Hello, everyone, and welcome to another episode of the Leadership Growth Podcast. I'm your host, Daniel Stewart, joined by my brother, Peter Stewart, and we have a fantastic guest with us as well. Dr. Reece Akhtar is joining us, and I'll introduce him more in a moment. But Reece, welcome to the Leadership Growth Podcast. Thank you, Daniel. Thank you, Peter. It's good to be here. Fantastic. And so we're going to dive into a topic that's on a lot of our minds, and it's what in the world is going on with AI and hiring and promotion? And frankly, how can AI help us as we seek to find the right people and as we seek to promote the right people as well? How can we leverage AI? And we're very excited to have Reece here, one of the experts in this field. So let me share briefly his background, and then we'll dive into the conversation here. So Dr. Reece Akhtar, CEO and co-founder of Deeper Signals, a company transforming talent assessment through scientific assessments and AI. As an organizational psychologist and data scientist, he is also the author of the book, "The Future of Recruitment,""Using the New Science of Talent Analytics to Get Your Hiring Right." He also publishes a lot of places and has a lot of other accolades. And yes, he's studied in lots of great places. And Reece, thank you again for joining us. Thank you for the very generous introduction. All right, Reece. So here's an opening question. Give us kind of a current state of things with AI and hiring specifically. There's a lot of things that are being experimented with and tried with and wanting to understand better. Give us a lay of the land. What is currently going on with AI and hiring? Yeah, I um... The first thing that comes to mind is that, quote, "the future's here. It's just not normally distributed." And so there is probably a lot more talk about using AI in your talent strategy than what organizations are actually doing. And I think that reflects, one, a lot of excitement about the promise of a new technology to solve probably one of the most important things an organization has to overcome or any leader has to overcome. Me and my colleagues, we like to say all organizations have problems. Most of those problems are people problems. And so if you can staff your organization, your leaders, your teams with the right people, everything else can be managed and sorted. So really, what is the promise of AI in the hiring or talent strategy space? The problem is finding good people, people that are talented, people that have the ability to get the job done to a high standard and to be reliable and dependable is extremely hard. For anyone that has had to lead or conduct kind of hiring for their organization will know that it can feel oftentimes like a 50-50 kind of chance whether you're going to get it right. What we do as IO psychologists over the last 50 or so years has been trying to move those odds from 50-50 to more like 80-20. And the promise of AI is that we're going to get to that 80-20 split or those 80-20 chances much sooner, much faster, and a much more affordable price point from time, money, and resource perspective. Whether we're going to get there is still being kind of tested and being proven out, right? There is lots of really exciting research and case studies to show that these types of tools can really, really work. But it's not a silver bullet. To use AI or machine learning in your hiring decisions, you need a couple of key ingredients. The first is you need to actually know what to look for and this is the biggest kind of tripping point for most organizations. They have loosely defined kind of frameworks or perspectives of their competencies or even how they are thinking about talent potential. Oftentimes it's potential for what... what are you actually expecting or looking for within a role. So that's the first thing you need to get right which is much more conceptual. The second is around your process. Good use of artificial intelligence to make fair, equitable, and predictive insights requires good quality data. And for those of us that have worked in HR will know that talent data or people data can be very spotty, it can be inconsistent, and it can also be somewhat unreliable. And so if you're trying to train an AI algorithm or machine learning algorithm to improve your decision making, you have to make sure that it has high quality inputs to get good output. And the third part of it is, I would say, more cultural or kind of more self-reflective. There is suspicion of or fear, let's say, of using technology and AI to improve or to make our hiring decisions for us. And understandably so. There are many cases where AI is perpetuating bias and unfairness. I think you only have to open up Google to kind of see the latest case of that happening. But we have to also counter that with a bit of self-awareness, which is humans are fantastic at perpetuating bias and unfairness and unequal decisions. And so really the promise of AI in that case is for us to just one, accept that we're pretty bad judges of talent and that we are prone to over-exaggerating our ability to do so, leading to bias. And actually start to trust AI as an additional tool to help counter that bias. It's not about kind of handing away our decision-making rights. It's about using AI as a tool, emphasis on a tool, to help us become smarter humans, not the other way around. So just to recap, it's about, one, having a good framework to know what you're looking for. Two, it's about making sure you've got the right data to train the AI algorithm. And then three, it's about kind of having more self-awareness about our own limitations as judges and decision makers so we can actually use the tool to make us smarter, not more incompetent.- That's so helpful as you lay out that perspective. And we could do a deep dive into each of those three steps and kind of looking at it as we started off with this conversation, kind of what's the state of AI right now in terms of helping to find talent, to help recruit talent, help to leverage it. Where has the early adoption been in that aspect? Where's kind of been that low-hanging fruit? And then as we discuss that, like, where do you see the next level, the next stage? Those are really now leveraging in a way to separate themselves from their competitors.- Yeah, great question. So people have been, organizations have been using machine learning in their hiring practices for a very long time, right? Machine learning uh you know is just another form of statistics and machine learning is just a branch of AI, right? So we can go back uh decades to see organizations that have been using uh data to make better hiring decisions and this is taking the form of um you know psychometric surveys as being like one example of how you can collect data, develop a scoring model using kind of regression or machine learning techniques to then give an estimate around someone's potential to be a good fit for a role. That's been happening for decades. And I would say most of the Fortune 500 organizations are doing some form of that. Now, that process is still quite manual. It can still be quite tedious to administer. And so what we're trying to do is increase the speed and the accuracy as well as the affordability of getting those types of insights. So then what you can do is look at organizations that are adopting kind of new methods, new ways of assessing talent, and then new ways of scoring it. So I would say the first kind of high profile example of this was maybe Unilever back in probably around like 2015, 2016, they were using a company that has since been sold to another organization called Pymetrics. And they were using kind of these short games to kind of measure different cognitive abilities. And there was machine learning algorithms underneath that to improve or to get a sense around someone's cognitive ability. And then that was used to inform decision-making around graduate programs. I would say that was probably one of the most notable success stories, at least that I first heard of. On the other side of that is um I think it was around a similar time uh Amazon tried to create a algorithm to screen resumes, to actually improve the selection rates of female engineers. This is like an often spoke about example because it went terribly wrong. The algorithm wasn't trained correctly, it went haywire, and was actually, in fact, penalizing female engineer applicants, which was the opposite intention. Fortunately, that's over a decade ago, and organizations have got better. And I would say organizations that are very large, well-funded, and ultimately have the volume, are all starting to use AI in some shape or form, mainly within the recruitment space, because you have a lot of volume, and volume is needed to train stable, reliable algorithms. Where I think the opportunity lies is, one, making these types of tools more suitable and ready for small to medium sized businesses. That's where a lot of work we do at Deeper Signals is trying to bring what was otherwise a very expensive, hard to use technology and make it accessible to organizations that are scaling. They don't have the entire infrastructure that, say, a Unilever or an Amazon actually has. Then the second is starting to think a bit more beyond recruitment in the high volume space and thinking more around employee development. How can we take all this data that we've collected about someone, about their personality, their strengths, their competencies, and so on in the hiring process, and actually start to use that data to actually help them be onboarded more effectively, give them more personalized feedback so that they can be more readily engaged in their work. They know exactly where they can develop, so on and so forth. So it's about kind of bridging the gap from just using AI to make better hiring decisions, to actually using that data and that methodology to improve their ability as a professional and their competence as a professional and their ability to grow and perform. The other part, which may be even more suitable for your listeners, is focusing much more on leadership, right? So how can we start to apply these types of models and frameworks to a set of individuals that are quite exclusive? Like senior C-suite executives, board members, like you don't necessarily have the volumes of data to train algorithms against. And, you know, the contexts and situations that these types of senior leaders operate in is very dynamic, very unique. And each one can vary from one organization to the next. And so how can you start to use AI in that respect to support those types of individuals when you don't have the volumes of data to train against? So it's a very interesting time. You've kind of seen as a market more of a maturation of kind of AI and machine learning in the assessment recruitment space. Because that's more of a continuation of what we've already been doing as an industry. And now we're starting to see a bigger push into using AI as a developmental or coaching tool for early career individuals all the way up.- So many of your points, they're all based on data. And you have to have data in order for an AI, a machine, to be able to take and do something with. As you pointed out, for the past many years, applicant tracking systems, you know, can scan and they can assess and they can analyze. They can look at keywords. They can then print out and help hopefully simplify. Doesn't always work, such as you pointed out, but simplify the number of resumes to then pay attention to. I hear all of this and I think, all right, data. We still need data and we need data to come in good ways and helpful ways. Does this mean that applicants are going to need to be able to just get more comfortable with spending a half hour or an hour or three hours, depending upon their level or longer, taking various assessments so that the company can have a ton of data? And this practice has happened for really 'seniorly' level leaders with really large organizations. But is this a form of a practice that will begin to need to be adapted more often just so that the AIs can have more data from which to then make decisions? Talk to us. How do we then get used to this idea of just getting a lot more data that's helpful and accurate?- Yeah, definitely. It's a great, great point. So, you know, it's like McKinsey what in like the, in the 90s, said there's like the war for talent uh you know everyone read, everyone heard about the article but no one probably read it but you know you all talk about it. I think when you actually look at kind of the state of the candidate or employee experience it's actually the war on talent. You look at like engagement levels are going down and down and down or people are just actively disengaged from their work or passively looking. But then even when you look at the candidate experience, you ask do candidates need to get more comfortable giving or dedicating like, you know, a few hours for an application. You know, they're already doing that and very unhappily like doing so because they have to dedicate so much time and energy, you know, jumping through the various hoops that an organization is asking them to do only to get kind of like no feedback or no notification whether they were successful or not. They just get ghosted. I think there is an opportunity, and this is kind of what we're doing at Deeper Signals, is to, or one of the things we're doing at Deeper Signals is to try to make that experience as painless and as easy as possible. So, you know, the promise of then AI is, you know, how can you design an assessment experience that requires actually the least amount of data to maximize kind of the accuracy or the assessment uh quality so a candidate can kind of get in and out of um of that process and uh and move on to the next. I think the other opportunity which is maybe more aspirational and um me and my colleagues have been talking about this for probably over a decade now but it's the idea of like a talent passport. Can I and, you know, just bear with me for a little bit as I claim my hopes for the future is, you know, as an individual, could I take a whole bunch of assessments, I could take all my kind of achievements, and all my qualifications, and all the other things that kind of make me great. And then it's like a single kind of point of truth that I could then give to each organization that I'm applying for. And it's a highly standardized process, it's very easy for me to do, and it's very structured for organizations to then understand and evaluate and, you know, build their models upon or even just, you know, understand who is applying for the job. That would be like an ideal place to get to, right? And... I know like blockchain this, blockchain that, like, yeah, maybe, but I think it can be far more simple than that. But the idea being if we can switch the uh the say social contract whereby we empower people more with owning their data and having more portability of their data I think we can make a better consumer employee like experience and also improve it for uh you know organizations who are looking to get such data to you know make better decisions about who they're staffing uh their organizations with. So this is more of a kind of a cultural, societal conversation we need to have which is around kind of who owns our data, how easy it is to kind of bring it from place to place and then, you know, how is that data then being used um you know after the fact. So there's all sorts of things like um like the idea of a data cooperative um or uh yeah like a talent passport. There are these concepts out there, people are talking about it. One day, I hope we get there. But I think there's still a bit of work to be done before that can happen.- Yeah, you're going down that path that my immediate, my mind was going down of if every, if we're trying to get data from applicants, and if every job is requiring their own unique set of data, like they think, oh no, we want you to take this assessment and fill out this survey and do this simulation. Oh, but then now Corporation B is saying, no, no, we don't like that assessment. We like this assessment and we want you to do this. And I mean, there's so many options out there. It is somewhat disheartening as I'm looking at this from an applicant perspective of, man, I'm having to invest so much time. And if there was somewhat of a credentialed for lack of a better word, but more of a standardized process where great, here's my talent passport to use your term. I've already taken these assessments. Here are the simulations I've performed. Here's my data. Click, share it with whoever needs it, and then they understand where it's coming from.- Yeah, it's as simple as that. Like, you know, there's no argument against it. You know, it's pro-consumer, it's pro-employee, and, you know, you are getting the, ultimately you're getting the data that's needed. And there's, well, when you think about, okay, well, you know, organization A needs something slightly different to organization B. Well, first off, there is like a jingle jangle fallacy whereby a company will say they need something, but actually it's the same kind of talent or construct as what organization B is calling for, but they just call it something different. For those of you that have, for your listeners that have worked on designing competency frameworks, you know, at various levels, I would say there's like three buckets. It's like your people skills, your emotional intelligence. Are you easy to work with, open to feedback? You know, not a pain in the butt. There is an ability piece. Are you smart? Are you competent? Are you going to learn? And then there is like a motivational piece. Are you like diligent, hardworking, driven, so on and so forth? Yeah, you can kind of break those down into smaller things. You may call them something slightly different. But those three buckets are generally what people are looking for. Those three buckets are always predictive of user performance, when you look at their literature. So it's like, you know, you could add like a standardized set of assessments or methods, let's say, to gather insights around how someone stacks up in these three broad competencies. And then, you know, you have some sort of interface on the organization side that says, okay, we're looking for this. This is what kind of, how it's been assessed in this person before. And so are they going to be a good fit? Yes or no? This is kind of where large language models may be able to help because, you know, they're great kind of understanding the semantics of language and the distance between certain words and so on. So we could get there.- Yeah. And it really helps kind of shatter that illusion of specialness that every organization wants to feel.- Oh, of course.- No, we have our own. Like, we have such a unique system and culture and process here that we have to have our own system. But as fellow consultants, as we look through across this broad swatch of organizations, they're really not that special. And they're really not that different.- Yep. And it's the opposite. It's true for, you know, what makes them great. You know, there's generally consistent things that make an organization great. There are the same kind of problems or the same things that make them bad, right, or kind of make them challenging. And organizations are far more similar than they are different. And it's actually true of most people as well, right? I think people are more similar than they are different. But, you know, that's kind of uncomfortable to admit because we all want to think we're unique and special.- Correct. And so what is one thing that organizations should stop doing as they, organizations, want to adopt AI or machine learning or just continue to evolve in helpful directions? What's a practice or a behavior they should stop doing as they recruit and promote folks?- Yeah, well, first off, don't say, "I want to start using AI" without knowing what you actually want to use it for, right? AI and its various different kind of branches is like the mechanism. It's not the answer to your problem, particularly within the recruitment talent space. So I think it always goes back to your strategy. What is your people strategy? How does it connect to the business objectives? And ultimately, what are the things, the skills, the competencies that you are looking for that are going to help you enable that strategy? That's like the first thing that you need to get right. And it's simple. And, you know, your listeners may think, yeah, you know, go figure. But you'll be surprised. Like so many people we talk to are like, you know, they've read about it in The Economist or they're hearing about it on podcasts. It's like, you know, how can we use AI? And, you know, you just sprinkle it over the top of everything you're doing and it's going to make life great. That's not going to work. So the first is, you know, please spend the time having a strong sense of your people strategy and what you're looking for. And then I think it goes back to what I said earlier. It's about having that process. Are you collecting the right inputs to feed whatever algorithm you're using or training? Or if you are buying an AI solution that's off the shelf, you need to get really crystal clear around how that algorithm works, what data points are being used to produce the outputs, and in particular, how is it trained? Was the algorithm trained on a whole bunch of students in, I don't know, North America, but you are an engineering firm in Europe, right? You'd be surprised, but that's how a lot of these things can work. In which case, how really valid or reliable are those predictions? So I think it's about, one, not reaching for off-the-shelf tools without knowing where it's coming from and the data being used. Two, it's about having the right strategy. And then three, it's about laying the foundation to collect kind of the right inputs around your hiring decisions. So going beyond just resumes and even beyond like assessments per se, but having like really strong structured interviews, that is one of the most predictive things you can do to estimate someone's likelihood of being a high performer. It's about getting better at measuring performance and look, having a more retrospective view around, okay, well, if we measure performance frequently and often against the right KPIs, what does that tell us about what we should look for? Oftentimes when we are looking to hire people, it's like based on our own intuitions around what actually good looks like. We should take a much more data-driven view and say, well, if we study one of our high performers, they have X, Y, Z attributes. Let's go look for that rather than the other way around. And finally, you have to kind of really think about the ethics of your hiring process. You can optimize for accuracy, you can optimize for price, you can optimize for ethics, and you're trying to like balance across all of those things. And ultimately, you know, if you are relying just on human intuition, you're probably not going to have the most ethical hiring practice. In the same way, if you over-index on kind of technology and automation to make those decisions for you, you're not going to be able to have a very ethical or very appropriate recruitment process. So it's about finding the balance between human and artificial intelligence or incompetence.- As you're going through those elements of successful adoption and obviously knowing that tool well, knowing it inside out, not just trusting it at face value so that you're really able to be certain, okay, is this giving us the output, the response that we really need? And as you know, with all AI, you got to engineer the right prompt to get the right response. You got to know what you're putting in there. But then you went on to, you were talking about strong structured interviews. And that felt very non-AI to me. So what helps? So you find that balance between obviously the data utilization of the AI. But what makes for a strong, structured interview? Because we've been doing interviews for decades.- Yes, yes. Well, okay, so interviews, you know, you can think of like unstructured, which is actually how most of them operate, which is, you know, you get in a room, on a call, and, you know, maybe you have some list of questions. But, you know, if I was to interview someone else, it would be slightly different. There is not like a very clear rubric. I just make notes and I walk away and I kind of form like a holistic opinion. A crude summary but it's kind of how it works. A structured interview um you know pre-AI, let's say, or pre, pre-technology or digital technology it's like I start with a set of competencies uh I then have a list of questions that are you behavioral or situational. There's a clear rubric around what good and poor looks like from a response, and I ask each candidate the same questions. And if I can, there's someone else doing the same thing with me. And then what I'm able to do is assess everyone against the same set of questions and I'm able to kind of have a degree or rate of reliability because someone else is doing it with me. Now, just me describing it kind of tells you how expensive and time-consuming the whole thing is. Makes sense maybe if you're thinking about your senior leaders, but as a graduate, you probably, how much time do you want to invest in making that hiring decision right?- Well, and you're really, you're describing best practices for top-notch qualitative research.- Yes. Yeah, exactly.- That's what you're describing.- And you're gathering like a... then you're turning that into some form of like empirical data because you have like rating scales for each of your questions. Now, how does AI and technology come into all of this? Well, one, because it is highly standardized. Okay, does that lead itself to being automated? And then two, like how do you then productize that? Well, a company called Hirevue was like maybe possibly the first to kind of create the concept of like digital interviews whereby you know you go on to a Zoom call whatever there wouldn't be necessarily a human on the other side of it, but there would be a question you'd get recorded saying your response and then you move on to the next. And behind the scenes, at least as far as I'm aware, like they are then scoring those responses. I think it's like either a human can do it or it's done algorithmically. And then the machine learning part of it can work on that. Again, I don't work for Hirevue. This is just from how I understand it. So I could be wrong. What that setup enables you to do is basically deliver a structured interview across hundreds, if not thousands of people at a fraction of the time and cost. So the idea being you're actually having or you're getting much stronger kind of candidates to go through because you've already been able to screen them against like the same set of competencies. And also they've gone through the exact same standardized experience, right, so it's very fair. Now, there's some problems with digital interviews per se because, you know, what other data is being kind of used beyond kind of just the things you say? Are they looking at prosody or facial expressions and so on and so there's some like ethical concerns there that need to be managed. But this is kind of like then well how do you then translate that to an even a more efficient form of uh measuring candidates and you know psychometric assessments are kind of the next equivalent of that, right? So like if you're asking people to respond to the same set of questions well why do I need to kind of sit down and talk to you about it, you know, I can do that via a survey or some other tool that can be easily analyzed. Now, the idea of interviewing yourself uh feels kind of a little like a dystopian, um and it is a bit weird for sure and uh you know you... there is something to be said about, you know, having the face-to-face time to kind of use some of the intuition to kind of see, you know, how well this person shows up, so on and so forth. And so that's why I said, you know, it's about getting the balance right between highly standardized, automated processes to counter your bias, but also, you know, you want to make sure that this person is going to be a good team fit or a culture fit, so on and so forth. But the idea being, the more standardized you can make your process, the higher quality the data, the fairer it's going to be, which is all going to make the use or application of AI much more accurate. And it then becomes much easier to detect where bias may be creeping in, and then you can do something about it. That goes back to something I said earlier around humans being kind of maybe more biased than AI. AI is just a reflection of human bias in that respect, The good news is you can train and, sorry, you can study the outputs of algorithms very transparently and then start to see okay well there's disparate impact or adverse impact for certain demographic groups, what is causing that, you can then look at kind of the weightings of the different variables and then start to control or remove those factors. Unlike humans, which, probably harder to train to become more inclusive, more open, less intuitive. AI, you can at least do something about it. You can manage it and control for those kind of damaging effects more efficiently.- Reece, so many things for us to keep in mind as we want to try to hire right, and do it in an ethical, helpful, positive way. Good candidate experience, good employer experience. So final question, kind of a Lightning Round question for you, Reece. What's the one thing that leaders need to keep in mind to best optimize AI as they look for and hire great talent? That's a good question. There is no one thing. There is no one thing. I'm sorry to be a classic psychologist, which is, well, it depends, or it's like giving you a non-answer. If you are looking to improve your hiring practice, it's about, one, trusting the science, what actually predicts performance. And I mentioned a little bit earlier what those attributes are. And then two, making sure whatever tool you're using, whether it's AI or non-AI, is clearly and empirically related to those three attributes. If you don't get that bit right, it doesn't matter what tool, technique, or methodology you're using. If you're measuring the wrong things, which is often the case, it's all going to go wrong. So please get clear around the critical competencies and then use that to inform your use of tools and methods.- Excellent. Reece, thank you. Thank you for being part of Leadership Growth Podcast today.- Thank you, Daniel. Thank you, Peter. It's been a real pleasure to have this conversation. Fantastic. And listeners, thanks for joining us. And please, like and subscribe in the future. Join us on other episodes as we dive into tools and ideas to help all of us become more effective as we lead others. Take care, everyone, and see you next time. If you liked this episode, please share it with a friend or colleague, or better yet, leave a review to help other listeners find our show. And remember to subscribe so you never miss an episode. For more great content or to learn more about how Stewart Leadership can help you grow ability to lead effectively, please visit stewartleadership.com.