In 2016, two computer scientists, both Google alumni, bet that artificial intelligence could help make smarter hiring and promotion decisions.
The company they founded, Eightfold AI, grew cautiously in its first few years. But it hit the accelerator after it raised $220 million in 2021, the pandemic boom year for venture investment, bringing the company’s funding to more than $400 million.
The Silicon Valley start-up’s staff increased to 630 at its peak last year from 140 before the pandemic. Revenue grew 100 percent a year.
But things changed abruptly in the second half of last year. Growth slowed and investors demanded belt-tightening. Eightfold laid off about 90 people last month, nearly 15 percent of its work force. And the business is an imminent target for government regulators from New York City to the European Union.
Eightfold’s experience offers insight into the potential and the challenge of applying A.I. to high-stakes decisions like hiring, promoting and charting career paths for workers. The company is at the forefront of using A.I. and data to assess a person’s potential for success in a job. That assessment is based on measuring skills and experience rather than on university degrees or personal connections.
The skills-based perspective has been embraced by labor market and policy experts as a vehicle for broadening opportunity in America, especially for the nearly two-thirds of workers who do not have four-year college degrees. Screening by degrees hits minority workers particularly hard, eliminating 72 percent of Black adults and 79 percent of Latino adults, compared with 58 percent of non-Hispanic white adults.
A.I. optimists say the technology, by harnessing data and clever algorithms, can make hiring fairer and more diverse. Skeptics say the technology is unproven and may just automate past discrimination in new ways.
A New York law, which is scheduled to take effect next month, will require companies using A.I. hiring software to notify job candidates that an automated system is being used and to have the technology checked for bias by independent auditors.
Europe’s Artificial Intelligence Act, which would regulate A.I. tools according to their potential to create harm and which is expected to be enacted this year, has hiring on its list of “high-risk applications.”
Ashutosh Garg, a founder and the chief executive of Eightfold, views the workplace as fertile ground for A.I. As an engineer at Google and then as one of the founders of an e-commerce software company, he spent much of his career building models to predict what a person might buy or view next.
The same concept, Mr. Garg figured, could be used to predict what job a person might be well suited for and career paths, based on the skills an individual possesses.
To do it well would require lots of data on workers, jobs and careers, and powerful A.I. algorithms. Varun Kacholia, a computer scientist who worked at Google and Facebook, agreed, and became a founder and the chief technology officer of Eightfold.
The company began by assembling data, the fuel for A.I. analysis. Today, by pulling information off the web and paying for proprietary data, Eightfold has amassed a data trove of 1.5 million skill sets and 1.5 billion worker profiles, in 20 languages.
Eightfold’s software is typically the matching engine behind a company’s hiring and career website. A candidate uploads a résumé or link to a LinkedIn profile, and the software returns a few roles it has determined to be good matches. A company recruiter can search by job and be shown a list of strong candidates, with the technology often sifting hundreds down to a half-dozen or so.
Eightfold’s A.I. can also show both the candidate and the recruiter skills that a person lacks but might be able to acquire with a short course online.
The Vodafone Group, a British-based multinational telecommunications company, began using Eightfold’s software at the beginning of 2022. Vodafone is seeking to expand its digitally skilled work force and to hire more women. Job candidates are shown a notification that explains how their data will be used, and they have to opt in to be considered.
Last year, Vodafone hired 19 percent more women and reduced the average time to hire candidates by 58 percent, said Marc Starfield, the company’s head of human resources. “For me, skills are the new currency,” Mr. Starfield said. “A.I. is a different and broader mechanism to understand skills.”
Micron, a big computer chip maker, is also emphasizing skills-focused hiring as it expands, encouraged by the billions of dollars in government money pledged to shore up American semiconductor manufacturing. Micron plans to hire up to 9,000 workers for a new factory complex near Syracuse, N.Y., and add 2,000 workers in Boise, Idaho.
The company is using the Eightfold software to broaden the pool of good candidates for its factory and manager jobs. Military veterans, for example, may lack college degrees but have acquired skills like team leadership and how to operate costly, high-tech machines, said Britt Thomas, Micron’s global director for talent and innovation.
Micron also turned on an Eightfold feature called “anonymous résumé,” removing the last name, gender and address of job applicants. The feature is intended to reduce bias in selecting candidates.
The A.I. technology has improved Micron’s hiring practices at times, Ms. Thomas said, calling it “extremely beneficial.” But the company has not systematically measured the technology’s effectiveness.
Forty-two percent of enterprises with 5,000 employees or more use A.I. tools, according to a survey last year by the Society for Human Resource Management.
Nearly two-thirds use the technology to screen out candidates deemed unqualified, while 38 percent deploy A.I. software to match candidates to jobs and rank them. And 46 percent said they would like more information and resources to identify bias when using A.I.
More information, some A.I. experts say, is needed. But so are rules, like clear notifications and independent audits.
The tech vendors, they say, should have to show how their software selects candidates. There should also be a requirement, they say, for empirical evidence, similar to the randomized control trials that determine a drug’s effectiveness.
“Unless you have proof that the tool works, you should not use it,” said Julia Stoyanovich, an associate professor and the director of the Center for Responsible AI at New York University.
Mr. Garg said he resisted layoffs for months, waiting to see if the outlook might improve. “But the market is not recovering,” Mr. Garg said. “And there is a lot of uncertainty.”
Cutting costs to shorten the company’s timeline for profitability, he said, was essential to appease investors and possibly sell shares to the public in a few years. “We have to be aligned with what the public markets want,” Mr. Garg said.
Eightfold, Mr. Garg insisted, still has a bright future, with good customers and plenty of money in the bank. By reducing expenses, the payroll cuts, Mr. Garg said, “bought us time and freedom” to build a business that is both sizable and profitable.
Though more slowly, Eightfold is still growing. New customers include Morgan Stanley, Starbucks, Chevron and Bayer, and more than 100 companies use Eightfold’s software. That, Mr. Garg said, “gives me confidence this technology can work and is working.”
Eightfold has set up an ethics council of outside advisers “to make sure we are putting enough safeguards in place in our company,” Mr. Garg said.
Eightfold, he said, supports regulation like the notification and auditing requirements in the New York City law, as long as it is flexible enough to accommodate fast-changing technology. And while his company’s software winnows candidates, Mr. Garg emphasizes that it is meant to be a helper. A human should decide.
Still, doubts persist. More data alone does not guarantee better results, A.I. experts say. Past bias needs to be scrubbed out. There are real challenges with measuring more subjective attributes like “interpersonal skills,” “presentation skills” and “can work under pressure,” or with measuring what skills and past experience actually determine success in a job.
Great care, they say, needs to be taken in the murky detail work of data collection, data cleaning and data curation. If not, the A.I. will founder.
“The spirit of skills-based hiring is absolutely the right direction,” said Andrea Jones-Rooy, a professor at the N.Y.U. Center for Data Science. “But there are a lot of questions about how to do it well and fairly.”
Source: Read Full Article