Synopsis: A.I. is transforming the job interview—and everything after

Allyson Barr
February 3, 2020

Pieter Schalkwijk, Head of Kraft Heinz’s Talent Acquisition for Europe, the Middle East, and Africa, is responsible for finding the best and brightest additions to his 5,600-person team. For perspective on how daunting this task can be, for around 40 job openings, his team has received 12,000 applications. This means that ~1 in 300 (or .3%) will get the job. The remaining 99.97% of candidates likely don’t hear back, and might have actually been well-suited for the job. Time-constrained recruiters either didn’t get to their resumes before finding their next new hire, or filtered them out based on GPA, past experience, or even based on their first and last name.

Before implementing pymetrics as a first-pass filter, Kraft Heinz recruiters did tend to look for top-tier universities. Now, Schalkwijk says, “it doesn’t matter if you’re from Cambridge.” pymetrics was designed to be mutually-beneficial for candidates and recruiters alike, allowing for a more fun and engaging candidate experience, and for recruiters to receive objective fit-scores to the role candidates apply for. Such scores are determined by the custom-models pymetrics has built with the data of 250 top-performing Kraft Heinz staffers. Their attributes, as represented by how they completed each of the 12 pymetrics games, serve as a benchmark against which all candidates are compared. This data ultimately helps decide job offers, essentially creating a machine-assisted recruiting class. Before implementing pymetrics, about 70% of Kraft Heinz trainee hires had business degrees. Last year, only about half did, and around 40% had engineering degrees. They have been so pleased with early results, Schalkwijk says, that they’re already using pymetrics in some U.S. hiring efforts.

Schalkwijk isn’t the only human resources executive putting into action this more cutting-edge and objective approach. In a 2018 Deloitte survey, 32% of business and technology executives said they were deploying A.I. for “workforce management.” That share is almost certainly higher in 2020—and is continuing to spread to some of the world’s largest companies. pymetrics clients alone include Unilever, Accenture, McDonalds, LinkedIn, Infosys, Hyatt, S&P Global, Mastercard, and more.

Furthermore, AI-driven tools are seeping their way into operations far beyond hiring, such as employee-engagement surveys, the identification of promotion opportunities, etc. Companies are increasingly relying on such tools to save time, and more importantly, mitigate bias in decision-making-- so long as it is done correctly. Frida Polli, CEO and Co-founder of pymetrics, has encouraged us all to proceed with caution, likening AI to teenage sex: “Everyone says they’re doing it, and nobody really knows what it is.” She argues that bad AI actors do exist because at the end of the day, humans, who perpetuated gender, racial, and class disparities, are also feeding the algorithms.

Unfortunately, some systems are trained on biased and homogenous datasets, so the outcomes naturally follow suit. For example, a resume-analysis program was trained on higher volumes of resumes submitted by men, and consequently taught itself that men were always the preferable hires over women. AI used for facial recognition has also come under scrutiny, as it often misidentifies or misreads faces of color. That is likely because, as Frida has asserted, “A machine-learning algorithm is like a toddler; it will learn from its environment...We haven’t had a diverse group at the table creating this technology to date.” So while we should be excited by the promise of such technological advancements, especially as they relate to talent, we also ought to think twice before adopting the next big ‘tech tool of the future’, so to speak. If you’re interested in diving into pymetrics’ uniquely ethical, human-centered design, head over to our Mission Page to learn about how humans and technology should work together to make better hiring decisions.(The article summarized here was first published by Fortune as part of a Special Report on Artificial Intelligence. Read the full article here)