Want to Know Why AI Won’t Replace Recruiters?
Chances are good that if you graduate with a degree in psychology, your first job out of college won’t be as an engineer at a renowned technology firm. But then, most people aren’t John Sumser.
Sumser, who will be presenting at the upcoming Recruiting Trends & Talent Tech LIVE! conference, ended up getting an engineering job at Westinghouse after impressing a higher-up there. Since then, he’s worked as an advisor, consultant and board member for a number of HR tech companies, including Salary.com and BrassRing. Today Sumser serves as CEO and principal analyst for HRExaminer, where he’s studied hundreds of HCM companies. He’s also the “Emerging Intelligence” columnist for our sister site, HRExecutive.com.
At the conference, Sumser will be presenting on a topic that he follows closely and writes about often: the intersection of AI and recruiting. As you’ll read in the interview below, Sumser isn’t afraid to take stances outside the mainstream—he believes bias, for example, isn’t necessarily a bad thing. What follows is our conversation, lightly edited for length and clarity.
You’ve been following the HR and talent-acquisition tech space for quite a while. What would you say are the most noteworthy changes you’ve seen?
Probably the first thing that’s worth saying is that there’s been an explosion of investment in recruiting tech, and HR tech in general. Of the dollars that are flowing into the category from venture investors, about 75 percent or so goes into recruiting. So if you’re a recruiting leader, it’s likely your in-box is flooded with solicitations from hundreds of vendors offering tiny little intelligent solutions that address just a small part of your problem.
Can you give me an example?
There’s a vendor that uses AI to take your job descriptions and improve them in a way that balances out the response by gender to your job ads. It’s a really great offering. The trouble is, it’s also just 2 percent of the recruiting process. It’s not the whole thing, it’s just a small string of bulbs on the giant Christmas tree of recruiting. So the question is, what do you do with that—do you just purchase this one-off thing, or do you have a bigger vision, and how do you go about implementing that? This is the heart of the problem for people charged with purchasing recruitment tech: it’s overwhelming. There’s a list of 50 or so different categories of solutions that have recruiting intelligence embedded in them. So, what do you do? That’s the big question.
Some people feel that one of the problems with start-ups in the recruitment-tech space is that many of the founders don’t have a background in recruiting. Do you share that concern?
No, I don’t. When you have hands-on experience with recruiting, the best you can do—if you’re a gadfly and change jobs every two years or so—is that you can see the insides of maybe 20 companies. A vendor can see the insides of thousands of companies, and what they see is that recruiting is very different from company to company. There may be some commonalities, but the processes and the end results are different. Getting recruited by Deloitte and getting recruited by Walmart are two entirely separate experiences and processes. It’s very unlikely that a recruiter will have had experience in both those type of environments. But vendors get the chance to see a broader spectrum.
Now it’s true that recruiting-tech vendors don’t build a one-size-fits-all solution, and if they try to do that, they end up with complaints that their product has a lot of functionality that goes unused. You can get a roomful of recruiters to agree that vendors don’t know what the hell they’re doing. But the truth is, all you’ll get is limited understanding if you just rely on actual recruiting expertise. It’s much better to build products using emerging marketplace tech surveys, and people who truly understand software. Would you rather have a car built by a car designer or someone who drives cars? I’d prefer a car designer. Someone who just drives cars doesn’t inherently understand the mechanics of one.
When it comes to AI and recruiting, what are you most excited about?
What’s really exciting is the focus it’s put on the levels of bias embedded within recruiting. Now, whether you actually want to eliminate bias in recruiting—that’s a big question. To me, the idea of completely eliminating bias from the process is one of the dumbest I’ve heard. Of course, you want to be compliant with regulations about bias. But if you prefer, say, data scientists with a Stanford orientation over those with an MIT orientation, then I think such bias is critical to the nature of your company. Part of the issue is that the word “bias” is being way overused. When people say bias, they often mean illegal discrimination. I don’t know why the word “discrimination” is no longer in vogue, but we should all be interested in stopping that. Bias, however, is inescapable and is often healthy, although there are places where it’s not healthy and you want to zero in on those places. Bias is a hairy category and you want to be very careful in how you talk about it. The reality is, however, that you want an elegant set of biases—you want to be moving your organization closer to a specific point of view without discriminating.
The other good news I’m excited about is that if you’re a recruiting leader who wants to have control over the capabilities in your recruiting system, there’s never been a better time than now. The recruiting function can customize its entire environment with these intelligent recruiting tools. However, you need to know what you want, and that’s probably the single biggest issue with respect to AI in recruiting. If you’re a customer, you need to have a clear picture of where you’re headed and what it will take to get there. It’s a huge mistake to try and test out these tools one at a time. What you need is a goal and some measure of how a particular tool will get you to it.
Look ahead five or so years—any thoughts on how TA might have changed thanks to AI?
What’s going to have to happen, at least in some places, is that people are going to have to figure out how to argue with the machine. This is a tricky problem, because the machine will have all the data. In the classic example, you give the machine 1,000 resumes and it will narrow that down to a list of top 10 candidates. The question is, how do you know it’s right? And, how do you test to determine whether it’s right? This ultimately means that recruiters will have to get more technical and smarter, and the idea that machines will eliminate recruiting jobs is highly optimistic. It’s much more the case that these machines are like little four-year old interns, and you’re going to have to manage them as such. Abstract reasoning is not their strong suite.
Can you share an example?
I’ll relate my own story. I got out of college—way too long ago—with a degree in psychology and couldn’t get a job pumping gas. While trying to figure out what to do, I ended up having this long conversation with a guy who ended up saying “Look, I want to hire you. But one of the guys who works for me has to do the actual hiring.” So he told the guy who worked for him to hire me, and that guy refused. This was Westinghouse, a big engineering firm, and the guy said “He’s totally unqualified for this job; I’d rather quit than hire him.” So they went back and forth, and they ended up hiring me with the caveat that my new boss wouldn’t be held responsible for what ended up happening because of me.
Well, I ended up becoming the fastest-rising engineer within this 1,000-plus group of engineers. Because I had no engineering degree, I was willing to take on the worst jobs they could dish out. I excelled at them. And my boss, the one who initially didn’t want to hire me, remains a close friend to this day.
My point is that this is one of the things that machines, even intelligent ones, can’t do. They can’t see the actual requirements of a job, they can only read the job description. Sometimes the job requirements are so situational that it’s difficult to articulate, but not impossible to intuit. And you need to be capable enough to know when to say, “I’m right about this, and the fact that the data doesn’t support this means there’s a problem with the data.” Never let an algorithm make a decision—the machine should function as an intern or junior employee in the decision-making process.
So, regarding AI and recruiting, are you optimistic about the future?
Yes, I am. I think we’re about to be done with drudgery and will be moving on to really interesting things.
John Sumser will be speaking at Recruiting Trends & Talent Tech LIVE! on Feb. 20.