Transcript: Can an ‘AI interviewer’ hire better than a human?

This is an audio transcript of the Working It podcast episode: ‘Can an ‘AI interviewer’ hire better than a human?’

[MUSIC PLAYING]

Ali Ansari
Some engineers say, you know, why am I being interviewed by cartoon character? But in the next version of our products, we’re going to allow the candidates to choose from three characters, one of which will be a realistic human. And we wanna make sure that the interview experience is not only on par with the human technical interview but in fact eventually better than a human technical interview.

Isabel Berwick
Hello and welcome to Working It from the Financial Times. I’m Isabel Berwick. I’m looking at how AI is changing the world of work, not in the distant future, but right now. Last week, we heard how digital twins are helping an executive be in two places at once, and found out that digital assistants of some kind aren’t far off for the rest of us. This week, we’re looking at another aspect of working life where artificial intelligence is saving managers’ time: the hiring process.

AI interviewer audio clip
Hey, I’m your AI interviewer. How are you?

Ali Ansari
I’m good. How are you?

AI interviewer audio clip
Great. Let’s jump into it.

Isabel Berwick
Managers don’t like taking time out of their packed schedules to interview job candidates, and candidates themselves are sick of sending out dozens of applications without hearing anything back. Surely there’s a better way to go about hiring people. A moment ago, you heard two voices. One belonged to an AI interviewer, powered by micro1. The other voice belonged to Ali Ansari, the company’s founder and CEO. Ali reckons AI interviewing systems like his product, gpt-vetting, will entirely replace technical interviews within a year. I wanted to find out more. Ali, hello and welcome to Working It.

Ali Ansari
Hey, Isabel. It’s great to be here.

Isabel Berwick
So could you tell us a little bit about who you are and what you do?

Ali Ansari
Yeah, absolutely. So I’m the founder of micro1. We’re building an AI interviewer to help companies screen software engineers at a large scale. And we’re giving candidates a much better chance at getting their dream role.

Isabel Berwick
Why does the software industry need this? What’s the problem with recruiting now?

Ali Ansari
Engineering managers and, you know, head of HR/talent, they get thousands and thousands of inbound applications. And usually what happens is they have to cut down almost arbitrarily, like take a sample size of those applicants and decide which ones they wanna interview. Usually what happens is that engineering managers waste a lot of time on screening sessions that end up being a waste of time because the candidate is just not the right fit. So what this tool allows engineering managers to do is to screen, you know, autonomously through an AI interviewer before they do a human technical interview with the candidates.

Isabel Berwick
And can you explain to me in fairly simple terms what it would actually look like to be interviewed by this piece of software?

Ali Ansari
Yeah. So you would go on the web application, you would enter your top skills and then you would get two theoretical programming questions per skill that an avatar asks you. And then you ask for those theoretical programming questions verbally. And then after the programming questions, there’s a couple of soft skills questions that you also answer verbally. And then after all that is done, you then do a coding exercise. And then from there we generate a gpt-vetting report, which is basically an assessment of how the candidate did. And then the employers can look at that to decide whether or not they should move on to the next step in their interview process.

Isabel Berwick
So you touch there on soft skills as something that’s being asked. Does the software measure body language, that kind of thing, because I’m hearing a lot about that in AI recruitment now.

Ali Ansari
So we measure body language to determine the probability of cheating. So the proctoring system that we have, you know, measures basically eye movements and a few other things. But in terms of soft skills, the soft skills assessment, we don’t take into account body language. There’s an AI assessment of the soft skills, but there’s also the audio of the actual answers on the report. So the humans not only look at the AI assessment for the soft skills but also listen to the audio, you know, of the candidates answering the questions to determine whether or not the candidate passes the initial screening threshold for the soft skills.

Isabel Berwick
And how widely is gpt-vetting being used at the moment?

Ali Ansari
So in its beta, it has conducted around 13,000 interviews. And we just got it out of beta about two weeks ago. And we’re already, you know, closer to 16,000 total interviews conducted. So, yeah, there’s a lot of companies using it, and mainly it’s being used for engineering right now. But there’s some companies using it for other roles as well.

Isabel Berwick
There’s a bit of a stereotype of engineers being introverted. Do you think this gpt-vetting might benefit people who would get nervous or uncomfortable in a high-pressure, face-to-face interview?

Ali Ansari
Absolutely. And actually, we’ve seen a lot of comments about that, which is candidates basically saying, I’m an introvert and this was really great. I was able to talk without being nervous that I’m talking to another human. And this is the feedback that we get every single day. In fact, tens of such examples come in every single day. And that’s exactly our goal: to really improve the candidate experience to eventually be much better than that of a human technical interview.

Isabel Berwick
What’s been the biggest challenge in rolling out this product?

Ali Ansari
One of the things was actually the sort of stigma towards getting interviewed by an AI system, and we’ve seen that reduce by a lot already. It’s one of those things where people will simply get used to it and they will realise that it’s actually a better way to screen candidates and also gives candidates a much better chance at actually getting the role. There’s some engineers say, you know, why am I being interviewed by a cartoon character? There’s going to be that viewpoint.

But what we’re going to do is essentially in the next version of our products, we’re going to allow the candidates to choose from three characters, one of which will be a realistic human. There will also be the option to not have an avatar experience and just hear the questions and answer the questions, which is what some candidates prefer. So our goal is to really adhere to what the candidates want and we wanna make sure that the interview experience is not only on par with the human technical interview, but in fact eventually better than a human technical interview.

Isabel Berwick
So you’ve said in the future that gpt-vetting will fully replace technical interviews. You know, can you put any sort of timescale on that?

Ali Ansari
Yeah. Our goal is to completely remove technical interviews by end of this year. The way that we see this is there’s going to be one soft skills interview at the very least, probably forever, between two humans. But really, the rounds of technical interviews that happen, that will be replaced entirely by gpt-vetting by the end of this year.

Isabel Berwick
So I’m slightly sceptical that this could replace interviews in, you know, less technical fields like creative industries, for example. Do you think AI will eventually be good enough to pick top candidates for all kinds of jobs?

Ali Ansari
I think, eventually, yes. But our focus and our goal for now is technical interviews. The roles that are much more subjective to assess will take longer, but I believe in the very long run those will also be replaced with AI interviewers.

Isabel Berwick
Ali, is there anything else you’d like to say to people listening who might be struggling to wrap their heads around the idea of an AI interviewer?

Ali Ansari
The feedback that we get online versus the feedback that we get of candidates that actually go through this are wildly different. It is 90 per cent extremely positive candidates saying, wow, this was quite an amazing experience. I was, you know, much less nervous. I was able to articulate my thoughts much better. And I think this is quite an important point of the sort of quick-glance stigma versus the actual product being used and the feedback based on that.

[MUSIC PLAYING]

Isabel Berwick
Thank you so much, Ali.

Ali Ansari
Thank you. Thank you for having me.

Isabel Berwick
Using a cartoon character to interview candidates might seem slightly uncanny. It’s definitely unusual, at least for now. But AI is becoming a big part of the recruitment process in subtler ways.

My next guest, Chano Fernandez, is the co-CEO of Eightfold, a platform that uses AI to help businesses manage their people. Right now, Chano reckons Eightfold does about 50 per cent of the work of managing recruitment and retention. But soon, he says, that number could be 80 per cent. Chano says Eightfold can help recruit candidates, guide employee career paths and even sketch out succession plans before members of staff leave. So I sat down with him to find out more and started by asking for some details on what Eightfold actually does.

Chano Fernandez
We do artificial intelligence around solving business problems, seeing HR, trying to help them to get the best talent, retain and develop. So, for example, creating job descriptions or role profiles or handling the process in terms of managing, scheduling, interviewing or ranking candidates or so on and so forth. So those are some of the areas where, you know, AI is basically increasing efficiency and productivity and helping to manage.

Isabel Berwick
Does it help people hire better or is it just faster?

Chano Fernandez
It should indeed help to, you know, find you the better talent because we are not any longer restricted by the university, the college, the gender or the race, or you know, the pedigree for where you come from. We are much more focused in terms of do you have the skills for the role to perform or don’t you? Right. And of course, machines will be understanding what are the skills that are more relevant. And they will be able to infer not only the skills you have today, but potentially skills that you could be acquiring. So let me give you an example. Someone that, for example, would have, you know, knowledge in terms of programming languages for Java or CPlusPlus, you know, the machine could say, well, this person may have capabilities for learning Python, for example. Right. Because kind of 80 per cent of the journey is done.

Isabel Berwick
Is there anything that humans are better at?

Chano Fernandez
Definitely. Tons of number of things. Right. I would say they’re better at reasoning. They’re better at consciousness. They’re better at judgement. I would say some of those things, we are still better at.

Isabel Berwick
So would you use Eightfold for like an initial process and then bring humans in a later stage. Is that a typical recruitment process?

Chano Fernandez
Yeah. So think about we use Eightflold to understand and say, do you have, I don’t know, 10,000 applications for a particular job? I’m going those one-to-one. It’s kind of hard. So if you use Eightfold kind of to understand basically who are the right matches and again can be internal employees and can be external candidates. But then once that process has passed, definitely you have the face-to-face in person or through Zoom or Teams or whatever. And humans are much better potential and understand how this person will fit with the particular culture we do have here. Is this person a great team player? Do we think that he has the right characteristics? Does he have the right grit or energy or passion? Some of those things are kind of very hard, you know, just to get out of doing parsing on particular résumés. Right.

Isabel Berwick
So your clients will probably be using humans at later stages of the process.

Chano Fernandez
Clients will be using human for a big part of the process. Right. But basically, the significant part of the process, which is the, you know, the interviewing and the managing processes for sure.

Isabel Berwick
Yeah. And I’m interested in using Eightfold for succession planning for managers, because that’s always been a huge issue inside companies. How can it help there?

Chano Fernandez
Yeah, that’s a great question. I mean, usually large companies tend to think about succession management for the top 200 executives. Right. We’re thinking succession management as something across the whole company is very important. You know, can you make it easier for me in providing visibility on what are the opportunities that are within my company, which is an interest for me as an employer as well, because, of course, I want top talent to retain, to stay here. So you wanna have, you know, the potential people inside the company. So we’re helping companies with succession management, basically for all layers of employees. How can I acquire those skills to be eligible for this opportunity in a way to expose different career paths, different journeys that I could have as a professional? What I’m doing today and what could I be doing tomorrow based on what I know, based on what the skills I could acquire?

Isabel Berwick
So that’s interesting. So does it give feedback after they’ve applied? So it’ll tell you this is what you need to get next.

Chano Fernandez
It will give feedback in terms of the skills you might be missing or not. Basically getting this opportunity, right? But this solution will be able or the software will be able to tell you this is why this, you know, candidate was not selected on the initial funnelling then, as you know, will be going more through personal interviews. And then should we people be able to provide that feedback?

Isabel Berwick
So let’s think about three or five years’ time, what’s the AI in terms of recruitment and retention gonna look like?

Chano Fernandez
Yeah, it’s gonna look like a very automatised process. Right. With the machine intervention is doing 80 per cent of the job and the human interaction is doing 20 per cent of the job. Right. That 20 per cent of the job being super, supercritical.

Isabel Berwick
What is it at the moment? The balance at the moment, Human-machine?

Chano Fernandez
Yeah, I would say is more a 50/50 potentially depending on the companies. Right. So that will be increasing significantly. Right. I think it’s very far away where we think like, you know, the whole process is gonna be managed without, you know, human intervention because, you know, I can see as I was, I kind of trust in that process in the next three to five years. Right, potentially rightly so. I wanna see, you know, would I have fun working with Isabel?

Isabel Berwick
Yes, obviously!

Chano Fernandez
You know, I would like to do work with Isabel. Isabel will want to see would you have fun working with me? Right. But that’s good. That’s what is going. Right. We see recruiters already increasing productivity 30, 40 per cent. Right. Today in terms of, again, just screening through resumes and what they can do. So there is a, yeah, there’s a significant efficiency increase play. Right.

Isabel Berwick
Chano Fernandez, thank you so much.

Chano Fernandez
Thank you so much for having me. Thank you.

[MUSIC PLAYING]

Isabel Berwick
Ali and Chano are pretty convincing on how artificial intelligence can make the recruitment process more efficient, but I found it telling that both of them are focusing on software engineering, which relies on highly technical skills that a machine is well placed to evaluate. People are still clearly essential to the recruitment process, which in a way is quite heartening. But it would be a bit shortsighted to ignore all the human shortcomings that AI might help us address, because we can be biased and candidates or employees can get overlooked for illogical reasons. And human error means we often fail to think in a joined-up way. If AI can help with that, even in part, it’ll be a very welcome addition in a lot of places.

[MUSIC PLAYING]

This episode of Working It was produced by Mischa Frankl-Duval and mixed by Simon Panayi. Manuela Saragosa is the executive editor and Cheryl Brumley is the FT’s global head of audio. Thanks for listening.