Episode 74

Rowan Jacobson on Transforming the Temporary Labor Market with Job and Talent's AI-Driven Approach

TIM: We are live on the objective hiring show with Rowan. Rowan, welcome, and thank you so much for joining us.

ROWAN: Thank you for having me. It's great to be here.

TIM: It's absolutely our pleasure to have you, and I'm really pumped to have this conversation, and I think a nice place to start would be just if you could give a brief introduction about yourself and the work you're doing at Job and Talent.

ROWAN: Sure, so I've been in the data world probably for around 15 years in a variety of different roles and different types of companies. I've now been from the small series A all the way up to listed companies and beyond, and now I'm at a company called Job&Talent, and Job&Talent is a marketplace focused on reinventing the way that we drive temporary labor to clients across the world. So we're doing this with our technology in two core areas: in recruitment, where we use our data to efficiently match people to the skill sets and job descriptions our clients need, and then we manage that labor so it becomes more reliable, efficient, and productive for our clients and also ensures our workers get more reliable employment. So we're doing this at scale, and we're operating in 11 countries globally, and every year we're placing around half a million workers in employment, so that's quite a good scale. I'm the VP of data at Job&Talent, so I'm looking after everything to do with the data pipelines, the reporting, and providing self-service analytics throughout the organization and also some data science and AI. Yes, I'm doing quite a bit.

TIM: Thanks for that introduction. I think, given the general topic of our podcast about objective hiring, it would be great to delve into a bit more detail about jobs and talent and how that product works. You mentioned a fairly staggering number of placements—that's a real product market fit achieved at the scale that you guys have gotten. so I'd love to hear more about the kind of specifics of the types of roles That you guys work and how that all works

ROWAN: That's a good question. Are we operating in the temporary workspace or essential workspace? Think about it as shift workers, for example. Our clients could be Amazon or DHL, and they're looking to fill up a bunch of shifts in a certain warehouse—warehouse packers, DHL drivers, et cetera. and they are looking to fill those shifts regularly and with workers that will not only show up to the jobs but also do their best in the particular employment, and our workers are also on the other side of that equation, also looking to get more regular employment, and I think what's really interesting is a great opportunity for us to grow. We've been growing in, like, many of our countries; the U.S. is our biggest market, but we are headquartered in Spain, and we also operate, like I said, in 11 countries, so we're in countries like the UK and France and the Nordics and the like, and I think what's important to think about is we're only just getting started, right? like we got a couple of products out there to help us build out these efficiencies and recruitment of our workers, but then also on top of that we can also offer more services to our temporary workers, so as an example, offering them banking services and lending services that today they struggle to access. So the opportunity is very big, and like I said, we are only just getting started.

TIM: That's really exciting, and it sounds like then the way you guys have productized hiring, there's probably going to be then some lessons for people listening because I feel like the traditional way hiring is done for professional roles is very tediously manual, and you guys, for the temporary workers, you've thought through a lot of this problem. I'd love to hear more about the mechanics of how you've productized it.

ROWAN: I think there's a big difference between almost the professional or white-collar jobs that you see on LinkedIn versus the kind of recruitment that we do. Like with us, what we've got to do is make sure that within every geo-jurisdiction and geography that there are specific laws that we need to meet. We need to also understand what the job description is and what the requirements for the job description are and then go through that hiring process, and that's all done within our recruitment app, and being able to find the right workers, being able to find workers that are good at that, are going to be good in the particular work that they want to do is not always that easy because you've got different verticals, so one contract you might apply for might be for a delivery driver, and then you might also be doing another contract as in a warehouse somewhere, and so there's not a clear—it's not so easy to understand, like, what makes a good worker. And so we've built, obviously, our main recruitment app, but we've also got another product that allows our shift supervisors to be able to understand who's arriving, who's going to come to the shifts, who's arriving, who's working, and then allow them to rate them, and in this way try to improve the efficiency of the shifts and the staff. for our clients, and in that way we can grow, and we can grow our share of the market within our existing client base and then grow our client base further, so we're finding the traction is really good, as I've said before, and also it's really a market that is really undigitized, and so we're probably the largest player working through the full digitization of all these different functions. And yeah, we're going strong and building up quite a good base of workers, like I said, which is key to other different types of products that we can offer.

TIM: I'm really interested in the comparison between these types of roles versus the, as you said, the white-collar professional roles you might find on LinkedIn, so how is it that you evaluate someone's skills or capabilities or validate their ability to do a physical, real-world job in a matching process that takes place?

ROWAN: First of all, obviously we are matching up these workers, like we are messaging them, and we are telling them, Okay, here are particular jobs. Do you want to apply? This is what's needed, right? So you obviously need the workers to obviously see that there's also quite a lot of touchpoints along that journey. So, like you, you want to be able to then, if able, call the worker or message the worker and confirm that this is really what they want, this is the hourly rate, this is the type of work; are they up for it? And then they have to submit their documentation, which needs to go through automated checks, etc. And then I think some of the key things to think about are that distance from the workplace actually plays quite a big role, and we need to make sure that the workers actually do get to their assigned workplace, I guess an hour before the shift starts; usually the shift supervisor doesn't really know who's going to arrive and if the shift's going to be filled up right. So, like, you're hosting a big event, and you need a certain amount of staff, and you just don't know if they're going to arrive, so the more information that we can get, the more touch points we can get with the worker, asking them, Are you going to attend? confirming with them, telling them to start traveling now. is really critical, and when they get to the shifts, then they start, and then they need to clock in right, and clocking in is directly linked to the hours that they work and the pay that they receive, and then obviously we start building up more of a detailed understanding of the workers once they start working with us. So at the beginning, it's less about the workers, but once they've done some work for some of your clients, then you know more about them, and you know how they're rated, some of their strengths and their weaknesses, and the types of work that they are doing, and so you start building up quite a big history and understanding of the workers. Now, because we're operating over quite a large number of clients and we've got quite a big worker base, you can think of this as how does temporary work become more permanent work so we're able to place those temporary workers in future contracts if they perform well, right? And so they start becoming employed more, earning more, and on the flip side, the client starts seeing an improvement in the efficiencies; they don't have to hire or overhire too much because now they know exactly who's coming. They know the workers ratings, so they know that these workers will be good shifts. They're able to understand for themselves through the data in the apps which workers are performing well and which are not. Do they have to have a chat with those workers that are not performing well, and how do they maximize their efficiency? So there's a lot of, like, sort of data analysis that goes along, but there's also a lot of reach out to the workers, and you need to do this at scale. So that's what makes this particularly interesting, and you need to probably what's quite interesting about this is that it's, like I said before, really we're coming from an undigitized environment, and we want to automate everything, and everything starts becoming digitized. And so some steps are very hard to automate, and I think with the advent of AI, there are more possibilities of automating things that you never thought you would be able to automate. So in other words, things like doing calls with calling up work and asking them to confirm their attendance or if they want to accept this role, you can start automating and thinking about how you can do this on a large scale, so there are really interesting things you can do with the technology, but like obviously the workers and the clients are the two sides of the marketplace that we have to think about.

TIM: That is so interesting, and immediately my mind's just trying to think of comparisons between what you've just described, that marketplace, and, let's say, the marketplace for data or professional hires and the differences in the process and what have you. What you've just described is light years ahead in terms of sophistication compared to the way that most companies would hire for data people because the whole process is digitized and and you've taken control of almost everything end to end so you can measure everything. You're not beholden to the way, let's say, a traditional recruiter would almost send their candidate to the company, and then who knows what happens after that? Like, they can only control so much of it. Are there any learnings you've taken from seeing this productization that you've then informed your own hiring for data professionals into your team, or is it just fundamentally different?

ROWAN: I think it is different, and I think some of the ways it's different are The CV can look quite good for a data professional, but you just really never know until that person joins, and I think there are some key things that I really look for, and they are their technical skills, the softer skills, and their personality, and these things have formed a big part of how successful somebody is going to be in a particular organization. and honestly it's not only about how good the person is or the person's character traits are, but it's how those match up to what happens to the culture of the organization and how the organization actually thinks of data, uses data, and how their teams are structured, which I always find like I've been doing this career for quite a while now, and I've been in various different companies, but each one is different, and so some people that you would hire for one type of organization you might not hire for another type of organization just because it works very differently, and so even from my position, coming in and having like a clear sort of bible of how I would do things doesn't always work because you have to adapt to how things work, but I think clearly there are things we could do better in terms of filtering out CVs. You put one role out, especially last year, and you'd get like flooded CVs, right? 95 percent of them, at least, are probably not relevant. The skill set's not relevant, and you can see immediately, like, it probably is not going to be a good fit. But then how do you weed out the other 5%? And so for that, I think there's no better way in this market that we are in to meet the person, to have really thorough, deep discussions, and to give them some real-world problems to see how they will react. We can go more into that, but I think personality and the way that they work is as crucial as technical skills.

TIM: I'm again interested in the parallel comparison between these two markets. Is it the case that fundamentally you're trying to hire, let's say, a data scientist? There's just a lot of skills, if you think about it, that they need to have the soft skills and technical skills to fit in with the business culture. Is that what makes it a fundamentally more difficult matching evaluation process compared to, let's say, placing someone into a warehouse to do delivery work or something? just, there's more data points you

ROWAN: I think because, like, your head counts are quite tight and they have been, like, for probably the last couple of years, so when you get the opportunity to hire, you want to make sure that hire is going to be there for a long time, and you want to make sure it's the best fit, and so when you're thinking about the other side, like, so you've got a short-term contract, and if it doesn't work, that person can go and apply for another role, you know, relatively quickly. They might be doing two roles at the same time, but, you know, in essence, you want You know you want to know what the definition of what a good data scientist is probably like; it's quite different from what the definition of a good warehouse worker is. And that's where the complexity lies, right? Also, I don't think we're just getting started on the journey of rating and digging deeper into that in terms of temporary workers, but I think even on the more permanent side of workers rating workers, or you do your performance review. your annual and semiannual performance review I think there's a lot of look; there's a lot of ways to do it, and I don't think there's every way any way is perfect, and I think it's still hard to get down to the meat and potatoes of how well all this person is performing and how good a match it is. Sometimes you hire somebody, and they're just a perfect match. and you can sense that, and before they start, but you never know until they've been there for six months; even in the data world, it takes like what, at least three months to get embedded, probably six months. You have to learn the tooling; you have to learn the data; you then have to learn the personalities and ways of working in the business. It's a much slower ramp-up, and you, as a hiring manager, want to be there; you want to be ahead of the curve, and you want to know quite early on if this is going to work or not, and I think that's the challenge; sometimes you just don't know until it's way down the line, and you have to start again. but other times you know quite quickly. Hopefully that answers your question.

TIM: Yeah, it does, and I'm just thinking then about, so it's almost like you're making a permanent hire; the salary is obviously higher, skill sets are more specific, and maybe higher as well, but then there's a level of risk aversion there, which is if you make a bad hire, that's going to be quite a complex thing to unwind as opposed to hiring someone on a one-day shift for an event. Okay, what's the worst that could happen? Then one of 50 people doing the role, so it's like a lot less; there's more risk tolerance in making a bad hire in a sense.

ROWAN: Yeah, and also I think that I think the environment can change, like you hire somebody to do one role, and then that can change over time, right? Like your strategy changes, the business changes, and you never know how that person will adapt, and so that also adds to the complexity, right? that's It's It's quite a complex situation, and I think also in the data world there are so many tools and technologies out there, and data scientists are going in so many different directions, and understanding exactly what you need and five years time is very hard to know, so you're almost hiring somebody for their flexibility and their openness, and the deep core-like knowledge base, I think, is what I try to do.

TIM: Yeah, I think you're right, and that's what I'm hearing more and more. That must be the case if you're in this incredible, almost breathtaking technological change moment in time, because the tooling is going to be chalk and cheese in five years, so to hire someone fixated on, do you know, XYZ tools is probably very shortsighted. You just need that flexibility.

ROWAN: the I also think about the roles, to be honest with you, like what will a data science role look like in five years time? What will a data engineering role look like in five years time? Right? I think that's where adaptability really comes into play, right? And I think the tools change all the time. I remember maybe it was during my age, like five to ten years ago, you know, you're working on large Hadoop clusters, and things are just like starting off right, and now everybody's talking about AI and the tooling around that, and things have changed a lot, right? So I don't think you can ever be static, and I don't think you could hire somebody who only has a very narrow focus, and if you have a specific need for that, I really try not to hire for a very narrow focus.

TIM: And do you know or have you developed any ways to get a sense of that in an interview, a sense of someone's adaptability? willingness to change, willingness to learn, growth mindset—call it what you like, but how do you actually get to the core of that?

ROWAN: This is the hard part. I think, like technical skills, you can always dig into—you can have tests, you can ask questions, you can even see the kind of work that they've done and the tooling that they've done and that they've used and everything around that—but understanding The softer skills are all the other dimensions that are really hard. So I usually find things like situational questions, or maybe I read somewhere that, like, maybe it was Harvard Business Review, these are the best predictors of future performance, so give me a situation when you did X and Y, but I do find, like, it does help, but it doesn't always go the full mile. So there's always going to be a gray area, and then you have to maybe then you—that's where bias comes in because you feel like, Oh, okay, maybe this person answered in a better way than that person, then how do you get such an—how do you transfer that or change that into an unbiased approach? It is very difficult, right? I think this is where things become a grey area also because there's the other side: you've got you're also working with constraints, right? You might need to fill a position quickly, or you might have a certain budget, and you might not be able to get what you need at that budget. So, like, how do you compromise? What are you willing to compromise on? And it becomes And almost an optimization problem we try to maximize what you can get, but it's not necessary; it doesn't necessarily always work, and then usually the person is not a bad hire, but you both have to adapt to the new situation.

TIM: yeah it's such a game of trade offs isn't it hiring like you could have You could say, Oh, let's hire as quickly as possible. Okay, pick the first CV and give them an offer, but that's going to be dreadfully inaccurate, or you could say, Let's make it as accurate as possible, but then you'd have six interviews, and it would go on forever, and candidates would drop out, and they'd get annoyed. it's too expensive and

ROWAN: And what you're finding also it It's who's the power who's got the power here, right? There's a time when the hiring company has the power, and they can set five interviews plus a large take-home assignment and other ones, and then you get a time when the candidates have three competing offers. You've got two days, and you have to decide: Do I take the risk and give them this take-home assignment? or Do I just say, Okay, it's okay, so that's judgment; that's also based on who else is in the pipeline"? You also like wondering about who else you've met; maybe you've met somebody for an initial interview, and they've been delayed, and you're trying to slow down the guy who's really at their last round just so the other person can catch up, and it just becomes this cat-and-mouse game, and yeah, it's quite interesting; it's interesting, and it can be frustrating, right? like it'd be super frustrating, so very frustrating. On the other side, we've all been on the other side where you're applying for jobs, but it's also frustrating for the hiring side, right? On the company side, I think it's not an optimal process whatsoever.

TIM: It's certainly not at the moment. I personally feel like even at the current state of large language models, they could help quite a lot in several stages of the process. What about you? Do you see some benefits to using AI? Have you started to dabble at all in using AI in your own Hiring: Have you seen candidates use it?

ROWAN: I'm still not convinced, right, because I still feel like I think they're always yes, you could automate some processes with AI, right? Like you could go through a CV and immediately just decide, Okay, this is out, and this is in, right? Or like you've got your internal recruitment team. and what normally would happen is you've give them like maybe some like filter questions to ask the candidates right and you try to show them what a good CV is versus a bad CV or what do you think would fit better or not and so you could definitely automate that quite well I think with AI so let's say you are a reasonably small Startup you probably don't want to hire somebody who's been at IBM for the last 25 years right so like these these You know You can immediately remove those candidates and maybe you could have done them with normal ATS with like traditional ATS systems although I've never really seen that automated filtering work very well up until now but I do think that AI can do that How can AI really help with the rest of it? Of course, like setting questions, you can always help your efficiency in interview questions; you can get some really good questions, but how do you evaluate people's answers and evaluate them against this sometimes blurred idea of what a good hire is? It's hard. Yeah, yeah, that's what I always think to myself: What is that secret sauce? What do you mean? Somebody, you go, Wow, this guy's going to be great. Like, I remember in one of my roles I hired somebody, and I knew he was going to be good, and it turned out to be quite good, and what was it? It was yes. Okay, he had the technical side, but he also had, like, sort of communication; he had an understanding for the business, and he knew that he would fit in culturally, so how do you nail that down? Even just asking that gets an AI to do that for you, I think, is difficult.

TIM: Yeah, if it's so almost intuitive that it's hard to put into words or codify, then yeah, I can't imagine that an AI is going to help us with that anytime soon. I feel like one big value-add AI could have large language models in particular that are even just almost like an interview assistant. You mentioned creating questions. What about creating a marking guideline for questions? What about doing the transcription of the interview? What about preparing the candidate feedback? Like all those things, I feel like at the moment it's just an absolute schlep for someone to do manually, which is probably part of the reason why candidates don't end up getting feedback.

ROWAN: I agree; I think the feedback cycle can be very slow at times, and it's also quite annoying because usually I have to say I'm just typing on the other screen because I'm taking notes, and just ignore me if I'm looking down that direction, and you know So I think those things can definitely help, although my experience so far is that the transcription isn't always fantastic, right? So you always need to go back and resolve it, so I think we were not completely there. I guess also the ad—I mean, if hiring managers can use it or people doing the interview—then the other side can also use them, right? And so that's where it also becomes quite difficult to understand if that's good or bad, right? I want to get to understand the person for who they are, not for how quickly they can or which what if they have an assistant to help them answer a question, so both sides I'm doing it for my efficiency, but I might also be doing it to come up with nice, interesting questions, and then they are doing it to get nice, interesting answers, and there you go. the cycle starts

TIM: Yeah, and then the next step will be to cut out the hiring manager and cut out the candidate. The AI can interview the AI, and then they don't need us anymore.

ROWAN: No, then I'm fine to collect my universal basic earning income and just sit on the beach. I'm fine with that. No, but yeah, I think sometimes we do, so obviously we would do some technical tests and stuff like that, and so then I'm reviewing it with maybe my team members who have sat through that technical test, and they might say, I'm not too sure about this or that. and I and my answer sometimes is if they can't do this technical thing, we'll know in the first week, and then that's not going to work out, so I think there's limited downside, but we're back to what I think is hard: finding the right personality, the right kind of curiosity, and the right kind of ownership of people is really hard because then if somebody is going to use ChatGPT or whatever it might be to come up with an answer in a star-based answer system of a certain situation they've been in but haven't, that will give them everything that you might want to know as a—or might want to hear as a hiring manager. They took ownership, they investigated, they communicated, and then they started their role, and then something like that comes up, and they just hold their hands up, right? And so you haven't figured that out; you haven't got that. I think this is where it's really hard, so it's very difficult.

TIM: I agree, and I feel like that almost highlights one of the obvious fundamental issues with hiring, which is that, certainly at the CV stage and that kind of behavioral interview, you're taking the candidate's word for it. It's a lot of talking about, in this situation, I did this; here's an example of that. and it's quite geared towards really well-prepared coach candidates who've already thought this through, thought forward, sorry, come up with all these examples, and they sound confident; they sound like they know what they're talking about, but it's so easy to be misled that they might not actually have any clue at all.

ROWAN: I would say, yeah, they are. I would say at least 60 percent of the candidates who believe that they're using it and they're talking with knowledge and with experience, you could probably tell straight away, right? And that's where you need to just ask more and more questions. more and more detailed questions and that's when it starts probably unraveling, but I think that's becoming like a more and more gray area, and it's becoming harder to understand, but yeah, I think it's an interesting time for all of us.

TIM: Yeah, I think that is still the key. I interviewed someone recently, Charles Shaw, and he was describing his interview method, where he basically asks only a couple of questions, but he tries to get down to what he calls level three thinking, so he'll ask and get a sort of superficial answer first and then keep digging down and down until he's almost exhausted his knowledge of the topic. and by doing it that way, a candidate who's got this vague idea or has got a ChatGPT answer, as you say, can't really stand up to that scrutiny, so it's really about just digging to find

ROWAN: Yeah, yeah, exactly. I've seen that work definitely, but then also the interview becomes very focused on one thing that can take 25 minutes, right? And you know, in the greater scheme of things, it might not be the total of the person's experience. and so you've got to be careful which part you dig into and how you dig into it.

TIM: Yeah, if I think back to the last time I was interviewing analysts, one question I had for them was like, Oh, I want to discuss in real detail a meaty bit of analytics or a project or a model you've built. We're going to spend 25 or 30 minutes going into this, so choose something that is recent in your memory that you can discuss in detail that you know we're going to get into. I've tried to do it that way so it primed them to know what to expect. They chose a good example; they didn't choose something from 10 years ago that they barely understood, and so then I tried to just dig as far as I could possibly go just to see what they'd actually contributed to avoid the candidates who might say, Oh, I did this model, but it was like they were just a bit of a massive team who completed it, and they weren't really that heavily involved. But yeah, I guess doing it that way you might get this myopic view of one particular example they've had, but maybe it misses a broader picture.

ROWAN: Yeah, that's what I would say. That's what I think, and also, like, for the more senior candidates, I think that's also where it becomes complex because you're managing a team of people; perhaps you're looking for an analytics manager or something like that, and what part of That role that you're trying to fill, are you most interested in it because zooming into the top, that type of analysis might not show their skill sets in a cross-departmental communication setting? Managing the team, like really, which skill set are you actually interested in looking at? and I think that sometimes, like I've seen people focus on one thing and ask, Okay, so what was the data set like, and how did you do this, and how did you do that? and then actually the hardest part of the future role is going to be something different, and they haven't touched on it because you've run out of time.

TIM: Yeah, I often feel that hiring processes can get derailed quite quickly, partly because I think the hiring process isn't really perfectly productized at the moment. Companies will use an ATS, but then people will come up with their own questions; they might write notes in some other system.

ROWAN: Oh yeah.

TIM: It's disconnected along the process. It's not like it starts with, Here's what we're hiring; here's the skills; therefore, here's the entire hiring process mapped out in a system like it. doesn't work like

ROWAN: Maybe I'm a little bit guilty of that myself because then I do read the notes of the people that have interviewed the candidate before my interview, but I really try to have an unbiased view, so I don't really want to focus on something that somebody else has seen. I want to really understand the person for myself, but perhaps that's just part of me also going off piste and doing my own thing, which I probably should pull back on, but we're all guilty of it, right?

TIM: And it's a tricky trade-off, I think, because when I hear people describe the hiring processes where they might have, let's say, several, normally several interviews with different interviewers, there's some sense of like a wisdom of the crowds, or one person might notice something that someone else doesn't. and you're getting like a more holistic picture, especially if you have a talent interview, hiring manager interview, or other stakeholder; then you've got quite a broad view of people's opinions, but then at some point, noise is introduced to the process. Is there a sense that the fourth or fifth interview might be looking for something that the first four weren't, like they just got their own idea of what you should hire?

ROWAN: Yeah, and how many of these people is that going to or are they going to meet, right? So you're like two months into the hiring process, and they still have to meet like the fifth business representative, and yeah, it becomes a little bit too much, so I agree, like everybody's got their own things they want to focus on. and then you're looking for these perfect candidates, and you've got to compromise, as we've said before.

TIM: I think it probably works okay as long as everyone's on the same page, so they've got okay, every single person knows the criteria for the role; they're not inventing their own criteria; maybe they don't even have to come up with their own interview plan. Here you go. This is what you should be evaluating.

ROWAN: But that comes with an overhead, right? You hear people that have been interviewing, and they're waiting three, four weeks for feedback because they can't get everybody in the same room to have a conversation and match up their experience right, and it just drags on so much that I think you'd lose the effectiveness there. So yes, if hiring can be prioritized and time is made for hiring and for the pre-work and the post-work of the hiring, then I think that would be an ideal situation, but we don't live in an ideal world, so you again, you're trying to balance, right? You're trying to balance how much time do I put in versus trying to maximize the best candidate that I can get out.

TIM: I feel like going one step up, then one of the meta problems I would see in hiring is that a lot of this stuff isn't really measured. You could imagine, in theory, a system that had all of this mapped out to it, and you're like, Okay, cool, you're going to have four interviews; that means, on average, you're going to have X amount of days waiting between interviews. So it's going to blow out the time to hire by Y. Did you think of that trade-off? Like, you could reduce the interview by one step.

ROWAN: That's a whole sort of area of analytics, right? And you could definitely start looking at that, but what I usually find is like you just don't have time to do that, right? Like you don't have the resources to do that; there's always something more important to do. But I do know some companies do that, and I think that's where maybe your writing tool can really help really start analyzing that time in between hires. Also, I find the time between hires and the time between interviews is super important, right? It's been like a month since your last interview, and you just never know; the camera just is not going to be motivated to perform very well. And the interviewer is not going to be able to really focus because they know that this is going to be another three months before you have that final discussion and correlate everything together. I can see this could be really useful in tooling; it could do that.

TIM: Yeah, I think that's what has to happen: the actual HR products have to build this in, and then they have to be used in the right way. Like, I can think of a few companies I know who've built their own little systems. Agoda, the travel company, I know their marketing team because they hire at such scale, and because they're marketers, and they think of everything like a marketing funnel, which hiring is just like a funnel like anything else. They then went to the effort to start measuring things like, Oh, which of our directors are actually doing interviews because they're meant to share it around among the directors so they have, like, just a leaderboard of that? Then they have a sense of who's the best gate because if you're an interviewer and you keep passing everyone through and then in the subsequent step they all fail, then there's something wrong with your

ROWAN: Point yes, yes. I think that's very useful. that's very useful, yeah, for sure I can see how That can be like defining the time and the resources to do that is hard.

TIM: Exactly, and it's probably unless you're hiring at scale, maybe not worth it compared to all the other things you have to do day to day on the job, but yeah, hopefully then that makes sense for some general HR tech providers to, at some point, solve this. And then spread that I think every customer

ROWAN: Yes, definitely.

TIM: You mentioned in passing before that in this market, in this current market, having that kind of live interview where you're getting a sense of the candidate's skills, the technical skills, the soft skills, and their fit is that partly because now the CVs may be even less predictive than they were before because maybe candidates are augmenting it with ChatGPT? Is it also because the take-home test maybe someone's using ChatGPT, so it's harder to get a lock on the real

ROWAN: Yeah.

TIM: than in real time is that part of the

ROWAN: I think so, but also I've always wondered about a take-home test, and I think in the past when you've had the ability to give take-home tests and the market wasn't so competitive, I think it was interesting, but I'd also like to see how people perform, like, in the job, and yeah, sometimes take-home tests are literally now if you really wanted to, you could, like, probably—I haven't done this, but you could probably put the question into ChatGPT or whatever and ask it to do the Python code for you and come up with some analysis. I wonder where those sorts of technical skills will eventually be tested, and again, what does that even mean because, like You could use a co-pilot to help you do stuff right, like you don't need to know everything; that's where there's like this balance between doing things live and getting the person to know it and expecting the person to know every single thing versus giving them leeway to investigate and do more things. And so probably that's that. Doing things, doing live business cases, or investing, or digging into situations, I think, is where that is really useful because you don't want somebody writing up how they did something in the past; you want them to think on their feet, and yeah, coding I think should probably be done in person live but to a brief level where they don't have to go into as much detail, yeah, so that you can really see where they are, and then at some point you need to do a very deep dive into how good they really are if it's a very technical role. but that can be done offline, and you have to take the risk. Yeah, it's making things really complex, right?

TIM: It is, I feel like it's, I think it's because it's because LLMs have come to product market fit so quickly, and because candidates can pick them up like that, because they're just an individual company, can't adopt the equivalent tools as quickly because there's AI laws, there's people data laws, there's just companies that are going to make decisions a bit more carefully than candidates. So the candidates seem to be ahead of the game, but then it's just several steps of the process because of that, and so we're in this weird in-between period, I feel.

ROWAN: Yeah, and to be honest, I'm not really thinking on my day-to-day how the candidates are trying to use AI to get through interviews, right? Like, I've got my day job, and the hiring is part of my day job, but it's not my super focus, whereas for the candidates, that is their focus. They're going to be working at night to try and find that extra edge, and I think that's maybe where the imbalance lies.

TIM: And I also feel like part of the difficulty is we're in this weird point where, let's just take, I don't know, SQL as an example, so you could get ChatGPT or Claude to write your SQL to solve a problem, and the accuracy now seems pretty good, but you're still going to have to actually use it in the real world and not make disastrous mistakes. You're still going to have to understand SQL, maybe not as well as you would have a few years ago, but you still have to know it, so evaluating the candidate SQL skills still seems relevant to me, but then we could be a year or two down the track; maybe that will no longer be the case; maybe no one will be writing code.

ROWAN: Exactly.

TIM: I don't know in which case, then what, yeah, what are we evaluating, then what is maybe that's a good way to ask it like you were talking

ROWAN: But maybe the roles are changing right Yeah, the roles I think maybe are changing at some point, right? And they haven't changed yet fully. Some companies are more ahead than other companies, and then changing their strategy and changing the way people work, but yeah, you're 100 percent right. I think it's really hard to know where things are going, and actually, to be a junior trying to start the career, I think right now it must be pretty interesting. Let's call it like that: where do you start? And it's not only—I would say it's not limited to data; all industries—like I was speaking to a lawyer friend of mine the other day, and she said, like, the role of their juniors in the firm has changed dramatically, and it's continuing; like, the pace of change there is just huge, right? and so like you can imagine, like everywhere it's difficult, and so I think as a hiring manager, you just got to really understand what do you need now, what might you need in a year's time or two years time, but you can't really overthink it; you can't really overthink it, but again, you need somebody who's got the right attitude and the right frame of mind, the right curiosity, the right way of talking and communicating to be able to hopefully bridge those gaps when you get there.

TIM: The way you've described that, then you would maybe be implying that if you imagine, like, the blend of technical versus soft skills that someone as a data professional needs, maybe that blend is changing where now it's going to be more the softer skills, more the fit, more And slightly

ROWAN: I think I must, as you say this, I'm thinking we mustn't overegg it because still right now there's a lot of technical skills needed, and like most organizations, the data is really dirty, and like these AI models are not going to be able to help you much because, like, even you or even a person in the role doesn't really understand, like, you know what's going on in this table and why this has changed. Nobody, you know, so I think there's a lot of ambiguity and a lot of, like, complexity, and especially, like, established businesses and older businesses that probably the data is really dirty and really a mess. So I think, yeah, we really mustn't overegg it, but I think it's something to keep in mind, and I think most businesses I would be surprised if this is not the case, but most businesses are thinking about, like, how do I improve the efficiency of my team? co-pilots automating documentation and so this is step by step going in one direction

TIM: Maybe there's also something to be said for the fact that, like, what's the total possible demand for analytical skills in a business? Maybe it'd be like if you're making every single decision based on data in the entire company, and we are nowhere near even 1 percent of that at the moment, even just hiring, as an example, most of it's not started-driven at all. That's just one bit of a business, so maybe if we can just make the data analysts or scientists a lot more efficient, then we can just analyze more stuff than we were before because we're not covering even

ROWAN: Yeah, but I think also, like, even before, I would say there was a big overhiring of data, like we've seen in the last couple of years, that, like, data was hit quite hard in, like, the change in the tech world, right? And I think this question is not only relevant just for AI, but I think in general data and tech, there's, like, now a flood of people who've got data science degrees or training to some level, whichever, like, where along that spectrum that lies, right? But there are a lot of people in the market, and companies have hired a lot, and also the stacks have become quite expensive. Right? This modern data stack became very expensive. You have to pay for every single layer, and it becomes very hard to justify why you need every layer and why you are paying so much. When in the end the data is still so messy that your first thing is like, How do I just get the numbers out and forget about all this other fancy stuff? I just need to do the basics right, and it's hard, and then that's when the data functions got hit really hard, I think the story is difficult. It's not an easy story, and this AI stuff coming in has changed some roles within data science. What is an AI engineer? I didn't even know what that was. I haven't really seen a definition of that. I just saw some companies are hiring for that, but what does all of this mean? and I think things are changing right, and it's not really what it used to be, and you've got to adapt, and you've got to just try to figure out, like, first, what do you need, and then try to think about where things are going in the next few years; you are hiring people that can adapt for that.

TIM: Adaptability then maybe should be the buzzword or the thing we're all looking for at

ROWAN: Resilience, adaptability, and yeah, I think that's what you—that's what I would advise.

TIM: Rowan, one final question. If you could ask our next guest one question, what question would that be?

ROWAN: Yeah, that's a good question, and I think the standard question would be where do you think that the roles will be going in the future, but I'd probably say I'd love to know people like the biggest mistake that person has made in hiring and the best or the opposite of the biggest mistake. What did they do extremely well? Because I think trying to learn from people's mistakes and what they've done well is something we don't generally talk about; like we generally talk about processes and how things can be improved, and normally people don't like to talk about the mistakes, but I'd love to know.

TIM: excellent I'll level that question at our next guest with a bit of preparation time for them to mull over that and meditate on it, and hopefully they'll give an interesting answer. Rowan, thank you so much for joining us. It's been a really interesting, engaging, and enjoyable conversation, and I'm

ROWAN: You.

TIM: I guess our audience has enjoyed it as well.

ROWAN: perfect Thank you very much.