Latest / Range Podcast / From AI skeptic to builder: Evert Kraav’s Journey in Transforming Compensation
Transcript
- Giac: Good morning everyone and welcome to another episode The Range Podcast. Today I'm Sir Evert Krav. Everett is dialing from Estonia, Tallinn, and is currently senior manager in global compensation adult. That is probably an understatement compared to all the things he has been doing in the past I've seen him on TV, I've seen him giving lessons at the Estonian Entrepreneurship University of Applied Sciences he also runs his own business called Cohorts. Everett, welcome to the podcast. What's the first thing you about the moment you start preparing for a holiday? What are you packing in your bag? Evert Kraav: Hey Jack, that's a really good question to start with. guess usually try to pack really minimal and take little items with me as possible so I can buy something from I'm going and take them back home with me. Giac: Right. So you're ⁓ you're a souvenir person, is that what you're saying? Evert Kraav: I think every travel you should have something to remember this place by. So you could say that, Giac: For that's listening in now, you should know that Evert has a trip planned to Rome and he's a fan of So if if was a souvenir that you'd love to take back from Italy, Everett, what would that be? Evert Kraav: Usually those tennis championships, they all have their own little caps the logo and stuff on. So I that's kind of my go-to. Giac: I'm completely unaware of what happens in tennis courts because I've never played tennis in my life. So that's that's was an interesting one to hear. ⁓ Evert, said about but ⁓ why don't you take a few moments to tell the audience what been up to lately? Evert Kraav: Well, I do work in compensation. So most of my job is around how much people ⁓ be paid. But obviously last year has been really interesting and exciting due to all of the new tools that we now have available to us. And I'm sure we will talk about them later as I think specifically the last six months or so have been really been about us building. different stuff with the tools that we now have access to. So I think it's really cool to see how much progress we can do in such a little time that we had. Giac: Yes, absolutely. We're all in that learning journey right now, right? We we know something is happening. No knows exactly where it's going, but nevertheless, you know, I think there's two kinds of personas, right? Those that are willing to take some risks and experiment as as they can explain the risks, and those that prefer to stay on the sidelines. There's no right and wrong. Everett, which category you belong to? The takers or the observers? Evert Kraav: If you would take a look at my strengths evaluation, you would say that I'm more of a cautious person. But think in reality, think we do take a lot of risks and we do ⁓ experiment a now, I think if you asked me this a year ago, the situation would have been different. But I think that now, yeah, it's basically a daily thing for us to try to figure out how we can do our work in a better, smarter way and how it can also more easier for all of our clients customers, which mostly are our managers in the company. Giac: A year ago, when first met, ⁓ Evert and I had a about, what what's happening in the AI space, specifically in Total I was very interested to know at the time how he was ⁓ Utilizing AI. ⁓ and at that time I think Everett, you were a little bit more skeptical as in, yeah, like I'm hearing all these things happening, the tools are constantly evolving, and that is quite distorting reality. That is quite confusing for a reward person because we all know that we need to make a big investment, but we don't know how big that investment is in terms of the time required. the mental sort of capacity, you know, the the the energy required. We all have day jobs, right? So this is just going on top of everything else. What was that trigger point for you, Evertt trigger that you said, right, from today, I start into the weeds of it and start learning how to apply these tools in my comp work. Evert Kraav: a year ago I did feel like the learning curve was a bit too steep and that it would take me hours to actually become proficient in them. And why also perhaps I wasn't so immediate in adopting But I think this our team got access both Claude Cursor, which usually given to developers only. And I think immediately the things that we were able to create in a matter of like or hours was something that blew us away. know, most in compensation have built their own little spreadsheets with ⁓ formulas and and stuff and how we were basically able to create the ⁓ much more version of these was something that us really rethink the whole approach. because last year as well I did spend I would say a significant amount of time to look at different vendors also provide different types of tools that should make our life more easier and efficient in terms of like job evaluation and ⁓ ⁓ creation all of that and they cost a lot to be honest and now to see like how much we can actually create in a I don't know how much it costs these days, 20 something euros maybe per month. we only take Claude, it's a huge ⁓ step and should be also concerning to all of the companies who are providing these types of services themselves. Giac: it is definitely interesting, right? Because for decades ⁓ we have saying that the reward function, but even more broadly, the people slash HR function, that there's been underinvestment in in this right? From a technology standpoint. we've been using, I mean, for decades, the same tools over and over again. We've been using Excel. For decades, we I think calm people and benefits people, they've always been building. They have a builder mindset. They've been models in Excel. But it's always been with technology that yes, has been upgraded over time, but technology has been around for what, 40 years? I think was the 80s when ⁓ first came out. So Now we do have that builder mindset, but we also have access to new tools. the is though, how do learn to use these Right, because we all know they exist. AI has been around for a while, but how like who's telling us how to use them? Is there someone that you ⁓ you learn from Albert or or is it primarily just self-learning, like getting your head heads down and starting to use the tools you get familiar with them? Evert Kraav: I would say the inspiration came from one of our business partners actually, who, because she's supporting team who ⁓ a bit more tech savvy people saw what was possible create. And then also created some things on her and shared them with us. And then we saw like, wow, like you created this ⁓ on own. And immediately wanted to do similar. And once we got access to the tools, it's mostly just testing and ⁓ figuring out ⁓ I say. is probably the best way learn. And also, ⁓ mean, the tools develop and change quite these days so any of online course or whatever is probably outdated in a few months so just to just keep yourself up to date and see what you can create. I think once you actually build something on your own and see the outcome and see like how easy it is to actually create and update these things you don't want to stop at least I don't. Giac: It is quite addictive, right? Like what it think the learning journey looks a little bit like a hockey stick from what you're saying. Like it's so hard to get started. But once you find that person or that group that inspires you and drives you know to that point then the learning journey like it's exponential ⁓ but to your point you you have to experiment on own but to your other point having someone that you can compare notes with someone that inspires you helps you also understand right now I'm using the tools I've made the I've made the effort But you never know, are you using them the best way possible? Are you using the right way? Because I mean, things keep on changing, don't you think? Like every other week there's a big announcement, and you find out about maybe a cloud skill or maybe another tool that can do something better. It's a matter of how you use the tool, right? ⁓ You earlier you mentioned it is, know, just astronomical is incredible what you can build in a couple of minutes or a couple of hours. And I know that a lot of people ⁓ get a little bit skeptical when you tell them, hey, you can build something amazing in a couple of minutes. Now, in your experience, Evert, everyone can build, I guess, once you have access to the tools. But what distinguishes tools that actually work well from tools that just look nice but might not be doing exactly what you think. Evert Kraav: If we're talking about tools that you can build on your own, think there is definitely a few aspects that you have to a little worried about. And that's obviously like anything connected to people's data and how it is actually used within tools. So ⁓ I that's probably the biggest concern some people also have is like the whole data sensitivity and how the tools create might actually leak something. So it is possible if you just like place sensitive data in it, then it becomes part of the code and it's possible also to extract that later on. So I think that's kind of one thing to always mindful of. The part also is like, just to remember that whatever ⁓ is it doesn't think on its own. ⁓ So it still produce errors if you have kind of kept those in somewhere the during the building ⁓ I've also had similar experiences. So for example, one thing that I've created for myself is like this little little dashboard tool, however we want to call it, like where I just can paste in the survey data that we receive ⁓ from of the providers. And then basically it will ⁓ create me the salary structure on the parameters that I give it to. So in I pull in the data, it will me the salary bands according to our But if I don't, and I didn't in beginning, code in like a little that you have to also monitor if there are outliers in the market data or any type of data, then flag those and just don't create something illogical. then ⁓ think that's another thing to look into. So especially if you're working with these types of data where outliers and mistakes be in. Giac: Right. Absolutely. And we we all know that especially in Europe, where population size tends to be smaller than US, you can get patchy data. But let's let's dive into that a little bit. So what I'm hearing is among other things that you've built at Bolt you build ⁓ Sounds like a salary band design tool, something that builds bands off of market data. Can you explain a little bit more of how it works? Like we all have different approaches to benchmarking. instance, one could use a multiplier approach, like it to the median and then apply multipliers. One could have like a predetermined pay grade structure. one could different approach like there's there's no like right and wrong right how does the tool know your preferences and your approach to benchmarking and creating salary bands at built Evert Kraav: So of all, we the data to it. So basically explaining like, you know, this is the market data, this is structured in a certain way. And then also explain what the output should look like and that there, the, for example, labeling structure is different. And then I map out how it should be converted to, and then also explain, or actually I don't even need to explain, I can just add. in the kind of or choices. Like, you know, if I want to build it according to the median, then I can choose that. If I want to build it according to something different, I can easily switch. Also, what else do I have here? Like, how wide I want the ranges to be. I do in it also the kind of checker whether there are outliers or some illogical outputs. So, I can easily flag that as then also I have a little like choice whether I want to create the ranges based on all of the levels or if I want to like sum up by grades and create a flatter structure. I would like you don't necessarily need to put in all the details. You can also add the option choose each time. ⁓ How do you actually to create them? And then play around and once you it then you can just also add in what is the default approach and will be something that you can then usually use in future as well. Giac: It sounds like an iterative process where you go back and forth until you find right output. Now the question for you, Everett, is, there are rumors out there that AI is actually not saving us time at all, that some tasks that are delegated to AI, that the entire workflow is taking longer than usual. Has that been your case? ⁓ Or do you think it's more a temporary, pent up investment at the beginning where you have to set up things to make sure they work, to check the output, but then it's like the the actual time to get tasks completed decreases over time. Evert Kraav: Definitely you have to take some time to invest in the beginning just to get the correct. But I would say like after that, if you keep using ⁓ you will time. I would say from ⁓ simple tasks restructuring files into how you typically want them to be, or like for example, combining survey data from different providers. You can automate that really easily. Once you have done that, you can use that file to kind of build the tool that will provide the bands for you. And then ⁓ depending how often you do it, often companies do it once ⁓ a Nowadays data ⁓ can received four times a year or even more. So essentially you are able to keep doing those checks multiple times per year without like taking huge amount of time to manually or with Excel through the data that you receive. So we'll take time in the beginning but I mean the same applies to any tool or any new tool that you would like to start using. There's always a learning curve and once that is done life will get easier. So I think it's the same with all the tools you might build yourself. Giac: Yeah, it compounds over time, right? Although AI on changing things, and you might need to spend time to review your approach. Might still work, but it perhaps has become obsolete because it's a better technique. It it still compounds over time. You're you're building over time. I wanna back to something you said earlier, which is really the elephant in the room here, because ⁓ We, you know, every compensation and benefits professional, we are all dealing with confidential data. ⁓ are regulations in place all across the world in terms of how you can handle this data. And when uploading ⁓ sensitive data to a tool ⁓ like ⁓ even if you have an enterprise license, there are still a few things that you want to in mind, right? So for instance, you mentioned someone on the other side of the fence, someone that is in the company that has access to these tools, perhaps a software engineer, might be able, you know, to open a backdoor. and have access to the data that you were uploading to the tool, which they might not have access to in their HRES or ATS system. is that something you've encountered and did you mitigate that risk? what was the process like at Bolt? Evert Kraav: Well, was a lot of self-learning. think that part we all ⁓ of knew instinctively that we shouldn't like just paste in all employee data into the tool. So we basically just give it enough to build something to understand the structure. And then most of the things that we have created kind of operate in a way that ⁓ you the tool that sits at the moment ⁓ offline your computer. you open it up and then it asks for you to a specific file that actually includes the data. So this way we don't give the employee data to Claude, for example, itself, but rather only the dashboard that we have built. And same with some things that we as well for managers, for example. So managers just receive a little template, then they download from Workday the data that they have access to and then they can drag and drop that into the tool and it will then show them the things that they have access to. But the of like somebody from ⁓ tech team have access to it, I mean that's ⁓ also true with of any other tool like Workday or Google Drive or like you know you might put files there, but someone ultimately has admin access and the risks are there anyway. guess that part, any company will need to work anyway. in that sense, again, I don't see it like it's super different with the AI tools. Giac: Right. Yeah. Yeah. It's less about the tool itself more about your process around ⁓ sensitive data and how those responsibilities are distributed in the company, right? To your point, someone in IT security might also have access to HRES in some shape and form and have access to that data ready. Evert Kraav: you Giac: And with paid transparency laws, which I know you're a big fan of, being rolled out across the world, you know, the perception of what can be shared and what not and for what purpose, I think is changing, right? Evert Kraav: Definitely. think also what did was we sent out the message to ⁓ all the team and also them like, now that you access, if you are building, please don't put this type of data into the tool. ⁓ Or if you of need to do something, then there are ⁓ different ways how can increase the security. One is like create this upload function. Second is actually Giac: Mm-hmm. Evert Kraav: make it ask for a password or something. So there are ways that you can kind of mitigate the risk, I would say. Giac: Indeed. So looking back at our conversation the past twenty minutes, we touched on a benchmarking tool builds salary bands off of market data, which is presumably a tool you use yourself or maybe your team uses, right? Is boosting ⁓ and your productivity. Evert Kraav: Mm-hmm. Giac: But recently you touched on another You said that you built something for managers at as well. Could you elaborate a little bit more on what you built, what drove you to build that tool, ⁓ and what the reaction was essentially. Evert Kraav: Yeah. So that was something that we built specifically for the pay transparency, which, you know, is coming and that we also wanted to increase and implement in, in volt. And if you have used like Workday for compensation review, then you know that once the review is done, basically what you can give to managers is a report that will just plainly display the data, but it doesn't, it's not really interactive or it doesn't visualize anything or help. Giac: Hmm. Evert Kraav: people understand it. So kind of overcome that step, we built our own little ⁓ tool, which I can also maybe show for a bit. And the first version of that was an file with formulas. ⁓ And ⁓ really to, yeah, exactly. And didn't look nearly as nice. So as I said, like we created the little dashboard, so we gave the template. Giac: Uh-huh. What we're familiar with. Mm mm. Evert Kraav: to all managers. They downloaded it to their own computer so it only works offline. And then we also said they need to another from Workday and is they can only get data that is relevant to them. And once they upload then it of presents the data that exists in Workday a bit of a better, more interactive way as I said. ⁓ We will show, we meet the other people ⁓ in the team, the levels, their performance ratings, salary increases. And then you can go in a bit deeper and see exactly how much was the increase, how it relates to the band, how it visually looks like ⁓ in the We've also added in some explanations, like where do we expect certain people to be? also added in the flag where Giac: Incredible. Evert Kraav: employees in, in which country, because for EU countries, we wanted to also make that managers share ⁓ the range when they having to talk. And then there's more data around total compensation, but also ⁓ what we, actually what team built was that depending on the specific scenario. So example, what the performance rating is, where the person is ⁓ the salary band and other data. there is like ⁓ ready-made talking point that can use. Yeah. So it's not like just give the data, but also like what you should mention based on specific situation you have at hand. So I think that was a really good thing that managers appreciated because ⁓ I will like a lot of managers maybe are not that experienced in having these types of specific Giac: Love that. Evert Kraav: specifically around page transparency and actually sharing the data that usually was only available to them. And because we have, or we are using terms that, probably managers don't think about every day, then we also added this whole kind of dictionary or type of explanations to the end. So you could immediately like answer questions if employees have ⁓ some. also links to all of the pages that exist in our conference. Giac: This looks great, Everett. Thanks for sharing. ⁓ just before we go into the dashboard itself, I did clock something very interesting at the beginning, the first window, which much aligns with what we discussed earlier in terms of sharing employee data. the runs entirely in your browser. So no data is sent to any server and your employee data stays on your device. I think that That call out really helps mitigating the risks, but also changing the the perception the managers have the tool, like creating that level of psychological safety that you need when when handling data and using a different tool. And also I also I also clock that the actual steps required for the manager are quite straightforward. All they need to do is download a report. That they would have otherwise downloaded anyway from workday, and then just upload it to this online that runs in time in your browser. ⁓ this is an illustration, it's not real data. ⁓ ⁓ what I think what we're we're looking at here is a ⁓ AI case. ⁓ Evert Kraav: Mm-hmm. Giac: In compensation, paycoms, ⁓ the end of a cycle, and it changes the perception ⁓ of what AI do for in our domain. we were talking earlier, talking about a tool that increased your productivity, right? Saved you time and maybe even surfaced insights from the market data that you haven't caught. With this tool. We're going much more beyond that, right? We are giving, we're building a tool managers and maybe also people partners slash HRBPs in a company that solves one of their specific problems. No one, like everyone, ⁓ likes do performance in peer reviews, especially when it comes to communicating the and the outcome. So this is where your tool comes in. I think it provides an opportunity. It it changes the relationship managers have with the compensation team, right? Because now you're much more on the front end and they can tangibly see the results of your work as not a black box anymore. Evert Kraav: Absolutely. I think ⁓ every year, time that we announced that now is the time to start these talks, we are like a little bit cautious, like what's now going to happen and ⁓ type of problems will surface. But I think this year, once we did that and we specifically said, like, ⁓ use tool, try it out, give us feedback. The morning was really different because basically The Slack thread was filled with people saying nice it is and very useful. And they were surprised that, you know, we have made this type of improvement. think we kind of also showed them that HR teams can also go with times and build something that actually looks good and is And the funny part with the whole process was that there were a few who were saying like, yeah, I wish it had had like this feature or that option as well. And then there was like one manager who basically said, yeah, I coded some new features in it. So try this one out. ⁓ it was basically our managers also building it themselves, which then we kind of said like, okay, let's try to keep it centralized. So they gave us all the feedback and created an updated version that had the features they felt that were lacking. Giac: Ha ha. Evert Kraav: So it was really good to kind of collaborate with them, but also show that we are willing to take the feedback and work with it. So it was a really cool week, I would say. Giac: Yeah, I think the adding like section, the tool, a button where any manager can submit their feedback without going on a tangent and create their tool, you know, that's probably desirable. And ⁓ probably not many managers to do that. Like this is all like additional responsibilities they have on top of the day-to-day work, you know, people management, communicating pay outcomes. But to your point, like having this direct access. with the management population and having and being for them being able to see a compensation tool does give you new ideas because I presume following this episode where this manager created his own tool, you definitely learned something from it and you reflected ⁓ the improvements and in the tool that you s run centrally. Evert Kraav: Yeah. And one of my favorite comments that came from one of the managers was that the, they're them saying they are embarrassed on behalf of Workday, this isn't already a feature ⁓ in the system. And think it's true for many systems now that they really to catch up because we're just going to like do it on our own and not really need them to, to create own because they're already late. Giac: Ha ha. Right. anyone from Workday has dialed in, please take notes. This is free product design advice ⁓ user feedback for you. excellent. thank you your time ⁓ and thank for showing us this tool. I wanted to end ⁓ our ⁓ with question. ⁓ you probably have ideas going on in your head now in terms of what you're gonna improve next in your existing tools, but also what you're gonna build next. So if you could kind of wave a magic wand and automatically fix one massive headache in compensation world by building a tool or maybe ⁓ Enhancing a tool you're really built, what would that be? What is your primary use case you're going after? Evert Kraav: Yeah, that's a good question. think one topic that usually brings us questions is probably evergreen one. that is the situations that might come up between manager and their team member when it comes to compensation. And I think if there was some sort of possibility to create something that the manager can kind of give in. the inputs of what the situation is and then immediately get like a really good ⁓ next steps of what they should do and what they shouldn't do, then that would ease up our life as well because I think from companies, usually I would say the biggest headaches come from the times when something is said that shouldn't have been said or promised. ⁓ So you would have that... Giac: Hay Evert Kraav: possibility to fix that somehow then I think our lives would be much easier. Giac: Well Evert, thank you for making it time to tell us more about your AI journey, your learning path, your AI bills and how they were received in your ⁓ Any last remarks, Everett, for those haven't yet started their AI journey, that are skeptical, that are afraid? What would you what would be your number one advice? Evert Kraav: I would definitely go and request access to a tool, like even if it's temporary for one month or two months, and just to see how much you can create in that time. And if you really take like a few hours per week, even, I think you will see that it will definitely improve the way you work. I know that even within our team, there are concerns like how much or how long we have jobs in that kind of world AI develops so fast? But I think, one way ⁓ to make sure that your future is safe is to kind of keep up with the times and ⁓ learn tools understand the tools so you are able to stay in the conversation as well. So I think that's why it's useful to like just be proactive and not kind of wait because then at some point something might already come that automates everything that you do. Giac: Keep up with the times. Absolutely. Thanks Evert. Thanks everyone for joining us. And we catch you on another episode of the Range Have a wonderful day. Evert Kraav: Thank you.