Kieran Snyder and the trouble with a single word

The adjective was “abrasive.” Kieran Snyder heard an engineering manager use it about a talented woman he wanted to promote. He also wanted to promote a man, whose impatience he treated as an ordinary imperfection. The woman’s personality had become an obstacle. The man’s came with a small indulgence. Anyone who has sat through a promotion discussion knows how much can depend on the temperature of a sentence.

Snyder had a linguist’s reason to linger over that adjective. A performance review turns a passing impression into a document that can travel: to a manager’s manager, to a compensation discussion, to the next decision about who gets a chance. She asked people in technology to share their reviews. In 2014, she assembled 248 of them, contributed by 180 people across 28 companies. An uncomfortable conversation had become a research question.

The resulting pattern was hard to overlook. In that sample, “abrasive” appeared 17 times, describing 13 women, and never appeared in the men’s reviews. The observation had limits: these were voluntarily shared documents, rather than a random sample of every workplace. It nevertheless gave a familiar complaint something useful: a count. Snyder would spend the next decade building a business around the consequences of language.

The professor who changed her application

Her own career almost belonged to a different institution. She expected to teach at a liberal arts college, putting teaching ahead of research. At the University of Pennsylvania, her doctoral work focused on linguistics, natural language processing, and computational linguistics. Six months before finishing, she changed her mind about becoming a professor. The destination had disappeared before the qualification arrived.

She wrote creative nonfiction and linguistics papers, and made extra money playing Scrabble. For someone trying to find another profession, it was a pleasingly literal way to make words pay. Software was not the obvious next move in her own imagination, even though she says she had been coding since she was three. Automated hiring screens were no more imaginative about her background. Her résumé kept failing them.

A person eventually saw the combination the filters missed and hired her. That experience later informed her enthusiasm for people changing careers. At Textio, she described colleagues who moved from research into engineering and from public relations into writing Python. She understood the difference between lacking a conventional résumé and lacking the ability to do the work. Her own first step into industry had depended on someone making that distinction.

A spell-checker learns to look beyond spelling

Microsoft gave Snyder a practical education in getting language technology into people’s hands. She worked on spell-checking and language detection, then on natural language and search. “Software hooked me,” she recalled. Across Microsoft and Amazon, she spent roughly a decade in product and engineering leadership. The academic question of what could be known now shared a desk with a commercial question: what would be useful enough to buy?

She also learned what happens when software must work at an enormous scale. Integrating Bing into Windows meant navigating production requirements far removed from a research paper. At Microsoft she met Jensen Harris, who became her husband and Textio co-founder. Harris brought a background that included music composition as well as interface design. Snyder has described recognizing the connection when he compared creating an interface to conducting an orchestra.

They founded Textio in 2014. Its premise was to help writers understand how language might land before sending it out into the world. Recruiting offered a concrete place to begin. A job advertisement can invite people to imagine themselves at a company, or quietly suggest they belong elsewhere. Textio applied language analysis to those messages, offering guidance while people wrote. The interesting question moved beyond whether a sentence was correct to what it might do.

The distinction matters to Snyder’s story. A spell-checker can rescue a mistyped word while leaving the writer’s assumptions untouched. Hiring language involves a messier transaction between intention, interpretation, and opportunity. Building a product there requires paying attention to the people who receive the words. It also gives a linguist a rather large supply of material. Corporate prose rarely suffers from a shortage of habits.

The annual review gets a second reader

Textio’s work expanded into performance feedback, bringing Snyder back to the kind of document that had provoked her early research. In December 2023, she described an analysis covering more than 13,000 employees at one large international enterprise, across two annual review cycles. That let the team examine written feedback alongside whether employees appeared in the following year’s data.

Employees receiving low-quality feedback were 63 percent more likely to leave than others in the analysis. The researchers controlled for performance ratings and tenure. That is a finding within a particular organizational data set; it should not be read as a forecast for every manager’s team. Its practical implication is still uncomfortable. A review that offers no usable direction can leave even a capable employee guessing about how to advance.

The same work examined who received actionable guidance, and how unevenly it was distributed. Snyder’s attention had widened from conspicuous adjectives to the information a review supplies or withholds. An employee needs to know what happened, what should change, and what progress would look like. A personality label answers remarkably little of that. It can fill a page while leaving the person who must act on it almost empty-handed.

There is also a basketball court

Away from software, Snyder has another place to think about instruction. She began coaching basketball in ninth grade, helping her father coach her younger sister. By 2023 she was coaching fifth- and sixth-grade girls at Seattle’s Lakeside School, with her own daughter helping. The continuity is striking: a teenager helping another teenager, then an executive still showing up to teach children how to compete.

Coaching takes her attention away from the running work commentary she described carrying into other parts of her day. “I am fully present with the girls that I’m coaching,” she said. She also described learning to bring the encouragement she offers children back into her management of adults. The transfer works in both directions. Specific instruction and confidence belong together on a court, and they have a place in a review meeting too.

Her stated hope for those players reaches beyond basketball. She wants them to feel able to compete for something important to them, whether or not they keep playing. It is a useful glimpse of the ambition behind her workplace research. Opportunity requires more than opening a door. People also need information and practice that help them walk through it. A coach has to make that tangible at the next practice.

“I am fully present with the girls that I’m coaching.”Kieran Snyder, 2023

A familiar problem arrives in fluent new sentences

In June 2023, Snyder briefed the Congressional Caucus for AI on people systems and bias. The issue had become harder to ignore as generative tools made workplace writing cheaper and faster. A system can reproduce an existing pattern of unfairness in perfectly serviceable prose. Fluency makes the result easy to accept; it does not establish that the result treats people fairly.

Her writing on AI asks readers to look at the data beneath a tool and the purpose behind its design. Those choices affect what a system notices and what it repeats. For a business buying writing software, that invites a more demanding conversation than a demonstration of how quickly a paragraph appears. What examples shaped the system? What outcome is it designed to improve? Who evaluates the result?

That concern connects the early research to her later work without making every technology equivalent. Counting adjectives in shared reviews and evaluating generated text involve different methods and different limitations. The recurring interest is the human consequence of a sentence that seems routine. Organizations can adopt a tool in a week and spend much longer discovering which assumptions came with it.

The founder hands over the job, then takes another

Snyder stepped down as Textio CEO on January 10, 2024. The board appointed Harris to succeed her, and she became Chief Scientist Emeritus while retaining her board seat. In announcing the transition, she returned to their founding intention: creating better experiences at work and opportunities for people who do not usually receive them. Handing over the operating role left that intention intact.

The next month she launched nerd processor, a home for research, writing, and work with leaders. Her plans included revisiting an earlier study of interruptions, asking how the dynamics might differ in virtual and hybrid meetings. She also coached founders and executives, taught data storytelling, spent time at Operator Collective, and worked on a childhood memoir. The former professor had found several ways to keep teaching.

Eventually, conversations with people building AI products produced a less decorous career signal: jealousy. She wanted their proximity to the work. In her account of returning to an operating role, she described being tired of talking about building and wanting to do it again. By December 2025, her return to Microsoft as VP of AI Transformation was public. Her current biography now lists her as VP of Product, AI.

That biography also lists board roles at Textio, Kit, and Atlas AI. The range makes more sense when viewed through the question that runs through her career: how do people get useful information, and what do they do with it? A hiring screen once missed her potential. Reviews supplied evidence that ordinary descriptions could carry uneven judgments. Now AI can distribute those descriptions at a speed no manager could match.

Snyder’s career has taken her from an abandoned faculty plan to a startup, from a CEO’s desk to a newsletter, and back to a company where she learned to ship software. The question has survived the changes in title. Before a word becomes somebody else’s opportunity, somebody has to look closely at it. She has made a working life out of doing the looking.