Artificial intelligence technology may be rendering miracles in healthcare diagnoses and self-driving cars and trucks. But in a business world striving for cross-cultural diversity, AI is still way to infantile for use in recruitment.

That’s not to say I’m a Luddite. I love technology and I’m supportive of the ways in which it has improved our lives, from decreasing infant mortality to improving telecommunications systems around the world—thereby giving us so much access to information with such great ease of use that it makes life easier for people everywhere.

But when it comes to recruitment, I’m not a fan of artificial intelligence because of that “too infantile” issue. Here’s the problem: As the technology stands today artificial intelligence is, at best, little more than an extremely advanced algorithmic machine for pattern-matching. It works mainly by finding patterns in existing data and replicating them. If you think it through, AI would, by definition, inhibit cross-cultural diversity if used for recruitment.

Oh, you have a Bachelor’s degree? You’ll do because all of our employees have those. Oh, you went to an elite college in the Northeast? Perfect! So did all of our executives! Oh, you’re a member of the Porcellian Club in Harvard? So is our CEO! You’re hired!

In a very simplified way, that’s how artificial intelligence looks at the world. At first glance, that’s perfectly fine. Who wouldn’t want to hire a Harvard graduate? And the Porcellian Club sounds pretty elite, right?

Of course, the algorithmic approach is looking at a resume and comparing it against the backgrounds of successful employees, which is a logical thing to do. They can even look at word choice during an interview and use that to match against words used by top employees.

The problem?

The Porcellian Club is an all-male club. There’s nothing necessarily wrong with that, but if the AI is being trained to look at the CVs of current employees to algorithmically decide if a prospective hire would succeed, then you could (and most likely are) propagating existing hiring biases. In fact, you’re probably making them worse and institutionalizing them because we fool ourselves into thinking that machines are impartial.

That fixation on the past means that AI would ultimately act as a homogenizing force and could hamper diversity efforts for many organizations.

Of course, you may be thinking to yourself, “Well, OK, what’s wrong with hiring for people with traits similar to those who succeeded before?” That actually makes a lot of sense! Well, one of the best hires I ever made was someone with absolutely zero office experience. This individual had some experience working retail and played a team sport at a semi-professional level. The role was to head up a new client-facing service team.

A machine would have chosen someone with years of experience, or the former start-up owner, or the recent Ivy League college graduate. Would the AI have gone for the candidate with next to no experience? Probably not.

Yet she became a fantastic client service leader.

Why? The way I saw it, retail workers have to put up with a lot of attitude from the buying public, so she wouldn’t be easily flustered. And team-sport athletes understand the intrinsic value of teamwork and looking out for one another. What is client service if not a massive exercise in doing what’s best for the team?

She was the underdog in that situation, but I pride myself in choosing underdogs when I’m hiring because they can give you the best surprise—and they usually have something to prove. Also, that sort of hire expands the diversity of the team because it begets looking for talent outside of your established talent pools.

Algorithmic science may one day get to the level where AI can mimic that kind of human intuition, predict diamonds in the rough, and drive cross-cultural workforce diversity. In fact, AI developers are striving to understand how human brains pull off their leaps of intuitive faith and teach their progeny to do it, too.

But until that day comes, I recommend sticking to old-fashioned, flawed, infinitely slower human brains!

Learn More: Request A Demo