More evolution than revolution, and gains in speed and efficiency rather than accuracy.
It is often said that we tend to overrate the short-term impact of technology but underrate its long-term impact. This mantra is a useful reminder that there’s no need to be overly critical or negative when our expectations of progress surpass actual progress in any field of innovation, including AI.
So, while there is much discussion on how AI “changes everything” when it comes to staffing, hiring, and recruitment practices, it is more accurate to say that its potential for driving fundamental improvements in this area remains significant, but that so far it is still very much a work in progress, at least when we dissect the key areas of impact, namely:
(1) Shifting from credentials to skills: Arguably the biggest opportunity for AI to improve both employers’ ability to find the right candidate for the right role, and increase fairness and meritocracy. Interesting, this is not so much about the how of talent identification (or the specific methods for inferring or assessing talent), but the what (redefining how we think of talent in the first place). Indeed, a logical extension of the fact that generative AI devaluates human expertise, and any hard skill related to IQ, formal knowledge, or objective problem solving capabilities, is that college credentials cease to be a central indicator of talent – and let’s face it, employers have long felt that the value of college credentials is not so much in showing what candidates have learned, but how good they are at learning, but there are much better, faster, and cheaper ways to test whether someone is curious, smart, and conscientious, than waiting for a higher educational institution to select them (or grade them on essay writing for three or four years). And yet, despite much talk about de-emphasizing credentials and college education in many jobs, roles, and organizations, there is still a premium attached to those who complete an undergraduate degree, and extra status to those who have the privilege of doing so in a highly rated academic institution. While there is little sign that universities are evolving and modernizing fast enough to adapt to the AI world, we shouldn’t bet against 200+ year old institutions that have managed to become highly coveted luxury super-brands, to the point of increasing their fees exponentially over the past few decades while offering the same or less value in objective terms. AI may not be able to change this; it’s employer demand (or lack thereof) that will.
(2) De-biasing assessment and increasing the accuracy of search and match: Mixed progress on this critical aspect of recruitment so far. To be sure, the general bar when it comes to making accurate inferences of candidates’ potential remains rather low: hiring managers and recruiters often favor their intuition over facts and data, and traditional hiring methods (from CV-based shortlisting, to reference checks, and the employment interview) still dominate the field despite their low reliability and validity, and propensity to elicit biases. The equivalent in the world of dating (where matchmaking is also a key goal) would be to not rely on online and mobile dating technologies to find your potential partner or perfect love match, but still rely on drunken nights out at the pub. So, it is not hard for the average employer to increase their ability to predict future candidate fit or performance, whether via simple search and match algorithms embedded in their applicant tracking systems, automated resume parsing tools, natural language processing or emotional sensing algorithms that scrape video interviews (they may seem creepy and biased, but not compared to many human interviewers), and science-based assessments that improve the candidate experience by shortening or gamifying the process. But there is a catch: organizations are often unable to measure the exact value employees contribute to their teams and organizations, so the outcome AI predicts is not actually a reliable or objective outcome in the first place. This is frustrating, since we have known for decades how to measure the quality of hire (individual job performance, productivity, turnover, engagement, organizational citizenship, etc). And yet, in highly skilled jobs, a fundamental paradox in management is that the more you get paid for doing your job, the harder it is to know if you are actually any good at it! Instead, your career success and official track record of “performance” is more dependent on your reputation, which depends largely on your ability to play the game, manage politics, and engage in the performative aspects of job performance. This is why if we train AI to predict who will get promoted in a company (which we do by training it on past organizational data), it will successfully identify traits, such as overconfidence, ambition, status-seeking, self-promotion, and low agreeableness as key attributes to select on. In other words, use AI to select your future stars and it will efficiently select politicians, not to mention those who replicate the status quo and the in-group in terms of gender, race, age, and functional background.
(3) Improving the candidate experience: Everybody knows it would be good to improve the candidate experience in recruitment, but there’s limited sign of progress on this over the past decade or more. So, while employers seek to give candidates “consumer-like experiences”, not least because they would benefit from hiring their passionate brand ambassadors or avid consumers in the first place, this is not quite the remit of AI, but of human recruiters and hiring managers. To be sure, AI has improved the process of hiring people on both sides: recruiters are clearly able to boost their efficiency and productivity (though by achieving the same output with less input rather than increasing their total output) and candidates are flooding companies with applications thanks to gen AI tools that enable them to send 100 perfect applications to 100 different employers in a few hours, rather than spending a full week on two or three applications. Indeed, AI hiring AI is no longer a future prospect, but a present reality. Alas, this introduces a great deal of noise and bias in the system, devaluating trust and revaluing old school methods, such as the analogue interview or an informal chat in the cafeteria, not to mention word of mouth recommendations, which lead to nepotistic appointments and harm diversity. What is clear is that the potential value of AI to improve the candidate experience and let even those who don’t succeed with their applications feel positive about a given employer (”I didn’t get the job, but loved the experience”) will not come from AI, but from the human and humane touch that AI can help to enable once it frees up hiring managers and recruiters to harness their EQ, empathy, and people-skills. This is not a given, and requires not just training, re-skilling and up-skilling, but also change management and incentivizes.
In short, though recruitment today does not look radically different than recruitment five years ago, there are visible changes in the horizon. So far, it is a case of mostly running faster in the same direction, but it isn’t clear whether this is the right direction to begin with. And if we don’t know where we are going, any road will take us there.