Needle in a haystack the term most commonly bandied about by hiring managers when they are looking for qualified candidates. Job seekers often feel like finding a job is searching for the impossible, given how many candidates compete for open positions (500,000 resumes for a handful of jobs at Tesla last year) even in a tight labor market.
So it’s no surprise that a raft of startups using AI have sprung up to reduce the inefficiencies and speed a process that can take more than a month in some cases. Today, another launches out of beta. Called Uncommon, it promises to deliver diverse candidates who may come from non-traditional backgrounds.
Teg Grenager, Uncommon’s CEO, says that the secret is machine learning that was trained on over 50 million resumes and 6 million job descriptions to be able to predict and determine the most qualified candidates. Amir Ashkenazi, Uncommon’s founder and president, maintains that the system is able to make predictions with 95% accuracy–a rate far better than a human combing through resumes with unconscious bias or even an automated applicant tracking system that relies on keywords.
Grenager says they spent a lot of time figuring out how to allow hiring managers to screen candidates to eliminate unconscious bias such as selecting candidates from certain schools. Ultimately, though, Uncommon doesn’t erase people’s names, gender, age, or other qualifying information. Instead, it builds a “merit profile” for each candidate based on previous jobs and skills. Candidates are ranked based on the prediction of how well they’d meet or exceed the job requirements.
This is what makes Uncommon unique. The predictive element is designed to guess the skill level of a candidate based on their previous experience. That means they don’t even have to have a particular skill on their resume. Grenager explains that during the beta period–where companies such as Amazon, Etsy, Lyft, and others took part– humans were reviewing a fraction of the predictions made by the AI. “In particular, the system analyzes the résumés of our applicants (unstructured data) and predicts their level and field of education, years of experience in specific roles, industry experience, hard and soft skills, etc.,” Grenager says, “Then a human checked the results and corrected them if necessary.”
He contends that across all of these types of predictions, Uncommon’s precision measure is about 95%. “It’s very unlikely that our AI system would predict a qualification for a candidate that a human would later disagree with,” he says. Candidates are able to verify or change their profile before they apply, he says. As the dashboard allows hiring managers to select qualifications, Grenager points out, “You see the impact you are having and who you are filtering out before you finalize your decision.” Theoretically, they’ll be able to see if they are setting the bar too high and eliminating non-traditional candidates. “It’s an ecosystem,” says Ashkenazi. One that Uncommon aims to make as equitable as possible.
Correction: An earlier version of this story cited a $5 fee per qualified applicant. The company has since changed their pricing structure to $9.95.