AI-Enhanced CVs and the Future of Recruiting with Tom Sayer from Accenture

This week's episode was originally broadcast during.

Our recent Social Talent live webinar. Visit socialtalent.com live to make sure you catch our next event. So I'd like to welcome and kick things off I to welcome our first guest, which is Tom Sayer from Accenture. Tom, you're super welcome to the chat. We'll let you just get your camera on and your audio on. And let's get that wonderful hair which I was complimenting you on before we went live onto the show. How are you, Tom? Good to have you here.

Thank you very much. It's great to be here. Thank you very much for inviting me again. Join. I appreciate it.

I'm hoping that although I can't hear you, I've had a problem with my speakers and the first time. And there we are. Go tomorrow I should be able to hear you now. Everyone else could hear you, just not me. Great to have you on the show. I want to kick straight into it. Accenture, which is an organization that hires a huge number of folks every year, a very big organization. You've seen a 30% increase in applications risk recently, which translates to about a million or so additional CVs or resumes. What specific trends or data points suggest that a decent portion of these applications may involve AI generated or AI embellished CVs.

Yeah. Thank you very much, Ronnie. And just I guess to give you a little bit, to give the audience a little bit extra context there as well, in addition to those increases. Yes. So our financial year runs until the end of August. So our financial year ended coming up six months almost to go now. And when we looked at the hiring numbers, which were about the same as they were the year before, before we hired around about 150,000 people, we saw this, this remarkable upturn in applications. Now, I think it's very hard to say definitively exactly what drives that. I don't think it's entirely down to AI. I think there are other things at play, such as the way in which we're the type of requisitions we're hiring, the locations in which we're hiring, those sorts of things as well. But I think there are other indicators that we are starting to see some of that uplift coming from AI for a number of reasons. So I think first of all, we've been looking into this probably for about 18 months now, really ever since ChatGPT came onto the scene and we started to see how it could be used within assessments. More specifically, we've been keeping an eye on it and see how it grows. And last year, part of what I do in terms of my role is I run three teams. One of those is the experience team. And we're interested in understanding the experiences of our business, of our candidates, our recruiters. And we polled our recruiters. So we have around about, you know, four or five thousand recruiters around the world, depending upon our demand. And we asked a set of questions around their experiences in AI. And the survey towards the end of last year showed that 29% of recruiters state that they frequently experience AI enhanced CVs, remarkably similar to the 30% which came up in the poll here. So we know we've got a problem. We know that this is something that our recruiters are starting to come across more and more frequently. And I think once you then marry that up with a lot of what the market is saying in terms of some of the research that's been released across the market, I've seen some reports that suggest up to 50% of applications have been, have been driven through the use of AI job application tools. We've, we've also heard that tools like ChatGPT are making on average up to 14 embellishments per CV. So we know that both industry reports and what our recruiters are saying are telling us that this is one of the reasons why we're starting to see an uptick in application volumes.

What's the impact, Tom, on your team, like, if you were to quantify on that scale what the impact of potentially embellished or perhaps falsified CVs is on the team?

Yeah, well, I mean, as we said, we're, you know, we're a large organization, globally speaking. We have 750,000 employees, we hire 150,000 or so people a year. So an uplift of 30% is significant, as you say, roughly approximates to almost a million additional applications. And what I'd say is that like many other organizations over the last few years, one of our key focuses has been on becoming a more efficient organization. Right. So we spent a lot of time driving efficiencies from a TA perspective further down the pipeline. How do we make our processes, how do we make our technology more efficient? And you do that by looking at how many hours you're going to start saving and so on and so forth. But if you've got this kind of leaky bucket at the top of the funnel that's then just pouring more and more and more into your pipelines, then actually you're losing all of those hard fought gains in terms of efficiency further down the pipeline. So in that regard it is a problem because that does drive more screening. Now we have, in many of the countries around the world, we've now released an AI powered screening assistant which helps to, helps to grade the candidates in terms of their match to the job descriptions. But there is always a human in the loop of that. We're not automated, we're not fully automated in terms of dispositioning candidates on that. So there is a very real impact in terms of those additional applications upon the workload of our, of our recruiting organization.

So it's interesting, you know, Chloe Morrison and. Hey Chloe, it's good to have you on the call. Was wondering about that. You know, surely it's not a problem. They're just AI enhanced and they're still good candidates. So, so what, what's the challenge when you think about the patterns that are there? What are the patterns? And maybe talk to me about the impact of what is the negative impact? What's the negative potential? Because obviously you're not too worried about AI enhanced resumes that represent the truth. So talk to me about when it goes wrong. What is that? What are you looking for? What are those patterns?

Yeah, I don't think there's anything wrong with enhancing your CV if it's helping to bring out skills that, you know, is an honest and accurate representation of your skills. I think where the issue starts to creep in when we're trying to drive efficiencies and one of the statistics that's read is that, you know, utilizing, you know, ChatGPT and using some of these tools which are more specifically designed to help you match your, your CV or your resume to a job description, actually helps to increase your match by between 13 and 20%. Now the problem about that is that whatever you're trying to implement to help drive the automation of that grading, to help your recruiter sift through Those volumes of CVs means that you're going to be looking at an increasingly big slice of pie. But what we're discovering and what we're hearing from our recruiters around the world is that, you know, and I think it was encapsulated a title that was in the press about three or four months ago, is that some of this is leading to a sea of sameness. It's very hard to distinguish between some of the experience that we see with the, the resumes or CVs that are coming through, particularly more at the entry level where it's, what you're seeing is repeating patterns in terms of the way that CVs are formatted, kind of the content on is rather generic and bland and it makes it harder to make that, that judgment call around suitability to a role. Now what we're finding is that that's, that's been, that has been reported to us increasingly frequently as an issue and that when you get down to those screening calls to try and dive into the detail, what often transpires is that some of those skills aren't the depth of which that CV represents. So there is a slight inflation problem and in some instances there are, there are examples of that getting through the cracks onto skills interviews and other things. And of course then that gets onto sort of business time. And what you don't want at this point, of course is business escalation about screening practices and not being able to, not been able to sift through, you know, those candidates effectively.

Can you maybe share with us some of those cases where candidates maybe excelled in the CV or resume presentation but maybe struggled to demonstrate that same level of proficiency? They weren't consistent in an interview or a skill assessment?

Yeah, I think there's a, there's a few different types of example of this. I think, you know, one of the ways in which we are looking to address this is look at the way in which we use digital assessments and how we position those further up the funnel to make sure that we're looking more at a skills match than sort of more of a CV or resume match, which actually is a, is a deeper match and could actually in an organization of our scale, help us to match people to closely aligned requisitions that might just sit in a different slight silo. So we believe that that's one way in which we can mitigate some of the risk around this. As we've started on that journey of looking at how we revolutionize the kind of the suite of digital assessments that we've got. One of the things that we've looked at, of course, is how do we make sure that they're effectively prompted. And you know, those are conversations with assessment vendors that you have and different vendors are going to have different proctoring styles, things like that. But interestingly what we found out is that as we've introduced vendors with better proctoring, and indeed in some instances we didn't have any proctoring, we're finding that those pass rates drop significantly despite the fact that our recruiters are telling us that they're putting through candidates with effectively the same skills from a CV perspective. So that would also suggest that there is a gap between what candidates are telling us they can do and then what's transpiring as they go through a digital assessment. And those drop in pass rates have been significant. So that's something that proctoring has helped us to screen out now. And that's an example of where that's. That, that kind of that, that mismatch has occurred historically.

Ross Taylor is a question that I wanted to also know a little bit about. Are there any trends, for example, increased AI usage for any specific industry or role types that you see in a business or even locations or any trends there when you look at that, because you do such a hiring at such scale that you can share with us?

I don't think there's anything specific that we picked up in kind of the research we've done through our surveys with recruiters and things like that. I think it's fairly broad. What I would say is that there's probably more of a distinction around level, but that might be leading with the fact that a lot of our recruiting is non executive. You look at the volume of our recruiting, it's at a non executive level. There's quite a lot of an entry level. So that's where we see the most instances of it. But. But that's where we do the most hiring. So it's not to say it doesn't happen at executive levels or above, but certainly instances that are reported to us are less of that. But no, I don't think there's any particular skill field or geography where we're starting to see this kind of emerge as a problem more than any other.

We've lots of people asking questions around the software you're using, the tools, is it proprietary? Is it not? Without going into specific tools, maybe. Tom, can I ask you more generally, what kind of tools, processes or trainings are being considered to better prepare the recruiters to handle this change in the way candidates are writing their resumes and applying to jobs?

Yes. So I think there's a number of different things that we're doing. So as I mentioned, increasing the use of digital assessment tools. It's not proprietary. We don't use proprietary tools. So everything we use is commercially available. And I think that the conversations that we're having with a lot of our vendors at the moment are increasingly turning to this topic of AI usage and cheating and proctoring and those sorts of things. So I think that's an important part of the conversations we have with our entire tech ecosystem at the moment. I think, though, that I think we start to get into difficulties Once we look slightly further ahead, because I think that, you know, we've seen some organizations which have gone down the route of, you know, stopping candidates using AI entirely. As I said at the top, right. I don't necessarily think that using it to embellish your cv, so long as it's an accurate reflection of your skills, is necessarily a bad thing or would be considered cheating. But what we really need to think about is what acceptable uses. So within the same poll that we went out to our organization to ask our recruiters about their experiences of AI, 72% of our recruiters said it's a fairly urgent matter, that we need to be thinking about specific guidelines and policies that we can provide both to our recruiting organization and hiring managers and our candidates around acceptable use of AI. And that is something that we are, are developing at the moment, although we've not completed. And I think it raises some difficult questions because I think when we look at the future of work and we look at the way in which, you know, all of the studies have said people who use AI as a co pilot within their roles tend to perform better, tend to work faster, tend to be more creative and so on and so forth. We think that in the future, right, we need to assess people's ability to work with AI as a co pilot. So inevitably this will come up part of an assessment process and we need to think about what the acceptable use is versus what is cheating. Which brings up a bigger question around how does this shift the meaning of cheating? If it does? So I think that really thinking about in terms of those tools and processes, what are we doing higher up the funnel to make sure that we've got appropriate assessments and proctoring in place, how do we make sure we develop a policy? So when recruiters or assessors come across instances of what they believe to be cheating, how do we help them to escalate that to the right people within their teams or within hr, within legal to have those conversations. But I think in, in addition to that, how do we help our candidates navigate this as well? Because I think we need to be out there very clearly on our career sites, within the communications of the applications to talking to them about what we deem to be acceptable use and where they can find out more around support for this? So that's another thing that we're considering. And of course, there are quite a few companies already that do a lot around this, which is great stuff. So I think that that's another consideration we've got to take into account. In addition to that, We've got to communicate the way, of course, which we use AI and be very transparent about the way in which we're using AI to, you know, to analyze their applications and indeed assess them as well. And we do see, we do see an opt out rate to people as they apply to Accenture in terms of their data being processed by AI. So we know we've got work to do there to address that concern.

We're going to move on to that transparency question with Paul next. And but maybe I might ask you one final question which is, you know, more tactical question. Michael has asked this one and I think a lot of people would love to know a little bit more about what you're seeing. But you know, is there a way to know a candidate's using, let's say AI during the interview, you know, let's say a zoom call or a teams call, et cetera. So what kind of feedback I think you said to me, you know, about 10% of recruiters are, sorry, 10% of interviews, according to your recruiters, you know, happen where you suspect or the recruiter suspects maybe AI is being used. What are they seeing? What makes them think that the there might be AI involved, let's say in a live interview.

Yeah, that's right. I mean, to answer the question, I don't think I know of any tool and I'd love to know if anyone does know of any tools which pick up during digital interviews, whether candidates using AI. That's right. Our recruiters have told us that in about 10% of instances they've had suspicions that in calls, in interviews, digital interviews, they suspect that candidates are, you know, cheating using AI. And I think the signs which we hear about most commonly are long pauses between asking questions and answering. Looks like they're reading off a screen. The answers are quite generic, well structured, but generic. And as a result of that, they are assuming, you know, the candidates are using some form of, some form of AI to provide the answers to those interview questions. We've had that from recruiters. We've also had that from interviewers. The business has raised this to us too. So we again, we do know that there is a very real problem and that's where we are developing this policy at the moment to help guide them as to how they start addressing those things. Because it is very difficult to definitively say we've caught you cheating when it's just an assumption. So it is a difficult one.

So Joseph and several others pointing out, and I would agree with this, that you know, a well thought of set of behavioral based interview questions with deep probing usually mitigates against those challenges and you find that it begins to fall apart. So maybe it's less a technology solution, more an enablement solution around great interviewing that can help solve that.

I would absolutely agree with that. And I think that's where people start to uncover the fact that they're talking to someone who's clearly using AI because I think that it doesn't stand up to probing. So I think the ability to do that and I think as to your point joining, it's a point of training rather than technology, certainly in the short term in terms of solving for that.

Tom, thank you so much for sharing those insights. The chat is exploding with more questions, more ideas, more comments. I encourage you to jump in. There's plenty of stuff that you might get from the audience as well. And I'd hope you could maybe just respond to some of those conversations because huge interest in the topics that you brought up there, Tom, but we're out of time. Thank you for joining us. We're going to move on to our next panelist. Again, Tom, always a pleasure. Thanks for sharing those insights.