For better or worse, AI as a term has fully reached buzzword status in HR circles (and far beyond). At this point, most practitioners can regurgitate what sounds like a good grasp on how it all works and what’s to come for the talent acquisition world. If you’re still a little shaky on the specifics though, and could use a breakdown of the fundamentals that would leave you more confident in your understanding, this post is for you. I won’t go deep into the data science, but there’s no reason to dumb down the whole concept either — it’s relatively easy to follow, with a little explanation and the right examples.
So in that spirit, here’s my quick primer on what AI in HR — specifically, in hiring — really means.
Demystifying “AI” and Busting Some Myths
The freak-outs around the idea that AI in the workplace means human-like robots will soon automate us all out of a job have largely died down after some early alarmist hot takes. But, just in case a small part of you is still worried about that: Don’t be. That’s not the goal, and not even a remotely likely outcome.
It helps to understand that when experts and analysts say AI, they’re often more specifically talking about machine learning. To briefly define them both: AI is a broad category of computer science that covers systems that can learn from experience without being explicitly programmed. Machine learning is a subset of AI that uses mathematical algorithms to identify patterns across large sets of data that can be applied to future data to predict an outcome. Even more simply:
- AI means “software that learns and gets ‘smarter’ as it goes, without being told what to do or what to learn”
- Machine learning means “using algorithms to recognize patterns in huge datasets”
- Algorithms are “a set of steps for processing data”
As a June 2018 issue of the McKinsey Quarterly put it, AI creates space and time to think, by filtering signals from the noise (meaning making sense of and finding trends within datasets that are far too large for humans to process — aka “big data”). The article further notes that with AI, you don’t know going in what the algorithms will reveal in the data — it’s always a leap of faith, and an opportunity to find disruptive insights in surprising places.
One other common misconception is that using AI means allowing machines to make decisions for us. This, too, is not the goal, and not the likely outcome — smart software simply provides recommendations, with which you can do what you choose. It doesn’t decide for you; it gives you more information to make better decisions.
If listening to the radio is traditional Hiring, streaming music on Pandora is predictive hiring
You know how when you play, say, the “Bruce Springsteen Radio” station on Pandora, it plays Bruce Springsteen songs, but also a seemingly random mix of others that you weren’t necessarily looking for but end up enjoying nonetheless? It’s not random. It’s based partly on machine learning, specifically predictive analytics.
In essence, Pandora algorithms “learn” which songs other Springsteen fans like by constantly crunching tons of user feedback data — whether a particular song was skipped, whether it got the thumbs-up button, and so on. With that knowledge, Pandora provides recommendations on other songs you may like. You don’t have to bounce around from station to station looking for tunes that fit your taste like you would when listening to traditional radio — you’re automatically served up data-based suggestions. And if you don’t like them, you can skip. Data gives you options, but ultimately you’re in control.
The same revolution is happening in hiring with predictive analytics.
Consider first the traditional hiring approach: You get a requisition, post the job, and begin sourcing. You search online and also go to career fairs and on-campus events. You solicit resumes, get back a huge stack of PDFs, and flip through them — you’re flipping through radio stations here. There are so many, and it’s hard to tell much about the person. So you spend about six seconds on each and look for easily identifiable criteria for filtering: good school, good GPA, past experience at big-name brands. Candidates who meet these mostly arbitrary criteria get interviewed. You subconsciously decide in the first four minutes of the interview whether you like each person, based on whether they remind you of yourself at that age (known as “mini-me hiring”). You spend the next 56 minutes in a state of confirmation bias, confirming the snap judgement from the beginning of the interview.
Yikes. It’s easy to see the breakdown points, and the opportunities for data and predictive analytics to reveal who might be worth bringing in for an interview and who might be best for the role, based on characteristics that actually lead to career success — because often, GPA, school, and even work history are not correlated with performance. Moreover, hiring based on these traditional signals can often lead to bias.
Imagine a Sort of “Pandora for Hiring”
What if instead of that huge stack of resumes, you started with a stack-ranked set of candidates based on likelihood of fit with your organization specifically? That’s predictive hiring at work — algorithms learn the patterns of people who have been successful at your company in the past, and use machine learning to apply that knowledge to candidates, revealing who is most likely to fit and perform well. With predictive hiring, we can drill in and understand candidate strengths and weaknesses and even get recommended interview questions — ones that are most likely to help us determine if this person will be a good addition to the team.
We all use our intuition to hire. Quick example: Say for an entry-level sales role at your company, your colleague Mike tends to hire people who played sports, while you look for high GPAs. The data could show you’re both wrong, and that you should actually be looking for characteristics like tenacity, strong communication skills, and problem-solving aptitude.
I gave a 5-minute talk on predictive hiring not long ago and covered a lot of the above info, as well as a bit on the specific candidate characteristics that modern research has shown are predictive of career success. If that sounds interesting, you can access the video here.
Josh Jarrett is Chief Product Officer and Co-founder of Koru, the leader in predictive hiring based on what really drives performance. Josh has spent his whole career using data to drive business outcomes at organizations as diverse as McKinsey & Company, the National Park Service, and the Bill & Melinda Gates Foundation.More from Josh Jarrett