I’ve held hundreds of job interviews. We shouldn’t trust AI with it

Congratulations, you’ve finally landed a job interview. After submitting your application, you’re invited to a phone call with a recruiter as the next stage.

You nervously prepare, rehearsing what you’re going to say in advance. The conversation begins innocently enough as you exchange pleasantries. “I appreciate you taking the time today,” says the interviewer. “So, for our conversation I’d love for you to think about a deal you’ve won in the past 12 months, and walk me through the sales process.”

You launch into a lengthy spiel about a success you’ve had recently. “That sounds like a significant win,” he says. “Can you tell me how you got in touch with that organisation?”

On the surface, this might sound like a regular job interview, except for one big difference, it’s not with a human. The recruiter is one of several AI ‘Interviewer Agents’ created by tech companies to streamline applications, and it’s having big impacts on the humans at the other end.

The hiring process was ripe for disruption, a slow and complex series of tasks with many repetitive steps. First AI came for resumes, and we’re now at the next frontier where AI avatars conduct entire job interviews. In the last few years it’s gone from a niche tool to a common filter for many mid and large-size employers around the world.

At its best, using AI to screen candidates can save time for the business. At its worst, it can add complications to the already emotional toll of looking for work, when you need to impress a robot to move to the next stage.

It took me many years of interviews to realise that I should always choose attitude over aptitude.

If you ever find yourself on the phone, or even a video call, with a robot who’s asking you questions, it’s important to understand exactly what’s happening behind the screen. As you answer each question, the AI system is scoring and ranking you, looking at what you say and how you say it. It’s analysing your words, structure and how clearly you’re able to communicate.

Many tools also track the pace of your words, the diversity of your vocabulary and how often you use filler words, with some even creepily tracking visual clues like your expressions and micro-movements.

All of these are fed into an overall score that ranks you against every candidate who’s gone through the same system, as well as giving the recruiter a breakdown of your skills based on your answers.

I’ve sat through hundreds of job interviews over the years, and all of them – the excellent and the tedious – have helped to refine my ability to decide who would be a good new employee. It took me many years of interviews to realise that I should always choose attitude over aptitude.

But if the candidates with great attitudes cannot articulate their thoughts on the spot, or someone with a disability is automatically marked down by a computer, how is that a better system? Some of my worst hires were smooth talkers, and many of my greatest colleagues would likely have been scored low by AI’s standards. Often the best hires are the unexpected ones.

The companies building this software claim it reduces human bias by standardising questions that are asked, and scoring everyone similarly. But in reality, we’re just replacing one bias with another, as many of the AI models were trained on data that already had embedded racial, gender and other stereotypes baked into it.

So the next time you log onto a job interview, only to be faced with an AI avatar who asks you how your day was, but has never actually experienced one themselves, ask them how their day is going, and marvel at the strange new world of work that we have somehow stumbled in to.

Tim Duggan is author of Work Backwards: The Revolutionary Method to Work Smarter and Live Better. He writes a regular newsletter at timduggan.substack.com.

Expert tips on how to save, invest and make the most of your money delivered to your inbox every Sunday. Sign up for our Real Money newsletter.