in profitable products.

“Kevin was in the small handful of people that took AGI seriously," Altman said in an interview.

Scott had to overcome skepticism inside Microsoft, including from then-Executive Chairman Bill Gates, who felt the company already had what it needed to compete on AI, said Hoffman. Scott and Nadella helped bring Gates around, he said.

Gates didn’t respond to requests for comment. But he’s said elsewhere that he’s since come around to OpenAI and the power of its technology.

The investments—$1 billion first in 2019 and a $10 billion round this January—came with consequences. Microsoft had to ration its resources as OpenAI needed ever more computing power to train its massive AI model.

Scott divvied up access for server time on the machines outfitted with the high-end chips needed for AI through what he dubbed “Capacity Councils." Every week, project leaders had to compete for access to computing clusters, a process aimed at cutting down a sprawling array of pet projects and very long-term research initiatives.

The virtual meetings typically included about 20 employees, representatives from different teams making their case in front of executives. They could grow rancorous, according to current and former employees, one of whom likened the meetings to kindergartners fighting over toys. If teams didn’t like the ruling of the councils, they could appeal. Scott made the call when a decision was appealed to the highest level.

Last summer OpenAI first showed off its latest AI model, GPT-4, at a dinner hosted by Gates that included Scott, Nadella and Hoffman. Sufficiently impressed that OpenAI had made a major leap, Scott decided to bet even more of Microsoft’s money and infrastructure on OpenAI. All of Microsoft’s most powerful computers holding its highest-end chips would be dedicated to running OpenAI’s tech. Every other team at the company had to make do with what was left.

While some people weren’t happy with his decisions, Scott said Microsoft needs to focus on projects that can make money.

“We do research, but this is not a research endeavor," Scott said “We are trying to build things that are useful for other people to use."

Write to Tom Dotan at tom.dotan@wsj.com