Come gather ’round people
Wherever you roam
And admit that the waters
Around you have grown
And accept it that soon
You’ll be drenched to the bone.
If your time to you
Is worth savin’
Then you better start swimmin’
Or you’ll sink like a stone
For the times they are a-changin’.
Lately the talent world has been buzzing with posts about the imminent demise of recruiting because of developments in AI.
AI is a disruptive technology with a lot of potential, but it’s not about to supplant recruiters.
First, what people label AI is generally machine learning — the ability of certain software products to change and improve at what they do when exposed to new data, without being explicitly programmed. Typically these products focus on a narrowly defined task like image recognition. True AI — the flexible, general-purpose intelligence of the type which allows an individual to learn to complete a vast range of tasks — does not exist.
Recruiting is one of the oldest professions in the world (not to be confused with the oldest), and it’s still about three S’s: sourcing, screening, and selection. The technology that has been developed until now has automated tasks associated with these, without necessarily improving the results significantly. Applicant tracking systems automated compliance previously managed through file folders; job boards brought online a process earlier completed through the mail; and online assessments eliminated the need to fill out a paper form. Fundamentally, there was little difference between the pre- and post-automation era. If a process was flawed, the technology just made the bad results show up sooner.
AI has the potential to incrementally improve all three, with varying degrees of impact. To work well, AI requires vast amounts of data to “train” the software. And data is getting to be more abundant and cheaper all the time as sources increase and people create more.
Sourcing relies on generic data — jobs and people in the labor market. Availability and interest can be estimated with a reasonably high level of accuracy using data on social networks, traffic on job boards and career sites, and non-traditional sources like economic activity. People tend to quit a job when they think they have better prospects elsewhere. The number of voluntary departures increases in fairly predictable volumes when economic indicators like consumer sentiment, stock indices, and the purchasing managers index go up. AI can aggregate data from a large number of sources to identify candidate pools for particular jobs, far better and faster than any human analyst.
AI can also be employed to create candidate pools from new data sources. One such source is biometric data such as the type collected by smartwatches and activity monitors like Fitbit. Biometric data has the potential to predict employee performance and engagement, but it can be combined with data from other sources to identify prospective candidates. Activity monitors, like mobile phones, track location which can be used in creative ways for sourcing. If an activity monitor travels the same distance five days a week along the same road at about the same time, it’s pretty obvious that is the wearer’s commute to work. Using the point of origin — the person’s home, and the end point — where they work, it’s possible to identify a candidate. If they have a long commute, they may be receptive to a job that involves less commuting.
The barrier to doing this is not technical; it is about getting access to the data mainly because of privacy concerns. Given the volume of data being produced by wearables, owned by about 46 million Americans at the end of 2015, it’s a given someone will find a way to sell the data.
The AI that powers Amazon’s home assistant Alexa is now available for use in other applications. The technology allows developers to build conversational, intuitive interfaces for applications and business data. The technology offers text-to-speech recognition service that reads in a natural-sounding voice in 23 languages. The potential for screening is developing career sites that candidates can interact with and have their questions about a job or the company answered. Chatbots using this type of AI can do basic screening.
It’s already possible to use Facebook “likes” to predict personal attributes like sexual orientation, ethnicity, religious and political views, personality traits, intelligence, happiness, use of addictive substances, age, and gender. Predictive models that do this become even more accurate as more data is collected. These models can be used for screening, especially for screening candidates out. Clearly there’s potential for discrimination and abuse, but the point is that AI can have significant impact on screening — good and bad.
Selection is often the most unpredictable part of recruitment. Unstructured interviews, inconsistent use of assessments, and environmental factors can often result in unlikely candidates getting selected. AI has the potential to factor in data from assessments, social media profiles, performance data of previous hires, to rank candidates while factoring out bias. This isn’t exactly a new idea. It has been known since the 1940s that using a regression model to evaluate candidates produces better hires than any interview process, mainly because it uses a consistent approach. But AI can make it much easier to use disparate data sources to do so.
The New Recruiter
So what will be the impact of AI? Sourcing is largely a labor-intensive, low-yield activity and using AI to create candidate pools can only make a recruiter more productive. AI tools can quickly narrow down a large pool of prospects to a small one of those most likely to respond to a solicitation. AI can also be employed to identify the kinds of content that should be used to engage particular candidates, but a person still has to talk with candidates. No matter how natural Alexa’s voice may be, not too many people are likely to get into a conversation with it.
Article Continues Below
Dice’s 2018 Diversity and Inclusion Report
AI can be a very significant time-saver for screening, especially for high-volume hiring. Unlike video interview products available today, products built using AI can adapt their questions based on responses and provide clarifications.
Selection is where AI has the potential to make the most impact, but it’s unlikely that hiring managers will outsource their own judgement to software. AI will, at best, be employed as a decision-support tool — producing a ranked list of candidates with a manager making the final decision. AI cannot read a manager’s mind to figure out what exactly they’re looking for in a candidate. So recruiters’ roles will become mainly about engaging candidates and working with hiring managers.
Thinking more broadly, AI can allow for hiring processes to become more informal while still being compliant. Manufacturing firms can buy a small box from n-Join that collects data from any combination of machines on an assembly line, and then uses AI to spot aberrations that predict a breakdown. The software in the box looks for correlations that indicate a manufacturing process is operating differently than usual and alerts users who can diagnose the problem. The same approach can be used to monitor hiring processes conducted over multiple systems — email, CRM, chat, social networks, etc. to ensure that they are compliant. Wouldn’t that be great?
AI will become increasingly more embedded into recruiting. But there’s no need to panic until the end of recruiting is discovered in the writings of Nostradamus, since he was never wrong. Then it’s all over.