Latest / Range Podcast / From a mirage of spreadsheets to one budget model

Transcript

From a mirage of spreadsheets to one budget model · Range Podcast

• Giac Soliman: Welcome to another episode of The Range Podcast. I'm here today with Neilan Vaug. Neilan works in people systems at Zip, where his focus is increasingly shifting towards people infrastructure, governance, and AI-enabled workflow design. He builds internal products across people and culture that turn fragmented judgment heavy work into more practical and government decision support systems. Neilan, great to have you today. Nilan Vagh: Chuck, it's really, really good to be here. I'm very super excited to be having this conversation with you today. Giac Soliman: Before we get into the whole whole AI thing, which obviously I'm super interested to know about, you know, the World Cup is being, you know like teams are playing at the moment. You know, that there's matches every day. And this podcast Nilan Vagh: Yeah. Giac Soliman: episode will only go out in like a month time, so by then we will probably know who will be the winner. I'm really Nilan Vagh: Yeah. Giac Soliman: curious to know who do you think will win? Nilan Vagh: No, I'm a massive Ronaldo fan, so it's it's the last dance for a lot of you know, fans that grew up watching Ronaldo. So I I really do hope. Portugal take it. you know, I think it'll be great close to his story. He he's he's an inspiration not just to football fans but you know to other athletes in the world. I mean his Suez celebration transcends, you know, even football. You have like people in NFL doing it, tennis doing it. I mean you name the sport and you know he's up there as like one of the greatest athletes of all time in my in my personal opinion. Giac Soliman: Ronaldo, if you're hearing this, there's a people operations and systems leader that's Nilan Vagh: Hmm. Giac Soliman: a big fan of yours. I'm sure this message will get to him at some point. do you know it's he's the first one, the first human being to play six World Cups in history. Like there's no other player that did that. So that's probably yes, testament to his perseverance Nilan Vagh: Yeah. Hmm. Giac Soliman: and endurance. I can barely walk the stairs, you know, go up the stairs for five floors now. that's interesting. I I I have a feeling it's not a great feeling that actually France will win the World Cup. Nilan Vagh: Yeah, they they look very, very strong. Like I they're they're definitely one of the Giac Soliman: Yeah, it it it's fun. Nilan Vagh: top teams. Giac Soliman: It's funny 'cause when you ask people who they hope it they're gonna win the World Cup, they all say the same, like, yeah. I hope France doesn't win it, but I have a feeling that France will Nilan Vagh: Mm. Giac Soliman: win. So it's quite interesting. but going back to your story, Neil and you've been in the space for in this space for a for a while now, and I'm really curious to know what what made you start building with AI? Where Where did do you have that like light bulb moment when you said, I need to start doing this? Nilan Vagh: I would I do consider myself one or an early adopter. So I think I use started using ChatGPT when it was a big thing. Like I I'm talking about like four years ago, you know, when it was a bit taboo to use it at work. So I'd be using it on my personal devices and stuff and then slowly and slowly, you know, as as all the models got better and stuff, I think, you know, it was just the exploration, you know. Originally it started out with like, you know, how can I get information? And then it was like, wha what are the opportunities for improvement and then from improvement led to like, you know, it was just more research like, if I needed to automate X, Y, and Z, how can I do it? So, you know, one of the things was I built a our A and Z intelligence workforce dashboard which started out as to like I need to automate this mm monthly manual process that my team is doing which it takes them day and a half because they're pulling information from two through two to three different systems and how can I automate it and that's when you know based on my research I found you know if all the information is there on Google Sheets then you know Chat G BD told me just use Google Apps Script. I'm like, what's Google Apps Script? I kind of fell into like you know the the initial vibe coding stuff. I've seen another term floating around is like vibe engineering because I'm I'm very you know throughout my career when I ever build processes I I do it from a you know what is what is the problem I'm trying to solve and I apply a dis Design thinking lens lens from I think about the end user but also the process along the way. So that's how Giac Soliman: Right. Nilan Vagh: I've approached all of my internal tool building as well. It's not just a prompt. It's not just, hey Claude, build me this app. It's I want to be able to build this app. This is the context behind it. This is what I want to be able to do. These are the different users I'm thinking of, and I have all of these things firing in my head. And you know, I guess AI is allowing people, teams to be able to create tools themselves. But what I really love about it is you're not you're not just building a tool, you're able to apply your own company's context to it. And I think that's what makes a lot of the tools that I'm seeing in the market a lot more fascinating than than what SAS sells you because SAS will try and sell it's whatever SAS or our HR Tech Stack is built, it's built for the medium, I would say. Like it it has to support 90% of the company's processes. But every company you've worked in, whether you're doing you know, you've done a thousand of RM REM reviews, but if you've done it in like six different companies, there's always that you know little nuance that you have to do a bit differently in this company because there's a thousand people and it requires you know five different approvals, versus if you're working at startup where it's just a CEO signing that REM review process looks so different and you Giac Soliman: Right. Nilan Vagh: know you you could be using the same REM review tool in those five companies but the context changes every time and what AI will allow you to do is kind of build that layer on top which allows you to capture that and that's that's what's really gotten me into building a lot of internal tools can I build a tool which may, you know, make the reliance on on whatever we're paying for a a lot more cheaper, but also how can I build something which has the context of the company I'm working for, which kind of makes the process a lot more easier, not for the just not just for the people team, but for the leaders that are involved, for the employees that are involved and for senior leadership. Giac Soliman: Right. I mean the answer to that question, can you build a tool? Of course you can. Nowadays you have tools like Cloud, Codex, Google Eye Studio, etcetera that can build stuff for you. The question is, should you build? Right. And I I really like what you said there. Like when you decide to build something, you focus on the problem, then you think about the desired outcome and you And you reverse engineer that and you came up with this fancy term prompt engineering, which I haven't heard before. How do you apply design thinking and prompt engineering to what you decide to build? How does that get baked in the decision? Nilan Vagh: I think it's all starts for me always is you know going back to the initial problem and then it's also like who are your different stakeholders involved? And you understand from the different stakeholder perspectives of how are they currently interacting with your REM review tool and you know what are the different things you're hearing about because at the end of the day they'll come back and say you know let's say people partners will come and tell you well we want to be able to do this bit this bit and this bit and then when you go back to your your account manager and saying can we do these adjustments and they'll say either they'll give you two things like that's not possible or it's in our roadmap it's in our FY27 Q4 roadmap so keep an eye eye out for it and by the time that roadmap comes you've you've already moved on so I think it's that's kind of understanding like if I were to build this myself You know what problem will I try to solve right now and also how much time and effort it goes because as I said, like a lot of people like doing love doing the vibe coding thing, which is just the high-level stuff is build me this beautiful thing that does this. what what will happen is Claude or whatever tool you're using will kind of build you something cool based off what's already out there. There, but the whole point of you try build a tool yourself is not to replicate what's out there, it's to take your local context, apply it that works for your company. You try to build tools, a lot of people try to build tools that will work for their companies, and I think no one has better context than you in your company. If you've been Giac Soliman: Right. Nilan Vagh: there for four or five years, the amount of knowledge you have, like no HRIS can replace. Light stretch. Giac Soliman: have to ask you this follow up question because it's so timely. Given all you've said right now, right, makes me wonder, do you think vibe coding will kill SaaS? Do you think that HRS, ATS, and others, you know, SaaS vendors that have been around for decades, do you think they will s like disappear or what mm Nilan Vagh: No. No, I think I think I think SAS has its place, especially when it comes to our world. you know, in the HRS space, like you we are driven by a lot of legal a lot of laws, you know, when it comes to holding the data, making sure it's private and stuff. Like you can't vibe code a super secure, you know, app Like it takes meticulous planning and a lot of effort and you know people just think it's front end engineering and back end but there is a lot that goes behind it like originally when I started in Google App Script I used to have just two files which was like a GS code and then index code which would be the web page and now a lot of my files look different because I've started understanding like you can't also have like 10,000 lines of code in your vibe coded app because it will slow down down it will break or it'll cause issues. So I I I at least for now I don't think vibe coding will replace S. but it is it is vibe coding has you know really reduced the barrier for people to give it a go themselves. Giac Soliman: I fully agree with you. I don't think we're about to see a SaaS apocalypse, Nilan Vagh: like some of the new HRS tools that I'm starting to see where I also feel like it'll become like a container where you still are able to store the people data that you need to, but then effectively within the tool itself you'll be able to wipe code your own things, but at least it's in that container which is secure. our our good friend Matt partners with or always a big proponent of shapes and so I've seen them do some really cool stuff and I think a lot more HRI scopes will go down where it's less about the modules they're trying to sell, but it's more about what the tool can do inside. Because I think people want flexibility as well and you know, vibe coding kind of gives them the flexibility. to do things themselves but then how can you do it and this is what I come back to is like you can you can try and do build a vibe coded app but like how secure is it like you Giac Soliman: Yeah. Nilan Vagh: know and I think that's one thing that freaks out a lot of the one of the reasons why PNC or HR has been very slow to vibe code is because we we handle so much sensitive data that we don't want to risk putting the data into a tool and then it gets leaked and then you know you can be up for law Giac Soliman: Yeah. If if you send the wrong like pay increase, the wrong toll reward statement, maybe it requires a signature, that says a contractual amendment, someone signs it, then you figure out, God, we made a mistake and by the time you realize they went to their trade unions or their labor unions, it is not very easy to backpedal from that situation, not really Nilan Vagh: No. Giac Soliman: to unwind it. So I think caution is definitely present across the entire HR spectrum, but so is it is excitement and and like being conscious about the risks is great, but it's no excuse not to get started in my opinion. Cause it you have a new capability now. So why not explore it? Just Gradual start with a gradual approach, right? Crawl, walk, and run. I think that would be the ultimate power move if HRES or any SAS vendor is able to show to in-house practitioners look it's more convenient and not just convenient, safer and more effective for what you want to achieve. If you vibe code inside my application than if you do it outside, you know, clawed code. That definitely like we would see a new Yeah, a shift in behavior happening. And it's it's it Nilan Vagh: Yeah, absolute absolutely. Giac Soliman: Are you are you starting to see like more players going in that direction? Like giving in house practitioners the tools and the freedom to customize the tools like for real. Like you can vibe code inside your platform inside the platform and and just attach or append new features that only belong to you. Like main platform belongs to everyone. Mm-hmm. Nilan Vagh: Yeah, I I've seen yeah, I've seen obviously shapes. I've seen a few others that are doing something similar in this space. I mean, you've got your big Sasses, they'll probably do it like the work day and Like they'll they'll have to get in, otherwise they'll start looking losing market share. But then I think it's the smaller companies that have that opportunity to really, you know, dominate this space early, which will be inter interesting to see. And and I think on the flip side is like a lot of this also opens you up to risk as well. I think a lot of companies, you know, whether even y you know, especially if you're working in tech, you know, your cybersecurity department will be asking for like line And lines of like, can we get this done? You know, get this, make sure you have this in the contract as well. I think a lot of those type of things are starting to become a lot more you know, stronger, there's a lot more scrutiny right now when you're buying software as well, you know, especially when it comes to AI. And I think the biggest fear that a lot of companies have, w you know, at least from a cybersecurity perspective, is you know, if this all has AI, are they gonna use our data to train their models? Because that's that's the other thing is like, you know, data is like currency, right? Like Giac Soliman: Absolutely. Right. And especially when it comes to compensation data, I think the stakes become even higher. I mean, yes, employee data itself is heights stakes on its own. When we're talking about compensation data, knowing what people are paid, how they're paid, you know, that that governance aspect becomes even more prevalent from a data residency, data privacy standpoint. To some extent I think it's it's to be welcome, this this new wave of vibe coding, because it it helps vendors focus their attention on what they're really good at, building secure software, to your point. But it also helps in house practitioners fill in gaps that would otherwise be filled by spreadsheets or the old ways of doing things that we've had, you know. been doing for more than thirty or forty years. I think Excel came out in eight nineteen eighty five. Right? Nilan Vagh: Yeah. Giac Soliman: and one of these one of these folks that I'm referring to that are building or vibe coding in the compensation slash remuneration space is you. It's one reason why I invited you to this podcast, right? And I'd love to learn more about Where you find the need, where what what did you build? What did you wanna to do with AI? Looking at the past, but also maybe later we can you can tell me a little bit more about what you wanna do next. Nilan Vagh: Yeah, so we had back in March, we kicked off our REM planning cycle with chat up, so that's the tool we use for REM review. with our rewards lead. So obviously I look after all of our systems. It was a rewards lead who is actually our parent to leave cover. so for upcoming cycle and you know, we got like, you know, we wanna be able to kick off in June to be able to launch in in August. We we run like a eight week process where Chathop helps build the entire REM review cycle. So then she turned around and asked Chathop, well, do you have budgeting capabilities so I can, you know, run the budget scenarios and then present it to the board? And they were like, no, we don't have this and then, you know, I I kind of told her, don't worry about it, I'll I can I can build something for you. I had built a a a workforce engine. I also built a calibration engine prototype which was you know it's gonna support our performance review calibration process. So I was like, yeah, let me let me do it. And then you know, I also wanted to be able to ensure that she's not stuck in shades. there was gonna be her first process at zip and you know, she's stepping in mid-parental leave and I wanted to help out, you know, you know, and give her competent. confidence and you know it's it's it's exactly what you said like it's not another spreadsheet you have to figure out and I think that's the biggest shift for me underneath was I think people teams can actually build when they understand the workflow deeply. Giac Soliman: And what was the first pain point you wanted to solve by building with you know, w w essentially Google Apps scripts? We can get into that in a little bit. I think you're a huge fan of Google App Scripts. Nilan Vagh: Yeah, I do I do predominantly all of my builds with Google Apps Scripts because again Giac Soliman: Mm. Nilan Vagh: it's a safe container. We're you know, we're a Google organization, so we use everything in Google, which means automatically If I try to share that whatever link outside to even to my personal email, I'm I I'm forced to log in to Gmail to be able to access that link, which means automatically there is at least a level of security that whatever I'm building is not gonna leave the organization at a minimum. I didn't start straight away with the code, you know, it it started with a conversation, you know. After that meeting, you know, I sat down with her, you know, she walked me through previous cycles that she had done at other companies. She was able to secure the documents and the sh spreadsheets, mirage of spreadsheets that was used last time. You know, we kind of understood what the normal total rewards REM cycle process would look like at zip. You know, I I mapped the workflow, I understood the logic. know, found the gaps, found the manual pain points, and then that's when I started building in ChatGPT and Codex as an accelerant, but they were not the architect. At the end of the day the business logic came from her and the workflow, not not from a prompt. I've actually built a a prototype for a a a fictional SAS company of two hundred and fifty people. So as you can see, what I what I ended up building was we started off with what was supposed to be REM review budget modelling and STI summary. But then mid builds, 'cause she was doing compensation analysis, which then would help her understand where she should set the budget. She's like, in the tool, do you think we can add a compositioning analysis? And I'm like, Yeah, absolutely, why not? Because we're using the same data set. And so effectively now it it flows a lot better. So as you as you see here We've got compositioning analysis. we've got some set business rules we're saying, you know, technology and product will get you know at market P seventy five. Whereas like all other divisions get at the midpoint. You know, you're able to look at how many employees are benchmarked, what's your average cop ratio, you know, an indication of how much correction is required and so on. And then you're able to look at red means danger, amber means it's getting there, whereas green means all good. And then effectively it gives you an action. So, you know, this is like to me the magic. moment here where originally you'd you'd you do this work in one place and then you'd have to pull those data points and then build it bring it to your you know rem modeling that you want to do whereas here I'm gonna press model and then it automatically feeds into My REM review scenario, again, you know, for demo purposes, I've got two pre-built scenarios and it's already telling me from the compositioning tool for the 26 employees, this is the correction cost, and we're suggesting an increase of 3.7% of the payroll. And then I can apply this to the cohort and then boom, it's modeled. And then what I've got here is a planning snapshot in terms of what the total budget is, how much you've got for reserves, how much is available for the different portions from a rim review, promotions, and then I've applied a waterfall approach in terms of you know you can clearly see how much reserves you've got, how much you've spent, and then how much budget is remaining in green. I I like it to keep it very visual because I think leaders respond to things visually, where if they can see. Something it makes a lot of sense. I've got a bit of impact analysis here where you can look at the impact of the REM review from a division perspective, a job grade, entities, we've got a bit of a summary explorer, Giac Soliman: All the breakdowns. Nilan Vagh: yeah, all the breakdowns. And then to me, a really important thing which allows this tool to be defensible is an audit trail. So you know every time you're modeling something, it all gets recorded in the tool. So one for your own memory you can go and see, what did you actually do Where sometimes you can lose track in sp spreadsheets because especially Google Sheets because it just keeps updating in real time. This allows you to have an audit and look back of, you know, what was the per increase did you do? You know, did you do any other adjustments and and so on. So I think that's really cool. And then another thing I was able to do is kind of a comparing two different scenarios side by side. So you could, you know, model one which is more conservative, and then you've got your market adjusted and then what are the differences, you know, and then you can do and lastly then you can you can then go and actually do a drill down at an employee level if you want to. And then that's the REM review modeling and then you know we built an STI summary which for these purposes we kept it like very super simple. We we assumed at a at a hundred percent. But what I think is going to be a future iteration is we've now have so much data of how we've actually applied STI in the past and we can go back up to like the last four cycles and start understanding that, you know, we've spent eighty percent of our budget at ninety percent for people that were rated at a three and twenty percent was for the fours and the fives and everyone else got like nothing. And then you can kind of bring That modeling here, and then apply it to your current data set, and automatically you'll have a much more better, you know, from a budgeting perspective, and then now she'll be able to take all of this data and then almost like bring it into chart op so you know we don't have to do a lot of work more a lot of rework with chart op because we've done the modeling and kind of know where we want to be, and it also gives finance kind of a a lot more. more confidence in the data we're providing because it's very you know, I think I've built a very defensible tool, but you know, I think this is only iteration one. I think there's gonna be lessons learned based on when we actually run the process to see how far were we off. And then, you know, again, I'm all about iteration is like how can we further improve this so then the next time we run this, we're even better. And you know, would I build an end-to-end REM review Probably not, you know, as we both discussed. but I think you know we we could get there where we could run like you know, maybe people partners can run the REM reviews in this tool so then by the time when it comes to run it in the real cycle they're all kind of have an idea. And you know, it's also understanding how much more data can we bring that actually impacts a REM increase, you know has someone been in a performance improvement plan, you know? That those type of little little decisions does make a impact and sometimes all of that data is sitting in somewhere else. So I think that's again the beauty is like how can you surface Other layers of data which often don't sit in the REM review modeling tool because it's often just holds the REM side of things. But you know, there's there's the full full 360 when you think of performance in REM because they're all they're obviously very much linked. So how you know, eventually I'd be I'd be wanting to build more and more data. So again, it it makes it a very robust and easy decision. Giac Soliman: Wow, this is impressive. I I do think a tool like this, I do think it would be helpful if I were a people partner, walking into a REM review, like having had visibility to a tool like this would help me get prepared f you know for that pay review. Would would help me feel less that I'm walking into the dark and then I have one week or two weeks to help my managers. figure out what the right salary increases are. So well done for putting this together. what do you want to build next? What's that one thing that you still complain about to your partner when you had a very tough day at work and you said, if only I could do this d a different way. Nilan Vagh: Yeah, there's just so many things I think, you know, you know, 'cause 'cause I sit in the intersection of not just total rewards You know, I I'm I'm I'm people operations all around. There's like certain things I'm looking to build. From a purely from a total rewards perspective, it's almost like a total rewards intelligence center, which is able to capture again, because data sits in different teams, especially equity data, for example. You know, for people partners, if we're Giac Soliman: yeah. Nilan Vagh: able to pull what what does Nylon's total REM look like? And total REM isn't just, you know, my salary, but like my salary, my potential STI, how much invested you know, LTE I probably have remaining is probably in important data set which only that employees across. and then if you if I and then what I wanna be able to overlay that is with the performance data And let's say their most recent risk rating. And then have again with these different data sets. If you had these different same data sets for your entire team, you almost will start having a heat map of like you have this cohort of high performers have like no vesting LTE left, which means they're at a higher risk. So whilst they might be very happy, they now have no future LTE to look towards. We may not have a refresher. And that doesn't mean we'll give them more LTE, but it just means like this person could potentially start thinking about their future because all of a sudden when equity drops off, like there's nothing to look forward to. So I think I want to be able to again, because we've got like a convoluted tech stack, like a lot of companies at our size, you know, we w we've just bought. As we've gone, a lot of these data sets live in different points. I want to be able to bring it in one layer, which that's what I want to call. Like it's a rewards intelligence suite, which kind of gives you when you're clicking on because at the end of the day, like everyone that works works for money. Like, yes, you want you know you have your own values and stuff, but like at the end of the day, you wouldn't be working hard for a company if they weren't paying you right, like, and when when When you feel undervalued or you don't feel you're being paid right, that's when you start looking for roles elsewhere. So how can we service this data a lot more quicker than the moment when someone comes and says, I want to resign, and then you have the people partner running to the total rewards person and saying, What can we do? Can we improve this? Like it's it's a bit too late at that point. Is so what I want to be able to build is like this, and it's something that people partners can frequently monitor. you know once once I'm able to bring AI within within the tools that I built because a lot of these tools that I've built is purely based off like no AI it's just rules it's just rules and data I think once I'm able to overlay AI and then it starts looking at insights and starts looking at the future data and stuff I think that's when it'll become a lot more robust but even even without the AI like having all of this data and having a heat map that you can just dive deep if you wanted to. It'd be I think it'd be really massive for people partners and not just for people partners, but senior leadership as well. So then they kind of get an understanding of how their cohort is sitting there. Giac Soliman: This is a concrete example of what we mean by when we say AI can help like can enable us enable the business and ultimately can help us improve the employee experience. I think this is precisely a a concrete example of what we mean by that. Neilan, thank you so much for coming on the show today, for sharing your story, how you started building, what you built, and your inspirations for the future. It was great to have you on the show. Nilan Vagh: Yeah, Shark. I've I've enjoyed this conversation and I hope a lot more people, you know, take the plunge and and and realize like you now you the barriers have been removed. You now have the capability. Like there's no other better time to be a a total rewards person or a people person than right now.