Machines will learn to make sense of myriad data sources to provide insights and recommendations
When a person thinks about artificial intelligence (AI), they might visualize Terminator, the iconic 80’s film starring futuristic sentient robots. More than three decades later, we’re starting to see a world that is both invigorating and frightening as machines become increasingly self-aware. AI has gone from Hollywood screens to many aspects of our lives, including the hiring process where millions of data points processed intelligently are able to give us new perspectives on a candidate than a resume and interview ever could.
Enabling this analysis to happen is the massive number of resumes and job searches flowing online, creating treasure troves of information. About 45 percent of adults have searched for jobs and applied for them online, according to a Pew report from 2015.
Add in unstructured data spread across social profiles, notes and spreadsheets, and employers could ideally have a 360-degree view of a potential hire.
Yet, even for a smart computer, organizing and parsing out this data can be as exasperating as sorting out a 1000-piece puzzle.
Eventually, however, machines will learn how to make sense of myriad data sources, and how to provide insights and recommendations to employers - both about the strength of a candidate and also the true needs of their organization. In this piece, I explore the influx of new data needed to be evaluated; the crucial need for employers to know the qualities in a candidate they’re looking for; and how we’re still at the early beginnings of the machine-driven-hiring revolution.
Mind-boggling data sources
There’s more data on the web than ever before, thanks to our social media profiles, blog posts, and frankly, whatever we post online. As Matt Campbell, Managing Director, Talent + Organization Advisory at KPMG US, puts it: “Everyone’s digitizing their life whether they realize it or not,” and that will allow employers to go deeper into what makes a person tick. It won’t just be about the employee’s skills but how they work and what they do when they’re not working.”
“This will mean moving beyond technical skills, which is where everyone tends to focus today, to what’s an agile person look like, so they can be as responsive as the business needs to be,” said Campbell.
“You’re going to start digging into the cognitive capacity of the person; their emotional intelligence and someone’s learning ability; as well as how quickly they can respond and work with ambiguity. Also their motivation and level of discretionary effort. As that information becomes more readily available, all of a sudden you’ll have big views across all of that content for potential candidates.”
But comprehending someone’s motivations and soft skills - attentiveness, nimbleness or assertiveness -- requires a level of interpretation that some recruiters don’t believe machines have just yet. There’s a level of sophistication that humans bring to the table that a pure AI solution cannot replace, at least not yet.
“You have a balance of quantitative information and qualitative information that go into the hiring process. As much as the numbers are often important and data points are helpful to reference, you want to be able to layer on top of that some quality interview questions and answers,” said Ryan Bulkoski, Partner in Financial Services Practice at Heidrick & Struggles, adding that often new data is learned that was never factored in.
“Roles may change throughout the course of the hiring process, or you may learn from the candidate pool,” Bulkoski explained. “You might say, ‘Wait a second, we actually met someone who has a BA in poli-sci who is very technical, and they taught themselves how to code,’ and if we relied on an AI solution, we might have missed that person,” he said. “You need a little bit of flexibility throughout the process to ensure it’s a good mix of human and machine.”
Defining roles with more precision
French philosopher Voltaire reportedly said, “If you want to converse with me, first define your terms.” In like vein, recruiters note that if companies want to find the right employee, they must first define the terms of engagement. In other words, employers need to take the time to identify the traits and qualifications they’re seeking.
“First, talent acquisition teams need to truly understand what is required for a role they’re recruiting for. Only then can they clearly articulate those skills, capabilities, and indicators of success in an effective way – both internally and externally. Technology products will need to incorporate these success criteria and be customized to the company that’s using it in order for the hiring AI to have a meaningful impact, said Ben Follansbee, Talent Strategist & Consultant at Mercer.
This will require the two sides to work together to figure out exactly what they are looking for. The AI company has to drive the process, Follansbee said, coming up with a number of characteristics of potential hires, then allowing the company doing the hiring to pick out which ones work best.
Defining the role is just the first step in the process. It then has to be refined and massaged as the hiring practice goes forward, and new information comes in. For that reason, human presence is still necessary, said Matt Aiello, partner in the CIO/CTO and Cyber practice at Heidrick & Struggles.
“At the beginning of a search we will undergo pretty rigorous intake sessions where we’re asking the question, what is the client looking for? We will, in the process, usually interview numerous stakeholders, because one person might have one version of the role and someone else might have another, so part of the art and science of what we do is to try to consolidate that and try to point out the differences and to try to come up with a true north. We come up with the five or six areas that we all agree we’re looking for and that’s typically a combination of both qualitative and quantitative factors,” he said.
“The position itself might change, or it might not be that well defined at first. It can actually change over the course of the search, so it’s hard for me to imagine that human judgment wouldn’t be part of any talent search process.”
Human reasoning isn’t error proof. Yet it may seem difficult for some to make dispassionate decisions when it comes to assessing a person’s soft skills. Today, there’s more comfort and trust in our own judgment.
“It’s sometimes difficult to know which behaviors and tendencies will make a candidate successful at the job they’re applying for. And it’s even harder to objectively test for it. But I think that’s what AI can (and should) do,” Follansbee said.
“The downside of this process is that if AI gets it wrong, trust is lost in the technology. And then we’re all worse off. Which is to say that we need to use AI as a tool for now. We need to continue to incorporate neuroscience and psychology into our hiring AI, and be patient for it to rise to the challenge of accurately predicting which candidates will be most successful.”
Most technological innovations rise to the challenge eventually. A decade ago, no one would consider using a phone to take a photo. A generation ago, it would seem absurd and a little scary to rely on an online dating site to find a spouse.
The apprehension to fully embrace AI today is not surprising. At the same time, it’s already part of the process. For example, companies can use a chatbot, which can potentially weed out candidates based on their responses to questions.
“This is an augmenting factor to a recruiter’s job,” said Heidrick & Struggles’ Bulkoski. “It certainly frees up the rest of the internal recruiting team to spend more of their time on less administrative processes, e.g. reviewing raw resumes, and I think there is an opportunity to eventually leverage similar capabilities at the executive search level, but that won’t be diminishing the number of people we have. I think it will only make for a more enriching, data intensive, process that we’ll be able to provide clients with.”
What Bulkoski means is that at this point, machine learning - when technology surfaces up new information or knowledge and/or makes discernments that are not hard-coded - is in a very nascent state when applied towards human capital or recruiting. Machine learning as a field is very robust, having been around for decades, so the discipline itself isn’t nascent. However, the application to talent as a function has immense room for growth.
KPMG’s Campbell agrees that he sees a lot more automation of work processes than decision-making from the business intelligence tools used for hiring.
“At this point, most business are more in analytics and intelligence rather than the actual artificial intelligence. The automation of work, and understanding the results of work, are where most recruiting functions are today,” said Campbell. “Just trying to take out the sheer number of hours associated with finding candidates. I don’t know whether across the market as a whole people are really engaging AI yet, but the big technology platform players who are data-centric organizations, are going down that path because they inherently understand it. If you go to a CPG company, they may not have the same bias or intent to use artificial intelligence at this point.”
Campbell thinks this will change in a couple years as more apps and platforms are built out.
A new world
It’s not hard to imagine a robot making a judgment call on who we are and how well we’ll function at our jobs by looking at a combination of our hard data - school, grades, work experience - and our personalities and predilections. After all, we’ll soon be putting our lives into the virtual hands of autonomous cars that should be able to discern whether a pedestrian waving their hand means, “Thanks for letting me walk across the street” vs “Go ahead, I’ll wait a moment.”
That type of intuition is already being built into machines. In the hiring process, the data to analyze is flooding in and it will require powerful and intelligent machines to digest it all; companies are realizing they need to be more precise about their hiring needs in order to get answers from machines; and already we’re seeing some machines conduct simple tasks, such as administrative matching.
A recent KPMG survey of 900 leading HR executives validates this point as the adoption of intelligent automation was viewed as a "paradigm-shifting force that will reshape the workforce and HR function."
Directionally speaking, whether AI can tell you the success of a new candidate before you hire them isn’t a matter of if but when.