Latest / Range Podcast / Treat AI like the hire you'll never get
Transcript
- Giac Soliman: Welcome to Range. I'm Jack Solomon and this is the podcast for competent reward people building with AI without losing their grip on the pay decision. Every episode, one practitioner, one honest conversation about what AI can actually do in our world and where you still need a human in the room. Be curious, stay a little skeptical, and here we go. Good morning, good afternoon, or good night, depending on where you're in the world, and welcome to a new episode of The Range Podcast. Today I'm with Teresa Cortez. Teresa is a tool rewards leader at Nirvana Insurance with 20 plus years across startups and enterprises. She's been experimenting in AI tooling lately and recently, building what the team needs instead of paying six figures for software. That does the same thing. Teresa, I'd love to open with a specific question based on a chat we had before. How long do you think you could sleep while Claude does the work for you without sort of wondering if what it's doing is what you asked for? Theresa Cortese: I'm not ready to completely trust it yet. I watch it make mistakes. I look at the output. I check everything. But the point is, it takes so much less time to check and validate the work than it does to do the work, to build out the spreadsheets. And this leaves me with a solid audit trail. If I build these tools correctly, I can go back and trace every error. Giac Soliman: That sounds like the right approach. you've been experimenting with the i tooling, right? I came across your LinkedIn profile, we said, Hey, we should exchange notes because we seem to be, you know passionate about this new capability we have that doesn't seem that anyone is publicly building with yet. Now there's a few more people building and showcasing it on LinkedIn. It's great. But what's the what's the first thing you built? Was it like in your personal life or even professional? Theresa Cortese: The first thing that I built was a, I'm still proud of it, a multi-tab merit spreadsheet that over the course of 15 years of merit cycles at retail companies, semiconductor companies, ed tech companies, HR tech companies, evolved into a highly functional machine to help with decision making. So this was I still remain proud of it. I'm never gonna give it up entirely, but I am pleased to say that after discovering AI and other system tools that can do the work, I have finally stepped back from using it. Giac Soliman: when was this? Like give us a little bit of context. Theresa Cortese: I built this 15 years ago. and over time as Excel added functionality, I added more complicated functions. There was a time I added macros. it just became really a a semblance of a compensation planning tool. and until I discovered the compensation planning tool I'm using now, I used it every merit cycle. Giac Soliman: Interesting. I wonder where like you can decide how to spend your evenings or your afternoons. Like you could watch Netflix, you can go for a run, you could play a video game, like you decided to build with the eye instead, right? And where do you get inspiration for your ideas? Like is it when you go to sleep? Is it when you take a shower? Like how how do these things happen? Theresa Cortese: In a different world, maybe I'd been born five, six years later, I probably would have gone into engineering. I've always been fascinated by the concept of building, of coding, but it's simply not where my training lies. So once my company made the tools available for me to bring that into my wheelhouse, I jumped on it. That's where my inspiration is. After a frustrating day at work, I unwind, I I have ADHD. So I will be watching Netflix on my phone to the left and working with AI tooling on my laptop on the right, doing two things at once. That's that's how it how it works best for me. And this is Giac Soliman: Rings a bell. Theresa Cortese: most evenings. I unwind and this is what I'm doing. Giac Soliman: Did you have any mentors or coaches? Like did anyone pull you into this or did you like did they show you how to Theresa Cortese: No one has shown me how, but I have the most amazing team at Nirvana that is so psychologically safe and so excited to experiment right alongside me that I ha I feel like I have a a sounding board and a room full of people to talk through ideas with and hear what they're doing. And some of the work my coworker Sarah is doing is inspiring me. Will something similar work for total rewards? So it's It's very nice because they have access to the same tools I do. But yeah, no mentor, just we're figuring it out all together as we go. Giac Soliman: I mean, as someone building a community precisely for that it just echoes a lot. It just resonates a lot. I seem to recall the first time we met that you were building a variable comp management system and that you were waiting to test it against April actuals. We're talking about sales actuals. This was I think maybe three months ago, maybe a little bit more. How did that go? are you still building it? Is that is that tool coming to fruition to life? Theresa Cortese: The way I like to think about things that I build is what version I'm on. So I'm shipping the V1 version now, meaning that I've talked to our IT team, shared my Git repository with them, and they're moving it to be hosted for live. I did the Q1 actuals testing and it passed for every employee. I've set it up to run 110 tests each time it makes a change to make sure that we're not messing up anything downstream. And after it passed testing, we were ready to move to live. So in Q2, we are primarily going to use the tool to calculate our payroll benefits. Admin will probably still run it manually just as a safety net for a cycle or two and then let it go. Giac Soliman: I another question to follow up to that comes to mind because when you said you think about it in versions, it it brings me to another hot topic. there seems to be a bit of a imposter syndrome feeling going on, you know, across our domain. People are building, but they're just not talking about it. And Maybe if they thought about what they're building in terms of work in progress version one, two, three, four, like there's no expectation that over like overnight you build a fully fleshed and tested enterprise grade end to end compensation system. Maybe if that wasn't the perceived expectation, people will be more feel more comfortable. in sharing even if it's just work in progress. Does that like align with your experience? Theresa Cortese: Absolutely. I can speak to what shifted my thinking on it. as you know, I rescheduled our call because I was sick with the flu. And I'm not gonna lie, I was relieved. I I was ashamed and embarrassed to show you what I built because it it isn't it isn't spiff. It isn't these tools that people go out and buy. But I saw a post on LinkedIn that was talking about women in technology and how there's such pressure to For everything to be perfect, that people will pick apart anything that they create and diminish their existence in the space. And so in order to avoid getting caught up in that thinking, the suggestion was realize you can continue to iterate. Just get what you're doing out there, start talking about it, start showing it, and you can benefit from the feedback either way, but don't view what you're putting out there as needing to be perfect. Nothing is ever going to be. It will actually hinder your progress. And that was eye opening for me. And at that point, I got the courage to ship my tested tool to IT and tell them, let's put this on a hosting site so it can be used across the organization. And gave me some of the courage to talk to you today. Giac Soliman: Like building with AI and feeling that sense of imposter f syndrome that it's not good enough, that you're putting yourself in danger, is a little bit like public speaking. Like I still hate it, but and yet I feel like I have to do it and I'm so happy when I do it because it's sort of I've changed myself and then it goes well. And you have this feeling that you're gonna go on stage and you think that people are there to listen to you. is there a like an analogy there that might r might resonate with an audience that has done public speaking but hasn't built anything with AI yet and they could use that as an analogy to get started. Theresa Cortese: I don't have any words of wisdom or analogy. I can say though that having done both, building with AI is so much easier. It's I I I've spoken to people who say, I haven't messed around in Claude Code yet. I'm I'm scared of it. Claude code is easier than the other AI tools. And the stuff that you build is so much more functional. Like the complicated and convoluted AI workflows that would be necessary to do a lot of the work in the total reward space. happens much more organically and much more quickly when you actually build a Python tool to do it. Giac Soliman: Teresa, why the heck did they call it Clot Code then? I mean that's probably the worst name ever, right? Cause it's not about code. Like it yes, it codes, but it doesn't require you to code. Theresa Cortese: It doesn't, but you know what? It's teaching me. It's it's kind of funny. When you watch the way it thinks and what it looks like, it teaches me how these tools think about the process and how they approach testing and validation and where they don't approach it correctly. And I need to directly address it with them. It's really been stimulating for somebody who always was curious about that world and simply doesn't have the training to actually dig around in there. Giac Soliman: Counterpoint, are they actually thinking these tools or are they processing and associ Theresa Cortese: The processing. Giac Soliman: associating words with other words they think are more likely to be associated with your words? I I don't I know a generally qu genuine question. Theresa Cortese: That's a great point. They're not thinking. They're processing, they're calculating, they're using algorithms and logic to put together a result. But that's also what my Excel Excel spreadsheets have been doing all this time. The piece that requires human input and insight still belongs to us. We are still the people who guide the work. We're still the people who review it with a trusted, vetted lens and add our understanding. If I didn't Deeply understand all 11 of our company's variable compensation plans, I would have been in no position to build this tool because I knew what to look for. I knew the logic behind what gets calculated first, what gets calculated second, what factors go into it. Without the human understanding, AI is not equipped to process this itself. Now, the way I approach working with AI is I approach it as Pretending I actually got a P1, an entry-level headcount added to my team. It's a tool that I can teach what I'm doing. I can guide it in processes and workflows, but it's not really ready to run off and make decisions on its own. It brings me data that it has calculated, processed, and output based on the criteria that I have entered into the system. And I can coach it and teach it for the future. So when I find an error, I say going forward after you take do X step, then do Y step. it it really is similar to coaching a human. Giac Soliman: That's fascinating. And one of the other questions like keep coming up is it sounds as AI can save you a lot of time, but in my own experience I actually find myself working longer hours. And if that happened to you too, is that a defect? Is that like a behavioral problem? Or is it something else? Is it meant to be like that? Is it because you're so productive that you are almost drugged into want to do more and more? Is it is it becoming addictive or is it is it due to all the back and forth because you're not probably prompting it right or is failing to retrieve the right context, the right time. Like in your experience, is it creating more work? Is it saving you time? Both? Theresa Cortese: great question and that goes back to how I approach working with AI. I look at it as I got approval for an entry-level headcount, meaning that just like with a human, it takes more time to teach somebody the process and the logic behind it and then let them run with it. It becomes a time saver, but not in the first few months. And that's the case with AI too. It's taking more time. I built an equity refresh calculator in Python using Claude Code. I had to do the actual equity refresh while stopping each step and it putting inputs into Claude Code. I was doing them in parallel. It took quite a bit longer than it would have. However, when I then ran the tool afterwards, it brought the same result. I had just spent a week calculating in less than a minute. So I will see the time savings next half. When I calculate the next cycle. But no, for now it takes longer, but it's worth it. It's I am a total rewards professional who likes working with startups and who never sees myself working with a non-startup in the future. At many startups, there's not even a desire to hire one total rewards person. There will never be a scenario where they're hiring a team under me. This is a unique opportunity for me to train up. Some help for my function to actually take some of the day to day off of my plate so I can be the strategic partner to the organization that I've been trying to be. Giac Soliman: And why do you say startups hesitate so much in investing in tour's function? It w Theresa Cortese: It's an afterthought. I I until you're at about 20 people, many startups don't even bring on an HR function. When they do, t they may join a PEO so they don't have to think about total rewards. startups tend to pay differently than enterprise companies, much more cash light and equity heavy. And there's When there is a small headcount, fewer than 50 people, the CEO can have a much more hands-on approach to making compensation decisions. They can all be very bespoke. Right around 100 employees is where companies start grappling with that question: should we exit the PEO and bring things in-house? Is the CEO ready to hand over some of this day-to-day decision making? And that's really become my sweet spot: is joining these organizations and guiding them through that transition phase. AI has it is not 10 times my bandwidth, it's 50 or 100 times my bandwidth because I'm able to be doing these extremely manual labor-intensive tasks that I've done in the past at the same time that I'm working on something else. If yes, I have to go in and provide inputs to Claude Code throughout the process, especially when I'm building enhancements or iterations, but in between. I can be reading Slacks and sharing my thoughts and insights. I can be reviewing decks for upcoming meetings and providing much more deep and thoughtful insight to them because I'm not so bogged down with the day to day. Giac Soliman: Or I could just make myself a coffee and chill on the sofa. Right. I mean it's also Theresa Cortese: yeah. Giac Soliman: potentially I I wish I do I was doing it, but potentially you could do that as well, right? it it's fascinating, right? 'Cause I think AI is really turning the tables around and providing us an opportunity to change the perception of our function, compensation and benefits. I've worked with startups too, and it's just They they are they're rushing to get done a thousand things, right? So understandably not always they see rewards as mission critical at that stage. Could be series A, Series B, pre-seed seed, who knows. But I think that based on the huge technology enablement for people that are non-technical in nature. like me, it AI is really giving us an opportunity to change the perception and bringing perhaps you know that that point where reward becomes mission critical, us. And that I wonder if that's that's what might happen in the future. The more we get AI savvy toll rewards experts that are able to work with startups in such a way that they see the value Much you know, much before the what we typically see. Theresa Cortese: Yeah. I think that we're seeing for the first time technology enablement in functions that historically have been kind of ignored in this space. it has always been a struggle to get budget to buy tools for the people team. there's there's pressure to prove ROI, and it's challenging to prove ROI in a non-business or non-revenue generating unit. So we've constantly had an uphill struggle, but when That obstacle is taken away from us. When we see a real-world business challenge that's costing the company money, five people at high levels in the organization having to spend a week of their time every quarter doing different pieces of the variable compensation calculation because we don't have a tool to lift that off their plate. We're not really big enough. For an enterprise-wide tool to make sense. So we're at a funny spot where it's costing the company a lot of money, not really enough to buy something out there and established. All of a sudden, we're empowered to solve that problem. We look at and say, what why? Why are we struggling with this? Well, it's because business operations, who owns the calculations, the metrics, can't see pay data. So we're maintaining two separate spreadsheets and validating them against each other every quarter. It's Perfect storm for human error to occur. It's tedious. It's taking an undue amount of time from our director of business operations, our controller. These are people who their time would better benefit the organization doing something else. And they're stuck doing this manual data entry. So by opening up access to AI tools and actually giving me an opportunity to build a tool like this, now BizOps can just go upload them and they're done. They're done with okay, the achievements. They they hit a hundred and two percent in this in this component and they hit eighty-five percent in this component to to know and to trust that it can just be calculated because the pay data is safe and secure and only the people team can see it, but inside of the same tool, that's been a long-term problem that. That's the the companies have made their entire existence on trying to solve these bigger these spiffs and and calculation tools. Giac Soliman: It sounds like you're talking from experience to some extent. And I we we mentioned the total variable compensation system that you built most recently. I wonder if that those were some of the drivers that led you to build such a you know a audacious tool. we were talking about sales quotas, target achievements. I get there's a lot of calculus, a lot of you know judgment, a lot of politics too that goes into that. Could you tell us a bit more the story around that that tool and Theresa Cortese: Yeah, I think it really came to to light during the planning cycle for the Q1 variable compensation. So we had the goals, the team that works on it every quarter got together and sat on a Zoom call for 45 minutes talking through what's our plan, how are we going to calculate this, and Are my I was I was asking, why can't we put it into a single spreadsheet? Is there a way we could put it in there? Is a way there way that Google can hide access to a single column? And it it can't. It's not a functionality that it allows. If they have edit access, they can unhide columns. And so on the call we were talking through it and and I said, This is why these tools exist. Let's look into it. Let's see if there's a tool that makes sense for us. But when we hung up that call, I sat back and I thought back to the directive that our CEO had given us at the All Hands earlier. He said Experimenting with AI is not optional. And build something this week. I'm like, build something this week. What am I gonna build? When he said that. And it it's sort of thinking about that call, thinking about doing an RFP for for a tool, it just sort of dawned on me. I was like, you know, this is the build. Is this even possible? I went to Claude, Claude Chat, just basic LLM, and said, Is am I have I lost my mind or is this something I can build? And it said, You really need to buy a tool. And I said, Well, if I have access to Claude Code, does that change things? And it said the phrase that just keeps resonating with me. That changes everything. And I'm like, that changes everything. Claude code has changed everything as far as putting the power in my hands. I didn't have to go to anyone for budget. I didn't have to ask. I just was able to do. They at our company they talk about a bias towards action. And what I appreciate is they Actually stand behind that. They're like, okay, we want you to take action. We want you to have bias towards action. So here are the tools. Use them. Do it. Just do it. We'll we'll figure it out later. So that's that's that's how it came to be. I tried it out. I ran test scenarios until we tested it against live date. I just wasn't sure. I wasn't sure it was gonna pass. following the Q1 cycle, it I I almost felt paralyzed. I'm like, it did pass. I have to, I have to show this to people now. I have to let the rest of the company use it. And yeah, that's where we're at. Giac Soliman: Speaking about action, you got me really curious about what this tool looks like. Would you be comfortable sharing like a just a few glimpses of what exactly we're Theresa Cortese: Absolutely. Giac Soliman: talking about? Theresa Cortese: Absolutely. Let me share my screen and I am going to walk you through kind of the basics of how it comes together. Giac Soliman: if you have a Tory Wards background, presumably one will be familiar with the key concepts, right? Theresa Cortese: Yes, absolutely. these are gonna be familiar terms to anybody who's been tasked with calculating or managing variable compensation in a commission style format. we don't have commissions directly, but it's it's similar. So it's not pretty. I'm not a software engineer. I I'm not a UI designer. I put together something that works, that is intuitive, that people can use to build a calculation or to use to run the calculations, but Prior to this project, I had never opened the terminal on my Mac. So I'm totally new to this. here I am, put it together. First thing is, I built it with Google SSO, and that's at the database query level, meaning that there is no possible way for somebody to see data we haven't given them access to. We assign people to roles, and the roles can only see certain pieces of data. So when we give employees access, They're going to be able to see their pay and their pay history, but there's no possible way for it to accidentally slip up and show them things they shouldn't. I have fake data in here for the purposes of demo, but to just give you an idea of a few different styles of plans that we have. So here's our dashboard. and this is where you can get the overall view of the organization. Metrics is where our business operations director can input the actual metrics and attainments and the quotas. They're entered here either manually or she can drop in a CSV upload, which is what is likely to happen. download a template so it's in exactly the correct format for upload. Calculations are run here. This would be run by either me or our payroll and benefits administrator. They run it, can view it, can export it, can look at the calculation history for each person. So I've been running it a lot during testing. So you see diff the same people multiple times in this list. If I want to compare one employee's payout. between two different runs and see if anything changed, I can go to the compare runs function and look at two different runs and see, okay, what changed, what didn't change. So that is helpful. helps you, helps us to maintain an audit trail and to be able to go back and trace what happened and why decisions were made. We are able to generate letters that talk about their payouts and what metrics went into them. This is a new feature. We haven't done that in the past. We do compensation letters at the beginning of the measurement period, but not at the end that talks about how their payout is calculated. And administration. This is the behind the scenes piece of it where I can take a person who's entered into the system and change their role or add a new person into the system. I am able to record a comp change. This is important because our merit cycle is effective two one every year and our quarter begins on one-one. So everybody receives a comp change in the middle of a measurement period. We can either input variable comp as a percentage or as an overall dollar amount, which is important because we have both use cases here in the organization. And all of this that you see that we can manually enter can be uploaded as well. So we have the option to either pull a CSV upload or to go in and manually make tweaks. Giac Soliman: Wow, this is the kind of functionality that you would expect in a enterprise grade, a hundred thousand dollar contract with with a vendor. Like I I'm impressed. And you said you iterated a lot, of course, in this, but this is probably version what? Version number Theresa Cortese: This is probably version three of how I view it. I mean, I do little changes at a time when an idea comes to me, like, huh, we need to be able to add this feature. And so I stick it in there. But as far as big, if you think about it in terms of your your phone updates, this is the third, like 13, 14, 15, but there's been dot z you know. One four two along the way. So I do little iterations along the way, but this is probably the version three of major updates. Giac Soliman: Gotcha. And and all you needed to build this was a bunch of Excel sheets with the data a or files connected Theresa Cortese: No. Giac Soliman: to your Google Drive? What? Theresa Cortese: No, all I needed to build this was clawed code and my long-term experience with compensation and variable compensation in particular. I just needed to talk to it and explain it has to be able to do this. It has to be able to do this. yeah. Giac Soliman: Do you have any advice for your f previous self? Like working on V one? What are the things you would tell Teresa four months ago? For building this. Theresa Cortese: I would say stop thinking about how you're gonna start using AI and just start doing it. Just cross that line. It's so much easier than you think it is. And really, the other piece is I would have told my earlier self to think about it the way I'm thinking about it now as a as a as an entry-level person because. There is a lot of stuff that still needs to be verified. My experience and total rewards is still a hugely important piece of how this, how this works, how every tool I've built works, because I'm the only one who can look at it and go, no, no, that's not right. That's not how that should have calculated. That's not how we think about it. And going back to the tool and coaching it and teaching it would have saved me a lot of time rather than. hanging my hands up in frustration when a workflow doesn't work the way I think it was. Sometimes the workflow is not the right tool for it. Sometimes you need to take a different approach, but AI is enabling me to do more. It's not doing things for me. Giac Soliman: I love that philosophy. It's just an extension of your brain or almost Theresa Cortese: Mm-hmm. Giac Soliman: like an intellectual sparring partner. But then you're in charge, you're making you're cutting shots, you're applying judgments, not Theresa Cortese: Mm-hmm. Giac Soliman: the other way around. Right, we are close to wrapping up the end of the podcast, and we all know AI has potential, a lot of potential, but no one knows exactly where this is going. In a year from now, what do you think AI adoption will look like in compensation teams? Theresa Cortese: In one year, I think we're going to see a very wide gap between people who are practicing compensation in the way we've always done it. These larger, older, enterprise level companies are not gonna be as quick to adopt changes and processes and workflows. There's too much red trape, there's too much bureaucracy, there's too many steps that it makes it makes quick change impossible. I The way I think about working with startup versus working with an enterprise level company is working with a startup is like building the airplane while you're flying it. But working for a large enterprise level company is like trying to turn a cruise ship. It's very hard and it's very slow. So I think in one year's time, we're going to see two completely different types of total rewards people. The ones who are using AI to increase their bandwidth, the ones that are able to operate in a space that requires it, that requires them to be able to do more with less time, with less Less monetary investment. But those are the compensation leaders who are going to be part of the strategic direction of the organization because we will have the space and the time to actually research and weigh in on these hot and but hot button topics. In five years, when the enterprise grade companies have had time to catch up, I think we're not going to see that slower function at all. I think it's going to take that long for the base expectation to be that. Total Rewards practitioners are also experts at leveraging automations, workflows, integrations, AI, all of these different tools that save us time. Giac Soliman: I agree. We've always been building. We've just been building on more, you know, rudimentary tools like Google Sheets and Excel. But now we have tools that have a little bit more potential. And that could definitely shake up things in the industry and change the perception of the compensation function in front of the business. That's a wrap. If this was useful, follow the show and send it to the one other comperson you know who needs it. Thanks for listening and I'll see you in the next one.