Episode 21

Peter Range on optimizing Data Talent Recruitment and the Future of AI in Analytics

TIM: All right. Peter I'd love to get your thoughts on this. Do you think companies are actually hiring the best data talent, or a lot of the time do they end up just hiring the best person based on their CV?

PETER: I think, as with most things, there's a mix of success in this area, but a lot of companies do get it wrong, and I think you need to have a really clear recruitment strategy, and it's got to start from the top. I think another just footnote here is I think also too many companies expect talent to be a silver bullet. And there might be bigger issues in terms of data literacy or data governance processes that you're trying to patch over with hiring in talent in the wrong places, and these kinds of companies can waste millions without addressing the root cause. And I think for me that's often because your decision-making and technical expertise are potentially sitting in different areas. and again, it is really important to have a data strategy that your executive is bought into when it comes to recruitment, but just to give you an example of what I mean there, you can try and hire in an army of reporting analysts, a data analyst, to solve things that really should have been solved at the engineering point. In terms of when it gets into hiring, for me it just comes down to the test. We've had some amazing talent come through in my history from online data conversion courses; equally, we have interviewed people with master's degrees in data science from really reputable universities who can't do the basics. You can't just read the CVS right; you need to really be part of the making sure that you're putting people through screening processes that can identify talent in hidden places. I mentioned to you before that I was reading this book on data games and how analytics has revolutionized Liverpool, and I think this is one where the moneyball philosophy is really important, right? You've got the kind of quotes about how they're trying; they don't want to hire a certain footballer because they're wearing gloves in the winter, and they're worried about what that says about their character. We can take that kind of into our setting; they don't answer a question about their career history in a totally convincing way or in a way that a salesperson would, right? But are you missing out on hidden value? And the quote there from Moneyball is the inability to envision a certain kind of person doing a certain kind of thing just because you've never seen someone who looks like ends as a market inefficiency, so if you know what stats and metrics really matter to your job and you can measure them, then you can exploit these market inefficiencies and deliver less with more.

TIM: I just remembered this week I don't know if you remember this, but it was not more than 10 or 12 years ago. that some of the top English Premier League players who even at that point were getting 100,000 a week, if they had quite a nasty injury, they didn't necessarily go to the sports scientists in their clubs; they went to a witch doctor in Serbia who used to feed them horse placenta. This is—I'm not making this up. This is what was happening only 10 years ago; of course, now that is unimaginable. I feel like hiring now is like the horse placenta of 2012, okay? And the way that most companies approach it is with as much almost like random magic, and will we get to a point maybe in 10 years where the average company hires the way Liverpool does or the way a baseball team did 20 years ago? What do you think?

PETER: I would hope in the data and analytics industry that we're slightly maybe ahead of Premier League football if only because the most senior people in these areas understand the value more than the kind of football men or women of the past. But I do think, as with football, it starts in one place and it filters out. I think the tech companies are famous for doing this really well, and then it trickles down slowly through some of the more laggard industries in terms of this approach to hiring.

TIM: A lot of hiring processes I've seen in the past have been like, Oh, let's do a vibe check. Let's do a pub test. Take him for a beer. I'll just have a quick coffee with him and have a quick chat. and this is layers of like just pure subjective intuition, gut feelings, and some combination of those things where someone will come out and go Oh yeah, like I didn't really get the sense that they were the right fit for this role. I'm not sure they're a great fit for our team. I don't think they failed the cultural fit test that we like, basically variations on the theme of that, and now maybe the opposite end of the spectrum is, yeah, what I'm guessing is your Amazon or your Facebook or whoever does where they're measuring things at each step. What have you seen in terms of data-driven processes that are actually working? What have you applied?

PETER: Yeah, using the football analogy again, right? What happens if you follow that? Oh, they didn't pass the kind of culture fit test, and you're not willing to test some of these hypotheses out; you're just going to end up overpaying for talent, right? Compared to your competitors, and in the long run, that'll kill you as a business, right? Yeah, the analogy again: if you're unwilling to take a look at a player because they've had an injury in the past, maybe you're missing out on an opportunity, and likewise in data, right? If you're unwilling to look at people because they've had a six-month career break, are you making the most of your budget in terms of—and I think really within that you need to ask yourself, can we measure these things? Has a career break had an impact on our success as a business before? Has coming from a red brick university really made a difference in terms of performance in our teams before? Sometimes the answer might be yes, and these conventional wisdoms exist for a reason; others don't, and that's where you find value.

TIM: And so it's almost, to put it in stock terms, you're finding those positive alpha stocks, the ones whose price is far below what it should be given its fundamental value, and I wonder if also those candidates might be the extra appreciative, like the undervalued candidate is systematically overlooked for whatever reason. imagine giving them an opportunity to see how they could perform

PETER: Oh, absolutely, absolutely. Yeah, I think, yeah, perfect analogy. You're looking for things that aren't priced into the market, right? And that's where you find alpha, and I do think if you are working unicorns, everybody wants a unicorn, right? Of course it's great to have a candidate that has technical excellence, passes the cultural fit, and has great soft skills, but we rarely have the luxury of having all of those things, and when you are focused on all those things, there's such intense competition. Are you going to have the loyalty that you would necessarily have from somebody that you've picked specifically for the job because your recruitment strategy is different, and they may be overlooked at different companies?

TIM: And I wonder if part of the issue is that companies aren't very good at measuring the false negatives in hiring. Like, you really remember the false positives—the person you hired who didn't work out. I don't think anyone forgets that like a disastrous hire who you had to let go, you had to fire, who quit within a week. That's just like a terrible scenario for everyone: the candidate, first and foremost, you as the hiring manager. Like, it really burns in the memory, but the opposite, I think, in hiring isn't that common. Like, you might get 500 applicants; you don't remember 499 CVs that you looked at for five seconds and rejected. you never know that Hey, it turns out I rejected the next president of the United States or whatever; some like up-and-coming genius, whereas in something like football, maybe you do remember that the young guy came for a trial at your club and then turned out to go to your rivals and was worth a hundred million pounds. That is very obvious. I wonder if we almost need to unlock that metric or that data set in hiring to give that feedback to companies to say, Look what you missed out on.

PETER: Yeah, it's an interesting idea. I think it's It's difficult to yeah measure the impact of the ones that got away right and it's easier to identify the cost of the ones that you didn't get right but you still maybe look through your LinkedIn of people that you maybe have connected with in the past and think about wow they're moved on quite considerably in their career and think about how could have I potentially got in earlier because now to hire them would have taken a lot more and yeah I feel left behind obviously challenges is with a lot of this data right with the ones that got away how do you get a shared resource of measurement that other companies are willing to divulge about what their performance was after the fact But if you could do it, it would be really beneficial.

TIM: Yeah, and then I guess again thinking about it, comparing it to football, maybe there are other motivations in companies. If you're a football coach, you need the best players on your team; if not, you're going to be fired. Like, it's a very clear relationship between recruitment and the success of your job. whereas maybe in bigger bureaucratic companies, hiring someone who's amazing could almost show you up Some managers could think that there could be some level of being threatened, and so hiring the best person for the job sometimes for some people might not be actually their best strategy. I would certainly hate to work in an organization like that, but I'm sure they're out there. Have you seen that in the past?

PETER: I think that you're always going to have people that kind of don't realize, or it's easier to carry on with conventional wisdom or continue with the processes that you've always seen or used, but those people are rarely impactful again. Most departments don't have unlimited budgets when it comes to data and analytics, and I think from a leadership perspective, the leaders who are highly coveted are the ones that can do more with less, right? And even in the football example, right, I think one of the reasons why it took so long to take hold in football is because oftentimes the lead times between changing your recruitment strategy and actually seeing an impact within your business are very long. and a lot of the leaders of the type that you're talking about, they're not thinking five-year plan; they're thinking, I'm here for two years, hold the fort down, and I'll be on to the next thing, so long-term results aren't actually what I'm trying to deliver, and you know what? If you have that kind of person, I'm fortunate for you. But they're never going to make that commitment right. The people who are looking long term, that's where you're going to get results, and going back to Liverpool, it didn't happen immediately. They had to make a lot of mistakes. It didn't work out with Damien Camolli, but it was starting a long time ago, and they tried a lot of things, and they're still learning today. But the book itself is extremely I think it's one of the best data strategy books that I've ever read because, right from the top, you have an executive team that, outside of the data science department, realizes that this is the way that we're going to run the company and FSG and Henry, like his background, is in commodities trading. He's a mathematician. He actually, in the book, they talk about the first thing that he tried to do in terms of making money at a university was to count cards in blackjack, so it gives you a sense of the kind of guy that he is and why you put such a large weight on data.

TIM: And I guess, again thinking about football, if Liverpool takes on that strategy, yeah, there's a bit of a lag time between them changing strategy and winning, but ultimately they do win. They win the Champions League whenever that was, four or five years ago; they win a Premier League. So then other teams obviously cotton onto that, and they realize, Hang on, we're going to have to change something about what we're doing to compete with them. Maybe a little bit slower for that to happen in business, but I could certainly see how a brand new startup could come into a new market in a very data-driven way of working. They've got three engineers and ChatGPT, and everything is now just they don't have any legacy systems and ways of working. They just come in from you and start to eat the lunch of companies, I guess like a buy now, pay later kind of business model suddenly disrupting finance and those kinds of things that must happen all the time, and so I imagine we'll start to see that even in hiring as well sometime soon.

PETER: Yeah, absolutely. I think, again, your examples of the smaller companies, where necessity and the lack of resources drive that innovation, are where you're going to see this used to the biggest extent, and again going back to the football analogy, but it's often true that it's not the teams at the top of the league table where this makes the most difference. Because a lot of them have nation-states backing them, and actually where you really see this make a massive difference is in Brighton and Brentford, right where, okay, your resources are seriously constrained, and you do need to, if you can, uncover a top-four player who's hidden somewhere in the championship. Sorry for everybody who doesn't watch football, but yeah, then that's going to really make a massive difference in terms of your performance; you could be a league above where you should be otherwise in terms of your resources.

TIM: And so for companies, then do you think we've gotten a bit drunk on cultural fit in the last 10 years? That's all I've heard for years and years, and still I haven't really seen a measurement of it. It still feels very vibey, and I, as a data person, get quite uncomfortable with things that are very vibey where there's not a number and something can't actually be measured because then it doesn't really feel real to me. What are your thoughts on cultural fit? Do you think it and our obsession with it are harming our ability to actually build good data teams?

PETER: It's a good question, and I think soft skills are always important. Cultural fit is important, but there's some kind of pyramid of importance or hierarchy of importance, and obviously lots of things impact where things sit in that hierarchy for your stakeholder management layer of the business. I think that is incredibly important. You can find a genius, but if they just have a horrible cultural fit, it rubs everybody up the wrong way. You're not going to be very impactful. You may implement a lot less of the good ideas than you wanted to because of that cultural fit, but for me, usually in individual contributor roles, technical roles, technical excellence is at the top of that pyramid, right? That doesn't mean that cultural fit's not important. I still love to have people in the interview process meet as many of my team as they possibly can. Now I'm not going to just flat-out reject a candidate because one person in my team was like, I'm not sure about the cultural fit, if everything else is a green light. But again, if you're choosing between two broadly similar candidates, right, and your team's this person really liked their vibe—they're a positive person—why wouldn't you probably use that as another tool? So again, I think it's probably we shouldn't take this in terms of the extremes that either cultural fit is not relevant or it is, but where how important is that? And if you can use that to differentiate two very similar technical candidates, then I think you should.

TIM: I feel like personally with any of these criteria that are a bit more on the subjective end of the spectrum, one thing that I've personally found helpful is to unpack them and try to quantify them even if they're subjective, even if it's an opinion. For example, you might have a process where you're evaluating all these technical skills and give a candidate like a score out of 10 for SQL, a score out of 10 for their data analysis, whatever There's no reason why you couldn't also score them on their cultural fit. It's down to the person who's scoring them and their view of what that means, but you can still reduce it to a number. Thanks a lot. And also you could almost unpack it, so for us we have eight business values that we try to hire for. So in theory, if someone is interviewing, you can say, Evaluate them across these eight values of our company. It could be a number, could be good, bad, or okay; come up with something, but at least that's better than just getting to the end of an interview and going, Ah, they weren't a good fit.

PETER: Absolutely, yeah, I think in our team we typically do that cultural fit based on a one-two-three-four scale right on values, and it's simple, but we also ask everybody to write that down, right? I want everybody who's seen them, even the maybe less outspoken people, to give their one-two-three-four before The rest of the outspoken people in the room have declared their opinion, and that way it's a little bit less prone to kind of groupthink, and you're getting a wide perspective on that cultural fit, not necessarily just the most outspoken people in your team.

TIM: Yeah, or deferring to your leadership and your view, which I guess would be easy for them to do as well, so yeah, that's awesome that you have that. I imagine you've had some scenarios then where maybe someone's given the person a four and someone's given the person a one. How do you then go about reconciling that? Do you have a kind of discussion to unpack it or something else?

PETER: I'm not sure I'm not actually sure that we've really ever ended up in that kind of situation where we've had that big of a difference on even something like cultural fit, which is so subjective, but I would ask people to give that their logic in terms of why they think that person is and it sets up a discussion within the team. At least you've now committed to that, and then we can discuss exactly why you thought that, but oftentimes you find that people are actually quite close on these things in my experience and that we're more talking about, yeah, twos versus threes and that kind of thing. There's less variation than I would probably have expected before implementing something like that.

TIM: I wonder if you don't have a big discrepancy in people's views over a candidate because your team's already well structured; you already have quite a good sense of what you're looking for. It's like a well-oiled kind of machine where then there's not going to be that huge variance of people behaving completely differently to each other and expecting different things.

PETER: Yeah, no, you're probably right; there's a bit of nuance there where if you're like a newly formed team, what is your cultural fit? As you start acquiring more and more people, it becomes much more clear what that cultural kind of identity is, so that's a good point. I think you really it's probably easier in organizations where you have that well-developed mission set. Yeah

TIM: One thing I hear hiring managers in data quite often say to me is something to the effect of, Oh, look, I really favor soft skills and technical skills. I can teach, or anyone can learn technical skills, which at a certain level I get because I think soft skills—your personality—is more or less fixed. The way you communicate can develop, of course, but it's probably harder to change a fundamental soft skill. Whereas you can learn a new technical skill if you're motivated, from nothing it's possible, but I also feel like the hiring managers, when they say this, are being a bit dismissive, and I feel like they have forgotten how ineffective they were as grads, okay, in terms of their technical skills. I'll give you a specific example. I can vividly remember from one of my first jobs as an intern at PWC—this is back 13 years ago—I had a spreadsheet of some kind of data. I was doing some kind of basic analysis, and then I was scrolling down this massive sheet of 50,000 rows and scrolling up again to see what the column header was again and again. I didn't know how to freeze the paintings, so my buddy came over and said, Look, God, this is killing me. Just quickly freeze the paint, so I'm like, Oh, cool, so that's just one micro example. I mentioned how shitty my skills were 13 years ago if that is just one of the things that was wrong with my skillset. like how ineffective was I? How lacking was I in technical skills to actually do something? And so I wonder if some hiring managers forget this; they're not humble enough to remember what it was like to begin with, and maybe they've undervalued that incredibly valuable skill set of actually being a data professional. What do you think?

PETER: Yeah, from my perspective, I think it's a bit of an oversimplification in either direction. Right? I think for me there's probably like a middle ground where it's light technical analysis or you need a little bit of soft skills, where I think you could make the argument either way, but on the extremes, there's some extremely technical endeavors that would be very tough to teach to somebody no matter how good their soft skills are. If they just don't think in a certain way, they might not be able to become a really advanced data scientist or principal data scientist, and likewise, no matter how long you spend with them, I think there are some people who might not ever be able to be a chief data officer from a soft skills point of view. So I think it isn't one or the other, and I disagree with that. Like, if you have the preference, why would you ever not hire someone with the skill sets that you need for the job? I think sometimes it maybe comes from a lack of clarity about exactly what is needed for the job. But often we have to make compromises, and where those compromises come from depends on the role, and then I think that's where you need to say, I could probably do that, teach a little bit of soft skills to this data engineer, or I could probably teach pivot table usage to this kind of business analyst. but I think it really depends on the role and what is more important.

TIM: And I would have thought then navigating those trade-offs would be easier if you took this approach you mentioned before, where you've measured as much as you possibly can in the hiring process, because then you've got this matrix of the strengths and weaknesses of candidates, and you can say, I can see these candidates are similar across these eight dimensions. This one's much stronger here. This one's weaker; here you can almost think forward to the training you're going to have to do. Is that how you approach it?

PETER: Yeah, absolutely, and you might even be in a position where you are offered a person with soft skills. It's just that, on average, their salary requirements are 20 grand more, and then you need to ask yourself, Okay, how much do I really need this? Could I? and you can make easy decisions. Could I send them on a five-grand training course? It becomes a lot easier to make these kinds of decisions, but yeah, training and development are I'm hugely passionate about training and development, but they're also time-consuming and expensive, so I think you need to really clearly know when you should bring it in versus build it internally.

TIM: You mentioned that, actually, the soft skills are attracting a premium in terms of salary. I'd never heard someone presented like that, so if you've seen this firsthand in terms of the softer skills leading to a boost directly or just candidates coming in with those really strong soft skills and expecting a higher salary,

PETER: Yeah, good question. It's certainly chicken and egg, right? We do the soft skills allow you to negotiate better and therefore get a higher salary. I think that's probably part of it. I think going back to the Moneyball example again, softer skills are easier to identify by people in the process who might be in talent acquisition or in more general management. It might be easier for them to identify those points than whether they can do an inner join properly. And therefore those might be a little bit more priced into the market, and things that are a little bit harder to unpick or harder to identify, harder to measure, they might not be priced in. Yeah, I definitely think, though, that soft skills are often much more priced in because of that initial wow factor when you talk to candidates.

TIM: That's a great analysis, and I hadn't heard someone explain it that way before. That made me think a little bit about talent acquisition and HR and the fact that it would almost be helpful if they were a little bit more cognizant of the inherent bias that they probably have, which is that they feel like they could evaluate soft skills and cultural fit quite well because they can see it, but they realize they're probably not in a position to evaluate technical skills. Do you feel like they are then overvaluing the component they can see on average, and is that something that they should at least be mindful of?

PETER: Yeah, I think that's fair, and that goes both ways, right? I think there should be a real two-way dialogue about what you're doing and what you're trying to measure on your interview stages, and if they're doing an initial screening call, make sure that you're really clear about what soft skills you want. and what kind of experience you're looking for. I think, from my perspective, we try to eliminate a lot of unconscious bias in terms of CV screening and stuff like that. I actually like to not see CVs until we've been through the technical testing. The reason why is, yeah, I don't want that to lead me; I'd rather be evidence-led off of how they perform in the technical testing in the first instance, but to do that I need to have a lot of trust in my talent acquisition HR that they're going to give me a group of candidates that aren't going to waste my time in the technical stages. So it needs that trust that good communication between those two areas, and yeah, I think that's where you can get the best results.

TIM: That's really interesting, so basically you'll be The second stage in the process, I'm guessing, after talent has spoken to the candidate and done some screening, then you're going to come into that interview without really knowing them at all. Is that like how fresh you are at that point? And you just have a blank slate like that? They're on the level playing field at that point, is that right?

PETER: Technical tests are the most objective or easiest to make objective, and let's go through that, and obviously we want them to have a visa; we want their salary requirements to be in the range they have; roughly the years of experience, but I'm quite explicit that I don't want to be overly selective about the companies, or I don't want to be overly selective about where they went to university or what they're asking for because we might not unearth that undervalued talent again, right? And I think after that, usually the last stage in the process is going through the CV and speaking about previous career history because at that point, you validated that everybody can do the measurable stuff well.

TIM: That's amazing, and so you've flipped the process on its head in a sense to really focus on, as you say, the most objective, most measurable bit first, and if it's like this is a take-home kind of technical test that you give them at that stage, is that right, or is it in person? How does it How does the test work? Without giving away the details to the market

PETER: Yeah, no, it's a live test. We may end up speaking at least once about AI today. The real reason is that it's just I think too hard to guarantee that an individual is actually doing the technical tests unless you do it live. There are a lot of drawbacks with that. A lot of people do not Enjoy that kind of pressure, and it can be a difficult environment, but we try to do everything that we can to kind of minimize that stress and make sure that we're getting the best out of everybody, but yeah, still live is the way that we do that.

TIM: Yeah, that's fair enough, and having worked in this game for five years now, everything's a trade-off, isn't it? Like, you could do a take-home test, which has a higher chance of cheating, but it's in maybe a more relaxed environment. It's going to take the candidates much longer to do. Probably you did the live thing, almost zero chance of cheating, especially if they're in person. That would take a remarkable effort from a candidate to cheat in an actual interview room. But yeah, maybe you could argue it doesn't quite simulate real-world life. I know when you people used to come and look at me trying to type sequel on my keyboard, I would lose the ability to use my fingers, so I can see how for some candidates It might be a little bit difficult to perform under pressure, but yeah, I don't personally see a better way currently of approaching it.

PETER: Yeah, I think you just have to do everything that you can and don't neglect that in terms of trying to set it up in a way that the candidate knows that it's not designed to be an intimidating thing that you're there to help them through this, and the main reason why we're doing it live is to ensure that there's no kind of copy and pasting of large blocks of code. I go to lengths to explain this isn't you can open up Stack Overflow and check where the comma goes in this particular kind of syntax. That's not really what I'm testing here, right? I just want to make sure that the basics—the tools that are required to do the job—

TIM: And yeah, this approach you have of initially seeing the candidate from your perspective in this technical test, where it's really focused on something objective and measurable—can they do the work that you need them to do?—is a great way to then prevent, as you say, those biases that can creep in. If you're fixated almost on their CV, initially CVs have potentially someone's photo, their age, their gender—all these things that really don't matter at all. Let's be honest: have you seen any other interesting ways that companies do or try to make the hiring process a little bit more objective?

PETER: Yeah, I guess there's lots of tools out there, right? And I don't think that there's a silver bullet here for data extraction from CVs and that being utilized to try and put things on a level playing field, remove formatting, and stuff. I've also heard that those things can have difficulty parsing certain CV formats, and then, you know, do you really want to reject a great candidate because, you know, their CV was in a PDF and that kind of thing? But it's interesting to try and do something like that again because really there are some people in your organization—the people who are putting together PowerPoint presentations—where they're formatting the CV really matters, right? Or maybe your storytellers in your organization, but there are others where it really doesn't matter, and it shouldn't matter, so yeah, if you can separate those and use tools like that, I think other things that, you know, tools that I've heard about making job descriptions less biased You can feed in your JD, kind of things that you're sending around comms around the process, and it will highlight Hey, you're using very masculine language here, or I'd recommend you replace this with this instead, and I think those kinds of tools are really interesting for me, and yeah, they can really help you out. Obviously, live coding platforms, as I've already mentioned, I think those are great, and the better the environment, the more easily you can collaborate and make it a collaborative kind of coding exercise more akin to what it would be at work, the better.

TIM: Yeah, interesting you mentioned the job descriptions and job ads. I was personally a little bit dubious of this myself just because I'd seen this stuff spoken about so often that if you just make fairly subtle adjustments to the job wording, you can have quite a profound impact on the applicants. But I just interviewed someone last week who went through exactly a fairly comprehensive experiment they did years ago to basically validate and prove this very point, and I was quite amazed at how big an uplift they got. And in that case, they were trying to get a more balanced gender ratio for their software engineering team. It started out as 90-10, and they got to 50-50 primarily by rewording the job ad and then changing their sourcing strategy to just target certain segments a little bit more aggressively. So it's a really clever thing. Interestingly, though, the way they arrived at this conclusion was using data Because they looked through their hiring funnel and saw exactly where the drop-off was and where the problem was, they realized that they weren't hiring enough women to their team because only 5 percent of the applicants were women. and I feel like this is where that kind of data-driven mindset is data for good. It's not. I feel like there's a perception sometimes in HR that data is like the evil negative. It's like anti-human or something, but it's a great example where you can use data to actually make good decisions. decisions to improve things, which I think is awesome

PETER: Yeah, absolutely. With LinkedIn and all the details that they give you, you can really see what the difference is in terms of, yeah, my candidate funnel from making these slight changes to a job description, and yeah, being evidence led sounds like that's what I think I'm advocating so far on this chat.

TIM: Yeah, exactly. What about a little bit of future gazing then? AI is changing, it seems like everything at the moment in every industry, some profound shifts happening. What about for data roles? So you've got your classic data engineer, data scientist, and data analyst. What are they going to be doing in a few years time? Will these roles even exist? They still exist but are just doing different things day to day. What do you reckon is going to happen?

PETER: Yeah, people often make fools of themselves by trying to predict where technology is going to go in the future, but I think at a five-year horizon, I'm comfortable with this one. First, first part of the question: Will they even exist? No question. In five years time, each of those roles will still exist. I'm pretty confident that what they will be doing day to day will probably change, and I think that will depend on a few different things as well. It will be unique to each of those roles, but also it's very industry specific, I think, what will happen, but yeah, to just use an analogy, right I don't think calculators didn't replace mathematicians; it just changed the way that they worked. It did make math more democratized. If you only need simple calculations, maybe you didn't need a mathematician anymore; you could get your kind of somebody with a reasonably numerate kind of way of thinking to do all sorts of stuff that you had to hire one for before. but then the mathematicians were still required, but they needed to up their game; they started peddling calculus and linear algebra and all these kinds of things, right? But I think we're in a similar situation here.

TIM: And so as an extension of that, I wonder if with the development of AI, then the barrier to entry to doing some basic analytics might be reduced such that we'll start to see a lot more of the kind of almost shadow analytics people emerging. Some of it is maybe like pretty handy with a spreadsheet, but now with ChatGPT Oh cool I can write SQL. I can get into the warehouse myself, maybe also the domain experts who have the knowledge about the marketing or the product or whatever, and now don't necessarily need to go through a centralized BI team; they can do a bit more self-service analytics. Can you imagine that happening, or is that a recipe for disaster?

PETER: No, I think it's definitely the direction of travel. Obviously, you need your data governance in place and really great kind of labeling and definitions to make that all work, but yeah, I think democratization is a huge benefit that we're going to see here. It's going to speed up workflows. I think where we're going to have the biggest impact on people in data is with the newest arrivals to data, like where the skills are the least developed at this point, because trying to differentiate yourself from the average business user, who's quite numerate, is going to be difficult with AI because, yeah, it's just so much easier to break into, but yeah, I think it's a matter of just new skills will become useful, right? So for data engineers, for example, there's probably loads of things, like in terms of tedious tasks, like creating ingestion pipelines and stuff like that, where they can work a lot quicker because they can just go into ChatGPT and say, Hey, roughly, or get GitHub Copilot and say, Here's roughly what I'm trying to set up. and boom, you've got the template. You'll need to check it. Yeah, it can't be completely automated. You need somebody who knows what they're doing, but they can work 50 percent faster than they could previously. But I think as well there's just like different skills that will now become relevant. Like that's always been the case in data. Like 10 years ago, nobody cared about Python. You'd see SAS all the time on job adverts. and it's so I don't think anything new is happening here. That's always been the case with data that new skills come in all the time, but data engineers that are doing those more tedious tasks or that's the things that they're calling out on their CVs versus data engineers who can talk about RAG or LangChain and things that are useful in the AI revolution—that's going to be where the jobs go.

TIM: Yeah, surely it's a great time to be in data. If you're willing and able and keen to upskill, you're in the perfect place. Surely the perfect time to really capitalize on that.

PETER: Yeah, and I think as well, for everybody who's in data and worried about this, think about how you can use AI to improve your workflows. Right? What could you do to make sure that somebody else who is using AI isn't more productive and comes in and disrupts? So I think people should be really looking at that and making sure that they're trying to leverage that again. calculators mathematicians example right it just made them quicker and better at their jobs and think about how you can do that I don't think people should be scared about it I think they should just think about how do I differentiate myself in that world Yeah, I do think, though, that you have to realize it's going to be a long time before AI takes hold in every industry, right? And, like, regulation is a huge kind of like axiom that I was thinking about here, where there are so many industries that are so heavily regulated that in the next five years it's unlikely they're going to be using AI in any kind of large-scale applications. That will stay, and we spoke about startups before, right? I think that's where you're going to see the biggest change, right? Like, why would I hire three junior analysts if I could hire a senior with AI that could do at least as much if not more? And yeah, you're going to have, at the opposite ends of the spectrum, more or less resilient to AI.

TIM: Yeah, I feel like with these changes, you could look at it on either side, can't you? So it feels like an existential threat. Things are moving so quickly; it feels uncomfortable, but the upside is astronomical. Just imagine pre- and post-internet; imagine being the first person to figure out how a web browser works, and oh my God, the potential upside, especially if you're entrepreneurial and you're willing to kick something off, is huge. This is basically unlimited from what I can tell at this point, so super exciting times, I think.

PETER: Yeah, and the Internet's a great example as well because even today you still have people in the organization who aren't Googling to its fullest potential, right? So the same will happen with AI that if you're interested and involved and starting to use that, you will differentiate yourself from a lot of people out there. Everybody's talking about ChatGPT, but how many people are really using it in the way that they could in their professional lives?

TIM: Yep, as a closing remark, I've heard from, I think, three companies I've spoken to in the last few weeks that basically said to their entire organization, Stop working. Don't do anything for a month. Just use Claude and Chachapiti in your everyday work and just do one task after another until you've automated away a lot of this kind of busy work that you're doing. and so I feel like that's at least interesting; at least it gives you a chance to really think consciously. Hang on, because it's so easy to get into habits like you've been doing the same thing for 10 years, and suddenly there's a new tool that could automate it for you until you have an opportunity to actually do that. You could easily just continue doing things the same way, so that's an interesting approach. I'm keen to see how it's worked out for them.

PETER: If you make the time for your staff to explore these tools and make them less afraid of it, and it's almost encouraged, I think you'll get the best results. That makes sense.

TIM: Yep, a hundred percent.