Could AI Be Screening Out the Talent You Need?

What if the biggest risk in AI-driven hiring is not what it helps you find — but what it filters out before a human ever gets involved?

In this episode, Gerry Murray talks to investigative journalist and author Hilke Schellmann about what happens when black-box AI tools start shaping high-stakes workplace decisions.

Hilke’s book, The Algorithm, exposed how AI was already being used to assess, rank, and filter people in hiring and employment. In this conversation, she shares some of the strange, troubling, and very human stories behind that work — and why leaders, HR teams, and hiring managers may need to ask much harder questions before trusting these systems.

Together, Gerry and Hilke explore bias, false objectivity, weak oversight, and the growing gap between the promises made about AI in hiring and the reality many candidates may actually face.

If talent decisions matter in your world, this episode is well worth your time.

Or, if you’re a job seeker, you might be alarmed at what’s really going on.

Enjoyed this episode? The Leading People Community is where these conversations continue to flourish. 

Free to join — Book reviews, Leading People Café recordings, Podcast episode previews, and a growing group of active HR and L&D professionals, leaders, and coaches who take their work seriously. 

Join the Leading People Community

Connect with Gerry on LinkedIn

Transcript

Why AI In Hiring Matters

SPEAKER_00 0:00

Welcome to Leading People with me, Jerry Murray. This is the podcast for HR leaders, LD professionals, and senior leaders who want to bring out the best in themselves and the people around them. Every other week I sit down with leading authors, researchers, and practitioners for deep, honest conversations about what great leadership actually looks like in practice. And how your own mindset, behavior, and presence shape everything around you. If you want thinking that challenges you, tools you can use, and conversations worth returning to, you're in the right place.

SPEAKER_01 0:47

So I thought I'd test this out. So, you know, I got a login from a company, set this all up, and then I thought, like, well, if I speak German, obviously it's gonna sound like gibberish. I didn't want to fake an accent or a speech disability. Um, I'm gonna get this error message, right? Because I'm not gonna get over the threshold, and a human won't be in touch with me. So I thought I'll just um read the German entry in Wikipedia on uh psychometrics. What was surprised me, although the transcription was gibberish, I still got a 73% qualified for the role. It's pretty high. Um so I didn't get an error message, and I was like, oh, this really like strikes me as weird. Like, how can the tool say I'm so highly qualified if I didn't actually say anything of substance?

SPEAKER_00 1:34

That's this week's guest, Hilke Shellman, Emmy Award-winning investigative reporter, NYU journalism professor, and author of the algorithm. At the moment, it feels like you can hardly turn right or left without AI coming into the conversation. For some, it's a source of optimism and possibility, and for others, it raises serious concerns about fairness, risk, and potential harm. In this leading people conversation, we explore what happens when black box AI tools start shaping high-stakes workplace decisions, and why leaders, HR teams, and hiring managers may need to look more closely at what these systems are really doing. If talent decisions matter in your world, this one should get you thinking. So let's hear what Hilke has to say.

Hilke’s Path Into Investigations

SPEAKER_00 2:34

Hilke Shellman, welcome to leading people.

SPEAKER_01 2:37

Thank you for having me.

SPEAKER_00 2:40

So, Hilke, uh, a few years back, uh, I discovered your book, The Algorithm, and I was fascinated by what your research revealed at the time about hey, how about hey how AI was being used to hire, monitor, promote, and even fire people. Um and I'm sure the listeners out there are going to be fascinated by the story behind the book and what you found out. But before we get into the book uh and the evolution of AI and the algorithm, please share with my listeners the journey that has led you to being an Emmy Award-winning investigative reporter and an associate professor of journalism at NYU. And were there any aha or epiphany moments along the way as your career evolved?

SPEAKER_01 3:25

Um, I mean, there are always sort of interesting moments, right, that change um how you see your work, the trajectory of your work. Um, you know, before I came to cover um AI in 2017-2018, I was already an investigative reporter, right? I I uh looked for systemic problems in the world and to report on them. And the work has gotten me to um, you know, document uh surrogacy when you pay a woman to have a baby for you in the United States. There are loads of problems with that. I went to Pakistan to uh report on an alleged gang rape um and what happens in a culture uh that tries to silence women. Um that actually did get me um uh the ME win. Um and um yeah, and so I was already at at NYU I'd found a couple years earlier um sort of this like deep satisfaction of teaching others. Um, and actually it made me also a better reporter and a better filmmaker, which is kind of a funny thing when you learn, you actually think through why you're holding the camera. You kind of learn a lot or why are you doing this interview this way? Um, and then uh one day I was at a conference in um uh Washington, DC in late 2017. It was a conference uh with lawyers about consumer law, it had nothing to do with AI. And I needed a ride uh back to the train station from the conference venue. I called myself a lift, which is sort of an Uber, um, because you know I was in a different city, didn't have a car, and I got in the backseat and asked the driver how he was doing. And he said, Oh, I'm having a weird day. And that has never happened to me that a driver said, Oh, I'm having a weird day. And you know, I'm sort of curious, and I was like, Oh yeah, what happened? And he told me that he had just had an interview, a job interview with a robot. And I was like, job interview with a robot, never heard of this. Um, I think it turns out that this was probably um a phone call with a pre-recorded voice, but to him it felt like a robot was interviewing him. And um, so we chatted about that. I got his phone number because I was like, oh, you know, I happen to be a reporter. I would love to look into this. And he's like, Yeah, totally, call me up. And uh, you know, I went on the on the train, took a note, and then kind of forgot about it until a few months later I was at an AI conference because my colleague was presenting there. And there was a session in the afternoon led by um Kelly Trendell, who had just left the Equal Employment Opportunity Commission. And she was saying that she couldn't sleep at night because companies use um, you know, this in 2017 use like these basic algorithms to go through their employees' calendars to check how many hours they're at work. And she was really worried that uh, you know, this could uh lead to unfair decision making, especially towards uh caregivers, which are mainly women or people with disabilities, who just have longer, often longer times that they're not at their desk. Um so we started chatting, and she mentioned that there's this uh very big company that does these one-way video interviews that I got really into it. Talk to the chief data science, former chief data science person the next day, and I maybe understood 2% of what this man was saying, basically his name. Like I was just like, I was blown away. I've never heard of this kind of technology being used on humans. And and uh he suggested I come to a conference, and I went to that conference, I was just literally blown away, and I walked into the room and they were showing how higher view is a company's called Higher View, does these one-way video interviews with AI, and you know, you were seeing these like um uh squares over people's eyes and tracking their facial expressions and and inferring people's um emotions underneath them, and they were talking about how they attract people's intonation of their voice, the words that they say. And I was just like, you know, maybe I was also a little bit naive at the time. I was just like, wow, I didn't know that that was that your facial expression in a job interview are predictive of how successful you'd be at the job. I was like, who knew? Like, so interesting. So I pitched the story to the Wall Street Journal. I was an executive producer and reporter on a 10-part investigative series, video series at the time. And they thought it was super interesting. I, you know, do what every investigative reporter probably does routinely. I looked in the archives, like what has been done about this. There's really nothing at the time. So I was like, hmm, either there's nothing to report here or we just don't know about it. But I was, you know, I had been to this conference, and this is early 2018. There was already so much AI in this space, and I was like, wow, I don't think people are aware of this. How much? And in fact, like every time I've talked to people who've done one-way video interviews, they thought a human watches the video. That has changed over the years, but at the beginning, they thought for sure some human has to watch all these videos. And I was like, well, that could be true, but some companies we know for sure use this AI product. Um, and then when I started talking to, you know, we went to the company, we filmed there, we did all these tests, and and uh, you know, as an investigative uh reporter, I do like trust but verify. So I was like, okay, this is so interesting. They do this emotion recognition and intonation of voice analysis. I'll talk to some experts who study emotions, who study this kind of um

The Rise Of Robot Interviews

SPEAKER_01 8:40

uh these kinds of algorithm. And I was surprised when they said, there is no science here. We don't have any science, but facial expressions in a job interview infer you will be successful in the job. Um, in fact, what this is based on is like um people who are deemed successful who have done the job interview maybe a year before, the same one, the the um computer model will sort of build an algorithm around what these people have in common, how you know what facial expressions they have, the successful people versus the non-successful people in these job interviews. And then it will infer and compare the incoming people into that and infer if they will be successful or not. Um, yes, a computer can do this and probably really well. The unfortunate thing is that there is no research and no basis on doing this. Yes, we can do this techno technological assessment, but it doesn't mean that you are smiling. You know, I'm sort of generalizing here, but if you're smiling when you when you answer what are your strengths and weaknesses, and it happened to be that other people smiled there, that that is actually um that's what makes you successful. It's just like a random statistical overlap. So that was a pretty decisive moment when I was like, wait a second, like this company is already using this on millions of Americans who have no clue what is happening, and there's like really doubtful science. Um, so um we published this 10-minute investigative video in the Wall Street Journal, and it like, I mean, I've never seen a story of mine explode like that. It was just like the most watched video on the website. It got hundreds and hundreds of shares and comments on LinkedIn. I got to know everyone on the LinkedIn news team because they picked it up, right? And they were really excited about this. Obviously, this is a place where people talk about this. So I knew this was a big deal, and so I started looking, you know, in the research for the first video. I had talked to many other companies and I sort of had found other technologies that it was like, oh, I don't think people know that AI is being used here, and we don't know a whole lot about this because a lot of these tools are sort of black boxes, um, right? And like often the vendors themselves don't know exactly how the algorithms work. They sell it to the companies who then use it to make high-stakes decisions. Um, and you know, I think I felt like, well, this warrants a closer look. And it seems to be like no one else is around to take this closer look, so it might as well be me. Um, so I started digging deeper and digging deeper, and that then eventually led to the book and and um all the other work that I've done around on this.

SPEAKER_00 11:12

Right. So um you you you're actually it's that's great because you're partway into the first question I was going to ask you around what surprised you or disturbed you about the way these algorithmic systems were already shaping uh you know, working people's lives. But you one of the things in the book, uh because it's a while since I read it now, I I read it avidly when it came out, but you interviewed a lot of people and tested out a lot of stuff, and you give some profound and even absurd examples in the book of how they were being used. I mean, I mean, I think I'm thinking when you're talking about the face stuff about the phrenology stuff that has no scientific basis, for example. But you had situations even with the German language, didn't you? You had situations with the language and all these things. So maybe you could share a few of those stories because they're they're fascinating.

SPEAKER_01 11:59

Yeah. Um, so I should say, like, I'm a person who has no computer science background, right? I don't have a computer science degree. I picked up some random knowledge about AI and computer science through through throughout the years. Um, and I'm trying to say this because I feel like if I can hold these tools accountable, everyone else can too, right? Like I don't have a distinguished technical background on this. But I was wondering um, you know, these these one-way video interviews that were um already um pretty widespread in the United States. Most, a lot of Fortune 500 companies were using them, especially companies that hire in retail, in fast food, in like places uh they call this um high tone or high high volume um uh work where you have to like hire constantly people, right? Like a big retailer has to hire constantly people. So they have to keep this pipeline going. Um, so um they they were using these on probably millions of Americans who who didn't know about this. So, you know, very intriguing that this is happening

Testing AI With Bogus Answers

SPEAKER_01 12:55

and um we should we should really look into this. So one of the things I was grappling with is like, oh, okay, so the tool um analyzes tone of voice, the words that I use in my facial expressions. And some tools only did the intonation of voice and words. But I did wonder like, what is with people like me who have an accent? What is with people who have a speech disability? Because the tool doesn't actually take the recording itself, it runs it first through a transcription service, which we use all the time in our voicemail, right? They're it's very ubiquitous. Um, but I wonder what is what is with people who have a speech disability or an accent. So I talked to the companies and they're like, oh yeah, these algorithms, we have it figured out. It's like totally accounted for this. And I was like, wow, I've talked to computer scientists and they think there's a problem here, but I these companies have figured it out. So I asked, they're like, yeah, there's like a specific threshold you have to overcome. I don't really know what the threshold is, but if you're under a certain threshold, uh, a human will get in touch with you because maybe you didn't speak at all or you had an accent that the computer didn't couldn't decipher. So I thought I'd test this out. Um, so you know, I got a login from a company, set this all up, and then I thought, like, wow, if I speak German, obviously it's gonna sound like gibberish. I didn't want to fake an accent or speech disability. Um, I'm gonna get this error message, right? Because I'm not gonna get over the threshold and a human will be in touch with me. So I thought um I'll just um read the German entry in Wikipedia on uh psychometrics, which is sort of the science of measuring um people's differences. Um, and I actually got a transcription, which was gibberish, makes no sense. It was like sort of weird words, right? Because, you know, the tool obviously was programmed to um, you know, uh inferring on English sounds and to try to make sense of my German sounds. Uh what surprised me, although the transcription was gibberish, I still got a 73% qualified for the role. Uh pretty high. Um so I didn't get an error message, and I was like, oh, this really like strikes me as weird. Like, how can the tool say I'm so highly qualified if I didn't actually say anything of substance? That is worrisome. I did this with like my other graduate students who spoke different languages at the time. They also had similar results. Uh we did this like all different different ways. Uh, we did it with silence that actually did trigger an error message. I also one time said just um I love teamwork 50,000 times. I also got a very high score, um, even though I did not say nothing else, but I love teamwork. Um so it really like makes me question like, what do these tools actually do? Obviously, AI isn't a thinking machine, but these tools should do better and sort of like um obviously have a much larger context window. And somebody also needs to monitor these. Like, how is it possible that I slip through with these bogus tests? Um, and then this spurred like a whole array of academic research where I work with computer scientists, a much larger um uh uh groups, and we did find that people, for example, with um accents who have an accent or speech disability, um uh in our case, one specific speech disability are um uh are unfairly treated by these algorithms. You know, I do have to say overall the technology is getting better and the data, the training sets um are presumably more diverse, but we still had a problem when we tested this um a couple of years ago. So it's not it's not totally gone. To me, this means this should give us pause, right? Because the problem is that like um these have black boxes and they make high-stakes decisions on humans. And that's what that that's what really worries me here. That like um, you know, this is not just some random spam filter and it works or it doesn't, but uh the losses are negligible, right? And if your spam filter doesn't work, your AI doesn't work in there, you'll find another one. But this is like high-stakes decisions that like if I get a job or not, like my often my happiness, my identity depends on it. And like, even more basic, my food, like feeding my children, the roof over my head depends on it, right? So, like these are we we need to make sure this is like fair. And uh, I'm not convinced that all of these tools that are out there in the market make fair decisions. And based on the research that I've done, we found out that they uh uh have make uh some some of the tools do make um racially biased and gender-biased decisions. Um, and on top of that, some of them just don't even seem to work at all as as advertised, and that all worries me.

SPEAKER_00 17:33

We'll be right back after this short break.

Subscribe And Join The Community

SPEAKER_00 17:39

If you're finding value in this conversation, please subscribe wherever you listen to your podcasts, and perhaps share this episode with a colleague who'd find it useful. Word of mouth is how good podcasts grow. And if you'd like to be part of the conversation beyond the podcast, I'd love to have you in the Leading People community, a growing group of HR leaders, LD professionals, and experienced leaders who gather regularly for live sessions, discussions, and the kind of thinking you've been hearing today. You'll find the details in the show notes. So one of the things that I, if I'm correct in remembering,

When Great Candidates Get Filtered Out

SPEAKER_00 18:25

there was a lady software person who had applied and was really super qualified and applied and applied and applied. And one was it was that the lady where um she hadn't she hadn't got an interest in was it baseball or something on her curriculum DK or a resume? Or was that a different story?

SPEAKER_01 18:45

And this this was something that was uh but I can I can I can talk about that. There was um someone I found, and we gave her the name Sally because she was really afraid that she's gonna um face repercussions when she speaks about like big employers that um you know didn't didn't give her the benefit of the doubt. Um what was interesting with her was that um you know she's uh she's a military veteran, she had been employed, sorry, she had been deployed um a couple times in war zones. She's African American, she's a woman, and she had two degrees in uh computer science, a bachelor and a newly minted master's degree. Um, so I mean you sort of felt like, okay, like people would bend over backwards to hire uh this this woman into their um uh you know the developer team. And what was interesting, she was very thorough. She had an Excel sheet, which she shared with me with like all of the applications and tracted, and she never got any traction applying. Um, she only got traction and got interviews and job offers um when she personalized, personally reached out to people on LinkedIn. She found out who's the recruiter, she sent them her resume, she told them a little bit about her story. And I think, you know, it's never 100% clear what happened in every individual case because I don't have any access to the algorithm that that makes these decisions on people. Sadly, the vendors themselves often don't know this, and the employers who use these tools often don't also know actually how decisions are made, and that saddens me. But um, so we can't say 100% what what happened, but like he she sort of um uh speculated that you know she's a bit older than others. Um, so she doesn't fit the normal trajectory, right? Because she was first in the military, she was deployed, she was older, she's African American, she's from the Brong, she's female, doesn't fit the typical bill. So she was wondering, oh, did I fall out some sort of um matrix that um was established by the AI? And I think what's interesting is that we found uh when I talked to other researchers um who have actually queried um C-suite leadership in companies where uh people use AI tools, and almost 90% um said that yes, they use AI tools and um they know that qualified candidates are being rejected by the tools. Um almost 90% of C-suite leaders know that the tools don't work as advertised. Still, I think so many um HR departments are absolutely overwhelmed with the amount of applications they get, and there is an even more bigger push to use technology, even though everyone sort of in the industry agrees that there is a real problem with these tools. So I can't say for sure what happened in her case, but I can say for sure that uh that even C-suite leadership are dubious about how good their own tools are. And I think what what I found out additionally when I talked to a lot of employment lawyers and sort of adjacent people who sometimes get, you know, when a vendor talks to an employer, sometimes they uh they bring in outside counsel to look at the tools. And that what I found out when I talked to people that some keywords um are more. Than suspicious that are being used. In one case, it was like the word Thomas that was uh, you know, get you more points. And in another uh instance, it was when you had the word softball on your resume, you got fewer points. And if you had the word baseball on your resume, you got more points. In the United States, like mostly women play softball and men play baseball. So this is, you know, probably gender discrimination that is happening, right? In another case, um, the tool um uh used uh African American or Africa as uh one of the words. So you see sort of like, and I think that the tool probably in the background did what it did, what it does all the time. It looks for statistical inferences. Um so if you train it, like a lot of companies um don't want to just use the job description um and like the keywords in the job description and see if it's on the resume, which is sort of what a lot of um tools do. Um they want to be a little bit more, you know, want their secret sauce, like what, you know, the people that we have here, like what do they have? What's what do they secretly have in common? That's the new people that we want as applicants. And I think um, you know, they often give um uh the AI tool like a pile of resumes of people who are currently in the company and are deemed successful. Obviously, another questionable criteria, but we'll we'll go with it. And uh the tool does what it does best, it does the statistical analysis. So maybe I don't know, you had a lot of Thomas's in your company, and statistically that is relevant, right? Like you had a lot of people that had the word baseball on their resume. And none of these jobs were for a baseball or softball trainer, right? They were just one-of-the-mill jobs. So um, your hobbies, who you are, your name, your racial identity, none of that should actually be taken into consideration, right? Because we know from the past that that is really problematic when we do that, because we as human have a lot of human bias. So we shouldn't put that into the machine. Um, and we really shouldn't leave the machine running unsupervised. And only if you bring in lawyers from the outside, right? Neither the vendors nor the companies caught this. They only caught this when they're brought in outside's uh legal help. Um, so that should have processed too. That like um this seems to be routinely built in. And I talked to a couple of lawyers who said, yep, we found a problem in every tool. Others said, like, oh, we found a problem in every fourth tool. Um, and they ran the gamut from like video interview was the biggest at the time that was looked at, but they also found it for assessment games that uh were being played at the time, resume parsers, right? They're very um um ubiquitous. And I think most people today now know that if you upload your application materials on LinkedIn, indeed, and other platforms, that AI is being used on you. That was not the case in 2018, but probably most people are aware. But how it's being used, we we have very we know very little about that.

SPEAKER_00 24:58

And at the beginning of this conversation, you

Hidden Bias In Keywords And Data

SPEAKER_00 25:00

talked about your early days in investigative journalism, looking at systemic problems. And when I'm listening to what you're talking about here is I'm I'm hearing this kind of linear technical solution. And what's missing, one of the obvious things missing is context. There, you know, you you can smile in a certain context and it can be seen as appropriate. But you could if you smile maybe at a funeral, uh, when they're you know, it might not be seen as appropriate. You know what I mean? So so these tools are not really able to to even in fair context because they're just literally taking uh, as you say, profiles, running statistical analysis on it, on taking words and not even understanding the the weighting of those words, like as you say, baseball versus softball or thomas, they're not seeing is that significant or in any way highly correlated. Well, I suppose there's a correlation because that's what they're inferring, but uh so tell me a little bit about the yeah, tell me about the systemic aspect of this then. When I mean, did did you go there with people like going, okay guys, but you're this thing is is busy hammering it away here, but what about the bigger picture and how the you know, so so because one of the challenges you have is when you're filtering, you see, you're not screening people into the to the recruitment program, you're screening them out, yeah, and that's how you miss uh some exceptional candidates, right? Yeah, um, and if you look at the success rate of tech billionaires, many of them dropped out of college, they wouldn't even get past the first the the first screening, right?

SPEAKER_01 26:31

So what did you discover when you yeah, um so yeah, I mean I think it's it's sort of like the the the bigger systemic problem is right that like we have this like new crop of tech tools um coming into the space that that promises the world, right? Like they're gonna democratize hiring, like all these companies said this is not biased, and uh, you know, everyone gets the same fair chance. Um it sounds really great, but when you start looking under the hood, you find problem after problem after problem. And I think that's part of the systemic problem here is that like the vendors come into the field and um no one wants to talk about this, like if there's a problem, right? Like the vendors don't want to talk about that, you know, maybe an employer dropped them because they found out there's racial bias in the tool. The interesting thing is like I wish employers would talk about these failed pilots or or when things didn't work out, right? Because the problem is if they don't share this publicly, the next company will believe the marketing and use the tool, right? And I think employers are, at least in the United States, just terrified of liability, right? That they have used a tool that then um singled out women, for example, or you know, anyone who had the word uh women on their resume, which has happened. Um, so they feel like, whoa, we're gonna have a million-dollar, probably multi-billion dollar lawsuit at our hand. And the job applicants, which are the third uh uh stakeholders in this triad, they have no clue why they're being filtered out, right? And they can't, like, it's even really hard to bring forward a case because you kind of have to show uh wrongdoing. And I mean, you're rejected, and every judge says, well, hundreds of people are rejected, right? I think that's also why this works so well, because we are sort of, as an applicant, we know we're gonna get rejected from so many jobs, right? That is routinely happened to anyone. So why would you question that? Why would you question this was a faulty algorithm? There's only so many people who did, and I'm so grateful for the ones that did question, step forward and showed me their documentation and their um material. And there was somebody in um, well, he was based in Spain at the time, but he applied to a job in the UK, and he was like, wait, I like, I like do like web programming and web scraping. Why am I being asked to play a Tetris game here as an assessment? And he like, you know, documented everything and did screenshots and and shared that with me. And he actually um and he got a pretty immediate uh rejection, and he was like, wait a second, it's a bank holiday in in the UK. Like I only did this yesterday. Like, how on earth am I getting a rejection so quickly? This cannot be like this must be an automated decision. And so he

GDPR, Transparency And Legal Pressure

SPEAKER_01 29:10

used the uh GDPR or the UK has sort of a similar uh uh private um uh data protection law, a UK-based one. And uh he got the company to reveal like how he was assessed, and it was like an incredible data trove. And in fact, there was an independent uh office in the UK uh that adjudicated his case. Um, and uh there was a settlement with the company, and uh they promised not to use it, not to use this anymore and not to not disclose that they're using AI, I should say. Um, so like those are like really interesting cases, but it's like one in a million people who sort of question what is happening to them, right? And then also know the applicable laws to get more information. Um so we know very little, and I think that's part of the systemic issue. There's no regulatory agency in in the world that looks at these tools um before they go to market, right? And I think a lot of vendor, a lot of companies that go buy these tools, they assume that like somebody must have looked at this, right? Like this must be real because it's a tool that you can buy and it costs hundreds of thousands of dollars. But in reality, yes, we have laws that guide um employment and hiring, um, that it has to be fair. But if no one looks, um, it doesn't mean that the tools are fair. Um, and we don't see a lot of regulatory appetite to uh to look uh to force companies um to open this up to government agencies, although I would argue uh some of them have um could ask for that, but they don't.

SPEAKER_00 30:40

And you do remind me when you mentioned GDPR and oversight, um, there is a directive at the EU level. Now I've I've I've come across it a year or two ago. I think it came in to effect, was it last year? But I think all it does is it just says there has to be human oversight. And like many of these things, it that could be as woolly as they get, right? Because you you could say, Well, I looked at it, or we did it, we did a random check on C resumes and CVs, and basically we think the tool's doing okay.

SPEAKER_01 31:11

Um Yeah, I think we have now, I mean, we do have like very different regulatory like philosophies in the United States, right? And in in in Europe, and I do think that like um if there's any regulatory changes, um and that maybe a more aggressive stance comes from the EU usually. And what we see in the EU AI Act, we do see that, for example, their motion recognition software is outlawed in most cases. I think there's like very narrow policing cases that uh may um uh get in get an exception, but you know, the bar is pretty high for that. Um, so basically, that should eradicate that problem in hiring. Um, and in fact, there has been a lot of public pressure in the years before, and a lot of vendors have abandoned it because there really is no science here. Um, and I think they really um folded under public pressure here. Um, but I think um, and I was really heartened to see that in the UAI Act, for one of the first acts, it actually puts hiring and work as a high-stakes place because usually that gets forgotten. People think a lot about policing and like unfair algorithms in that space and in sentencing, and that's all very important, right? Social uh benefits, um, but not many people think about hiring. Um, but there is sort of like what is the implementation going to look like, right? How are we checking that these algorithms are fair? That's really not resolved yet. Um, and what often happens is that the vendors themselves who build the technology offer to do this fairness check, and they often check for uh gender and racial uh bias, but they don't check for for disability and other biases, and they also don't check for, so for one example of one of the game-based assessments, um, the company did pay a computer scientist to test the tool. And when I talked to him while he was testing it, he told me that, like, yeah, that you know, it's fine when you look at uh women against men and different racial um uh identities versus each other. But when he took a little closer look, he found that like African American women versus uh different identity categories of men, they actually were under what it's called the four or fifth rule in in the United States. So there was a presumed larger bias than is is allowed under under the guidelines. Um then the company pressured him to not put that in the final report, so no one knew about it. But like we just don't look, and the vendors have no incentive to look deeper here, and that's another colossal problem. It's all in this black box of algorithms that the information gets lost, and that is um really, really worrisome.

SPEAKER_00 33:59

You're listening to leading people with me, Jerry Murray. My guest this week is Hilke Shellman. Coming up, why this is not just a technology issue but a people issue. And what leaders and HR teams should be asking before they trust AI in high-stakes talent decisions. Stay with us. Yeah, and and you've reminded me that uh on the the back cover of the book you you even call this a civil rights issue. Oh, yeah. Yes. Uh because as you say, I suppose people just accept that you get rejected. I mean, I have daughters in their twenties, and and when they started, when they graduated at university, it was just like banging off applications and getting no response. Now, I suppose the clever thing about the AI is it will respond to you because it doesn't require human beings, it'll it'll invent some sort of it'll select an answer from a group of three or

Generative AI Floods The Pipeline

SPEAKER_00 35:00

four, and and maybe say to you, We're at this we're not processing you any further in our so they will get a response. But there's a sort of acceptance that this is the way life is, right? And you you're just gonna bang out as many resumes and CVs as you can, and at some point you're gonna get some progress.

SPEAKER_01 35:17

You will get money, somebody will pick it up, and and and maybe some will at one point, right? And like, and like, you know, I think there is also a difference, like most very large companies use these tools, right? Because they're inundated. Like, is a small nonprofit with three employees that hires a new employee every two years? Are they really gonna use these tools? Probably not, right? Like, um, so who knows um uh who you come across, you know, what day on what desk. Um, but I think um you want to make sure that your uh resume is sort of like machine readable and has all of these things to like do the first pass. I think also what we see now, so with the advent of generative AI, um what we've seen is like generally the volume of um applications has even gone up more. So there already companies were complaining the last 20 years that they get too many applications because it was so easy with one click. Now you have AI who can just generate your resume, generate a cover later. You even have AI that can do the application process for you. You can have agents for that, so you can just lean back. Um, so that has raised the volume again and again. So we see job openings closing after 24 hours, right? Like, can we get an application in? Uh, you see much tougher criteria. Um, so I think that's also problematic. Um, and you may get like, you know, like this laundry list of things that you need to know um or skills that you have. So you may have people who get through, um, but there may be mediocre in all of these things. But the people who would be really outstanding and have two or three things that really would uh uh uh put them in front and have like a good skills skill set, they get rejected because one or two criteria they they didn't meet, right? So um it's really difficult. It's really difficult out there, and it's getting uh worse and worse. And generative AI, I think, was like helping some early adopters really quickly, and now it's just leads to more proliferation. You know, like HR teams complain about this and they feel like, oh, they're all using AI. And I sort of feel like, well, I mean, we kind of started using AI. You can't really fault uh job applicants for using it. Um, so I think what we're seeing now, which is a little bit unfortunate, when I went to um uh HR Tech, which is one of the biggest conferences in the field, probably the biggest ones, where like 10,000 people or so convene every year in Las Vegas, and you have employers, you have the vendors, all there showing the technology. And people were very keenly aware of this. And some companies openly talked about how they moved um their hiring intake from like LinkedIn, indeed, and other sources, and moved much more to um employee referral. So, like um, and we have like 10 years or so, the industry sort of moved away from that because it was very clear that like if you rely too much on uh employee referral, um, usually employees refer the same people that they like, that they have beers with, right? You're gonna hire the same people again that have the same similar socioeconomic background, probably went to school with uh the people that you already have. It doesn't lead to diverse like uh or workforces. Um, it leads to the same thing. But we see that like now HR teams moving back towards that because they feel like, well, you know what? An employee will at least uh recommend a human and somebody who's probably somewhat qualified. Because there's also not only a problem with the proliferation um of application, there's a problem with um uh, you know, people swapping out an assessment. So like I apply for a um software developer position, but I'm not really not that good. So I let Suzy take the test uh because no one is watching, right? It's like sort of mediated on a computer. So Susie takes the test, ACES the test. I get hired, I show up, and it takes the company two months till they find out I'm actually not that good at what they thought I would be good at. So it's a huge problem. And then now we also have avatars taking the job interviews. We actually have avatars on both sides. Like some companies, instead of one-way video interviews where you have like sort of a pre-recorded video or something of like, hello, thank you for applying, answer these questions. Now we have avatars conducting them so they can be responsive to the applicants, what they say, and sort of ask more targeted questions. But on the other side, then applicants also sending their avatars. So you have like sometimes avatars are talking to avatars, and so it's sort of like I don't know what we're doing. How we can find any substantial material information here that would help you with hiring is unclear to me.

SPEAKER_00 39:42

Yeah. Um Andrew Palmer, the economist on his boss class uh podcast in the last series, he actually tested this out and he sent, he got an AI generative to send the applications. He told them the type of job he was looking for, and he said it. I mean what you just said, it became absurd because he said he was getting he was applying for things and getting rejected for things he didn't even show any interest in because his AI thought he might be interested in these things, etc. And and he was I mean, he's got a very witty, dry sense of humor, so he presented it in quite a uh a tongue-in-cheek kind of way, but he did he did expose how

When Algorithms Help And When They Harm

SPEAKER_00 40:20

crazy it is out there. But I want to ask you a question are all algorithms um bad? Because in the good old days of normal psychometric testing, they all had some form of mathematical algorithm computing behavioral traits or stuff like that. Um, and even the great Daniel Kahnman, God rest him, um, in his book Noise, talked about the the importance of uh having some sort of algorithm to reduce noise and bias in the hiring process. So when you were talking to all these experts, it it did any of them sort of say, well, this the traditional algorithms were okay, but once you let AI loose on them and they start um amplifying things that they shouldn't amplify, then the whole thing becomes that the whole value of an algorithm just gets yeah, reduced. Yeah. Come across anything like that?

SPEAKER_01 41:11

Yeah, I mean, I think it depends what experts you talk to, right? Like there is sort of this divide of like um industrial organizational psychologists who used to build these assessments, right? And they come with like a clear job analysis. They will check like, well, what do you actually need for the job, right? And then try to build assessments um that are targeted to that job. It's very cost intensive, right? Um you can you can sort of assume this this takes a while and it has to be job specific. So data scientists say, like, we don't need all of this. We just take the people who are already in the job and will copy how they do the job. The problem is with that, you bring in um uh sort of all these problematic things, right? You bring in this correlation soup, uh, which, you know, maybe this correlates to something, but that doesn't make mean that it's causally affected. Yes, it might be that people with brown hair are very successful as accountants. Um, that is a correlation, but that's obviously not meaningful. Any human would know that because brown hair doesn't qualify you for anything. It just means that lots of people in the world have brown hair. So probably that was a criteria, right? So I'm sort of pushing this argument a little bit, but that's how computers do this, right? They have no, they have no morals, they have no ethics, they don't understand what like sort of the moral implications of hiring is, right? That we shouldn't look at identity or gender or or any kind of like hobbies or anything that we should really look at, like your capabilities, your uh uh, you know, your experience, um, your skills. That should be the criteria, not all these other things that we as human bring with us. If you're smiling at a certain time, it's not relevant. Uh, but computers don't have this knowledge. So uh yeah, I'm actually like I use technology all the time. Um, I just think we need to use it responsibly. And I think the the way the promise and the reality of these tools in in hiring is really not um uh does really not live up to the promise. Um so that worries me. And I think also the the lack of independent testing um worries me here. Um and uh all other things. Also, like, I mean, the the I think one of the real problems is like we know from psychometrics and other science that the highest validity is like uh meaning the most successful thing is like testing people on the job. Tells you how good they will be at the job, right? Um so like personality tests, which a lot of I think American companies have gravitated to, um, are actually have very little implications how good you'll be at the job. Um, it's like 10% on a good day. And that's really on a good day. So 90% of like how you are acting on a job has nothing to do with your personality. So you might want to move away from that at all. Or if you have to do that, like combine it with assessments about the actual skills on the job, right?

SPEAKER_00 43:56

Yeah.

SPEAKER_01 43:57

Um, so I think there is a way that we could use technology. Technology for the Bettermen and actually think through like, well, what are the things you need to do at the job? But that like really requires a lot of thought and a lot of thoughtful building. And I'm not sure if the companies are up for that. But I actually think with like virtual reality, you could do a lot. You could like actually preview a job for people. Right, you know, it's obviously not feasible to hire 100 people and then let 99 go after 100 days, right? That sounds feasible, but you could sort of really think through like, oh, can I have a computer simulation of this? And uh, can I ask people to do the job? Um we don't see a whole lot of that. Um, so I think a lot of people believe what sort of data scientists have said, like, oh, we'll just take this, like more data is better. We'll just take everything that we have, and now we have a lot of data in people, and and we filter it through and like make make a prediction. And the problem is the computer cannot understand what is good data and bad data, right? Like, and if you have uh, we all have uh human bias in the hiring process in the past, and now that gets put into the training data and gets uh re-jurgitated by the algorithm, and then it brings its own statistical inferences and problems into the mix. Um, so that that worries me. Um, so I use AI all the time. I just think when we use it for high-stakes human decision making, we have to do better. And there is also a question, for example, like if you have a disability, like how on earth can like a uh person, you know, and disabilities um uh show up differently for everyone, right? Maybe you have somebody who has an autism diagnosis in the training data, but then you interview the next person who also has an um autism diagnosis, but their autism may show up very differently. So, how on earth can that be part of a statistical inference? That is really, I really do not understand how we can account for people with disabilities and say this is a fair hiring process. Um, so there are lots of questions that that that haven't been answered. Um so, yes to technology, but only in like um a very clear functioning way. And I think the other way in a very um thoughtful way, and the other question is also like, you know, with hiring, you are predicting the future, right? And it's like, and it's not just your skills and capabilities that are important, it's also like the makeup of a team, right? And then your personal situation, something that will never, hopefully never be part of the hiring process. That like, you know, I might quit a job before because I need to take care of my parents. I have a difficult pregnancy, who knows? Like, it's incredibly hard to predict the future. Um, because now we see predictions of flight risk and other things that, like, um uh, you know, sure there are some behavioral parameters, but there are also things that like you can never be predicted. So, like, can we actually predict uh kind of success um on the job based on these ephemeral um uh uh things that just show

AI Reveals A Broken Hiring System

SPEAKER_01 47:00

up in a resume? I doubt it. And I think what really happens here is like AI sort of exposes a semi-broken system that we already have, right? Like, like resumes, we have them, we love them, a whole industry is dependent on it. They're not that great uh to understand how good you will be at the job, right? Because I might look for a senior software developer, and you know, it says in the job description you need to have Python knowledge, yada yada yada yada yeah. So like 99.999% of applications will have Python expertise and skill on their resume, right? But the resume doesn't tell me at all what kind of skill level you have at all. Um so it's actually not very actually meaningful criteria, but we are really relying on it. So I think it just exposes this problematic system, right? We rely on interviews where you have people talk about how they do the work. They're not actually doing the work. It turns out some people are very skilled at talking about how they do their work. It doesn't mean that they're actually good at doing their work, right? We call this uh uh uh uh you know a uh uh a competency uh problem where like people are really good at talking about it and they sort of exude confidence, and then we believe as humans that they can do it.

SPEAKER_00 48:11

Um men are better at it than women apparently.

SPEAKER_01 48:15

Very, very, very, very much so. And I think everyone had that um experience that you feel like, wow, that guy is really confident, but they're actually not competent at the job. How did they get the job? Well, because they exuded this confidence, um, and that is a real problem because they actually are not competent. Um, so um, so actually, like job interviews, I'm sure they're not gonna go away because people want to meet each other, right? And they're at the end of the hiring process when humans still step in and make a decision, they're probably not going away. But really, like science tells us they're really not a good criteria. Um, we have a somewhat broken system already, and now AI is just showing us that it really doesn't work by amplifying this um already problematic process.

SPEAKER_00 49:00

I actually have a few things if you if we could just uh unpack a couple of things there, but one thought that came to my mind was from my background as a musician, you you never hire a musician by looking at, you know, looking at their resume or that. You you sit down, you play music with them. Now, I happen to play Irish folk music, jazz musicians would do the same, rock musicians the same, but even in an orchestra, they have to audition, they have to show they can do what they're supposed to be able to do. And yet, when it comes to the world of work, we seem to not have we don't seem to see it that way for some reason. And the other thing I wanted to just comment on was you know, because you talked about the data scientists saying, yeah, we don't need all this stuff, etc. But you see, for me, the big issue here is we're dealing with human beings and we're dealing with human beings and their lives and their livelihoods and that. And as you were you were explaining a lot of stuff earlier, even uh I I mentioned Kanman in a different context, but Kanman also said, and I've had this experience um because I've been privileged to work with a lot of very senior people. He said there was no secret source, there was no secret formula. Most of the people they were maybe above average, but the guys who get the guys often, but even people who get to the top of organizations, it's a certain lucky thread by first of all getting in, getting that first job that gives them the you know, whatever sort of visibility and credibility in that. And yet there's so many other people out there who could possibly have that path, but if they can't get they can't get into the train or whatever it is at the beginning, they can never get to the destination or to any of the stops on the way if they're not on the right train. So it's I think we have to rem remember that this is about human beings. Even the Pope is weighing in on this. Uh, in the last few days as we're recording this, he's just come out with one of his um what are they called again? They bring them out every so often. It's called Magnifica Humana or something like that. And he's he's talking about the the and and actually, believe it or not, because I've seen a I've seen a pressy of this, somebody has done a kind of synthesis of it on summary. He talked about recruitment and hiring people and how AI it is fundamentally probably unfair to be using this kind of technology because of the impact it has on people's futures. Yeah. The Pope is weighing in on that.

SPEAKER_01 51:24

Yeah, I mean, I think, you know, I think, you know, sometimes we feel like um, you know, maybe even as journalists, right? Like, why is like like why is this so slow? Like we've been talking now for for years that we found these problematic things in these tools. And I do think like um uh like actually some companies have moved away, right, and abandoned some of the uh the the most out out outrageous usages of like the emotion recognition um on people's faces, for example. And there's often other companies that six months later come up with the same thing, and you're like, no, no, no, we already just show it, there's no sign. So that can be uh um you know, can be a little frustrating, but um, there is progress, things do change, companies do abandon some things because there is public pressure. And I also see that like, you know, now the Pope has like understood this is a problem, right? Like I also see this when I go to like even like school classes of like middle schoolers, right? Like when I first showed them my first video in 2018, they all wanted to know, like, oh man, this like facial expression analysis, how that, you know, how does that work? And la la la. And today, the first things the kids say is like, wow, that could be really gender or racial bias. I mean, it's like so clear how like the public has become much more educated on this issue and does understand that these algorithms governing in our lives uh can have profound unfair decisions. I think we just haven't built up sort of the regulatory um uh structure here to actually understand, well, how does it work case by case, right? And how and how can we regulate it well? And I think what's clear um to most people is that the self-regulation of tech companies does not work unless unless it hits their bottom line, right? If the public outrage is so vast that you like, I don't know, that you have like black Nazis or something that your your tool generates, you'll stop generating that, right? Like there is like uh some change, uh, but a lot of times criticism just goes nowhere until it hits a company's bottom line. So they're not interested in reforming. Um they might say that in uh you know hearings, um, but they're not actually uh walking the walk. So like there's there needs to be other ways to do this. Um there needs to be maybe like more public AI, different ways, alternatives. Um there needs to be um more clear regulatory structures, maybe also more fines, that it really hurts a company's bottom line. Um, you know, we've seen some changes on like city level, state level in the United States. Um, unfortunately, as a lot of people had said, in New York City, there is an AI hiring law. And a lot of people in the process of the hearings around the law pointed out that this law has no teeth and that they're worried. Um, and exactly that happens. We have no not seen any um regulatory um execution of this law or policing, that actually companies are required to do this. So it's sort of like didn't work, but we see some changes. I think we're getting smarter every year. Right, right.

The Accountability And Regulation Gap

SPEAKER_00 54:26

And um something I suppose we have to acknowledge is that I mean HR people are I work with a lot of HR people, they're stretched, they're really stressed and stretched by the amount of stuff. So they're actually coming from a very I mean, and I I've yet to meet a HR person who didn't care about people. Well, you meet the occasional one, but for the most part, they're in that work because they really care. And if they're kind of constrained. So if we were to, if you were to like if you were advising a leader or HR director or some people managers today, what sort of questions should they be asking before they start trusting any of these systems, like an AI system, to make or even influence people decisions?

Due Diligence Questions For HR

SPEAKER_00 55:05

Because these are expensive decisions. You know, it's always funny. I always think how a finance director and these guys, as you say, the 90% of the C-suite can say, yeah, we kind of know it doesn't work that well, etc. But a finance director would never uh go out and buy a company or merge without due diligence of the highest order, and yet you could be putting some of these people into roles where they can do serious damage to your business by yeah, by being incompetent, for example. So if you were advising somebody out there who's in a position to make decisions about this, what sorts of things would you say they should be looking into or asking before they just go out there and buy this stuff?

SPEAKER_01 55:43

Yeah, I mean, I think like, you know, never give in to the hype, right? We often see in um, you know, I think m and we we see this now as well. Like the CEOs would be like, you know, like uh uh texting their uh CHRO, like, hey, like we need to use AI, everyone's using AI. And it's sort of like, okay, like what is the use case here? Like, is this actually a problem AI can solve? Um, or are we just putting something that we can't solve to an AI tool that also can't solve it? And now we have like an objective algorithm. And I think for some companies, the efficiency and the labor savings that these tools uh bring them is what drives this decision because HR departments usually don't bring in money. They're not seen as strategically the most important parts of the business, although I would say that human labor is probably the most important thing. But uh I'm I you know I don't argue with uh with uh with CEOs, you know, they don't usually bring in money. So they're seen as a cost center. Um, so I think um um they're kind of at like an kind of a little bit of a of a dis disadvantage here, right? But then I would like bring some um expertise in-house. Do not believe what the vendors tell you because they will tell you the moon that these tools like democratize hiring, they're fair for everyone. Like, really understand like what is the training data built on? What does it actually do? How is it validated? Look, if the company doesn't have what's called a technical report or validation report, I would not at all do business with them because they need to show how they tested the algorithm that this algorithm actually works on this. And then uh maybe bring in outside counsel, definitely run a pilot before you unleash it on everyone, right? Like run a like sandbox pilot where you use parts of your of uh the people to run this through while you also have your regular hiring process um to actually test like, does this lead uh to a more diverse candidate? Does this lead to more qualification? And you can really test this, right? Like you can just see, like, oh, Hilka got uh uh a no in hiring, uh, but she was recommended with our traditional process. Let's hire her and let's see what happened in two years, right? What actually happened? Did she turn out to be a no or was she a yes? And you obviously have to do that en masse, right? Because one Hilka doesn't mean any statistically, absolutely insignificant. Um, but we do this kind of uh testing with medications and other things, right? Where we test on like large populations. I hope some company out there is doing that and hopefully publishes this. I've yet to talk to them. For some reason, HR seems weirdly decoupled from often from the rest of the company, um, that they don't run these like large studies. Um I would demand if I spent millions of dollars on a tool that I understand if it works or not. Like um, but we don't really see this. Um, maybe also because things change in HR departments. People aren't in the same job for five years or so that they could actually run this kind of study, right? There are all kinds of different things that influence this. But um uh yeah, those are the questions I would and I would bring the the um like either hire outside counsel or have really you also need to monitor once algorithms are deployed. So you really should have some sort of in-house expertise and don't leave it to the vendors that have built the technology because that's obviously a conflict of interest if they um have to audit their own algorithms. I mean, I'm shocked they don't find anything problematic. Um, you know, we have we have this, we had a whole uh uh worldwide recession, right, around the housing crisis where um uh uh companies were paid to rate uh mortgages by the people who originated their mortgages and they were all rated triple A. Um, it's a conflict of interest. Um, so we know where that goes. Um so I wouldn't recommend

Takeaways For Employers And Applicants

SPEAKER_01 59:24

that.

SPEAKER_00 59:24

Okay, so coming to the end of our conversation, I think we could talk for hours about this topic.

SPEAKER_01 59:29

It's fantastic on and this looks so fascinating.

SPEAKER_00 59:32

Yeah, if my listeners could take away just one big idea or one big takeaway, what would that be?

SPEAKER_01 59:39

I think, you know, like really um and you know, I think I have different takeaways for for different stakeholders, specifically for people who who work in companies. Like I wouldn't trust anything a vendor says, I would always trust but verify. And I urge everyone to verify and then share that knowledge. Please, please, please find a way to share that knowledge so that the next company doesn't do the same mistake, right? And then for job applicants, I want to be very clear that like, like, you know, don't give up. Um, like it's not you, it's probably the algorithm. There's sort of different ways to mitigate it, right? Like make sure your um uh resume is machine readable and like, but really think through like, okay, who can I like, who do I know with different companies? Can I do more uh networking? Um, because it is a numbers game, but you can influence us a little bit, but everyone is suffering right now. Everyone is sending out hundreds, if not thousands, of applications. Everyone feels the strain and the hardship. Um, so it's not you. Um, it's probably this algorithmic system that we built, and it sucks.

SPEAKER_00 1:00:42

And and one thing you mentioned earlier about going to this idea of recommending people, you you know, I've had this experience even just working as a consultant and running a small business. Uh, your friends won't recommend you unless they, you know, they don't take the risk for their own reputation unless they really believe in you. So it can have a very, very high probably reliability when you get a recommendation. And by the way, listeners, uh, if you are interested in in the whole job market stuff, Hilka has loads of stuff out there. You you can find her on YouTube doing interviews about how to actually approach the application process and that we just don't have time for it today. Um, I'll put some links in the show notes, but how should people reach out if they want to to strike up a conversation with you or get your advice? Where's the best place? LinkedIn or your website, what what do you recommend?

How To Reach Hilke

SPEAKER_01 1:01:32

Sorry, my my website is fine too, but like LinkedIn is probably uh the best place. And there's only one Hilka Shalman on this earth, so you'll find me. Um there's also, you know, my email address is also on N NYU's website. Um, but I do converse a lot with people on LinkedIn because that is sort of the place where everyone congregates to talk about jobs and hiring, right? So it's like a really natural uh place. Um, so that's probably uh uh the the easiest way. Um and I'm happy to help people with with resources. Like I have to sometimes people ask me, like, how should I, like, what should I do if I'm being asked in a job interview uh that I was uh laid off. And like uh I want to say like I'm not a recruiter or career coach, right? Like I've sort of investigated uh systematically what goes on, so I can tell you in general, here's what works with your resume, and to to uh make it through this algorithmically driven world. Um I'm certainly not a career coach.

SPEAKER_00 1:02:30

Okay, so the one and only Hilke Shellman.

SPEAKER_01 1:02:33

Thanks for sharing that is a gift of having a very unique name.

SPEAKER_00 1:02:39

Yeah, I well I'd just like to thank you on behalf of myself and my listeners here for sharing your insights, your tips, and your wisdom with us all here today.

SPEAKER_01 1:02:47

Oh, yeah, thank you for having me. I love having these conversations. Obviously, um, you know, these algorithms are a passion of mine. I'm not I'm not letting them go quite yet.

SPEAKER_00 1:02:58

I I I sense that. And that's it for this episode of Leading People.

Final Thanks And How To Support

SPEAKER_00 1:03:05

If you're new here and enjoyed today's conversation, I'd love it if you subscribed. It's the best way to make sure you never miss an episode. And if you're a regular listener, the best recommendation is always a personal one. So if this episode made an impact on you, please share it with a friend or colleague. It might just be the best thing they listen to this week. Until next time.