Recall Sessions: Why Two Finance Leaders Are Ditching Excel for Claude Code | Jeff Cobourn (Gusto) & Rohit Divate (Tide)

0:00We've got a board meeting tomorrow, investor meeting tomorrow. A year ago, that would have been two weeks of work, five people updating a slide deck, updating the backups in Google Sheets and trying to get the right data in and reconciling all the pieces. And today, our team is creating pull requests in GitHub against an HTML file that is on board back. I want like a company Jarvis, right? Ingest everything and I can ask it through voice, you know, give me the answer of all the meetings you ingested, all the Slack and all the Google Drives.

0:42My guests today are Rohit Devadeh and Jeff Coburn. Rohit recently joined Tide as VP of finance after seven years in corporate finance and strategy at Gusto, and Jeff is approaching nine years on Gusto's finance and strategy team. Between them, they've spent the better part of a decade inside one of the functions that determines how a hyper-growth company actually scales. The finance stack is being built in real time. Today, we're getting into what that actually looks like from the inside, how the role of finance is evolving, how teams are using AI to transform the work, how that's shaping headcount decisions, and where the actual debates are happening from token management to build versus buy.

1:19Let's get into it. Jeff, Rohit, I've known you guys for a long time. I've been wanting to do this. I feel like finance is the center of everything right now, given the amount of kind of tension and spend that's kind of going off the roofs. At least that's what's been written about these days. I'd love to just see what's happening in your world. Maybe we can start, you know, Rohit, Jeff, both of you guys are finance leaders in your own respective domains. Tell me what's happening in your role and function and how it's changed over the last few years as AI kind of hit the road. Yeah, I can go first if you'd like.

2:03So I think, you know, when I think about AI, I don't think it's really accelerated until probably the last three to four months. I think even before, like if you look at the last two years, we've had ChatGPT, we've had Claude, but mostly ChatGPT. And it's been used for writing, research, but not necessarily doing anything like really finance centric. And I think it's because the models haven't, you know, you couldn't really use it for modeling purposes. It was more like writing and research. But over the last three months. I would say it's gotten quite crazy in terms of the change. And I predict it's going to continue going on like that, at least for the next six months, year that I can think of.

2:51I think even thinking beyond that, I don't think anybody really knows. What has changed specifically? Is it just how the models have just gotten better? Or is it like, you know, we've played with it the last few years, and now we're willing to lean into it more? Like what is what's going on? Because, you know, for the last, you know, few decades, you know, the spreadsheet was king, like what really changed here? Yeah, I think for me, personally, it was, you know, using cloud code that change. And the fact that you can throw in a model, talk to it, and actually model things out is is what the big changes because before you put something in chat GPT, you kind of were just getting like a summary of it.

3:33You couldn't really do anything with it. You could just ingest whatever you put in there. But I think now you can actually start coding, right? You can start coding. You can start building things as a non-technical person, as a non-engineer. You can write to databases. So that's the more exciting piece is you can use it as a partner to build things in the finance domain that you couldn't do before. So I think for us, a year ago, let's say, Our day-to-day was using Google Drive, was using Microsoft Office, right? Those were our main tools. And using SQL, as most of our team, you know, have no SQL, and using Redash to get data out of our data warehouse.

4:13And that was kind of our core toolkit. We were in meetings, creating presentations, you know, creating models and spreadsheets. And today, all of that is gone. And 90 % of my day, my team's day, is just using Cloud Code. And so all day, every day, we're running multiple sessions and accomplishing different tasks. When we get a question, we immediately go to CloudCode and we're leveraging the tools there. And so when you think about the recurring processes versus like the one-time or project-based, the recurring processes have changed massively. We've gone from, we've got a board meeting tomorrow, investor meeting tomorrow.

4:54A year ago, that would have been two weeks of work, five people updating a slide deck, updating the backups in Google Sheets and trying to get the right data in and reconciling all the pieces. And today, our team is creating pull requests in GitHub against an HTML file that is on board deck and has all the same functionality because I built the Google slide commentary. I built edit inline and in HTML, so you don't need to do a pull request to edit long work. And so it's massively, massively changed, you know, those recurring processes, and then we're leveraging those for the projects as well. And our instinct is just we're using Cloud Code, drafting up a new session, leveraging skills we built, and that's our day to day.

5:39So it's been night and day, and I think I can really remember like the day that it changed, which was February 3rd. February 3rd, we all had access to Cloud Code and everything changed after that. What is, I mean, are you even, you know, going back to the spreadsheet anymore? Or is this like, what's, you know, spreadsheet was like the killer tool for finance. Like, when do you still use a spreadsheet then? We're not really. So we were migrating on the referring processes. We're migrating all of our models and financial models that were in Sheets into Python back web pages, right? that are hooked up directly to our data warehouse.

6:19So those are kind of, we're still using some of the old ones in spreadsheets on a recurring basis, but we're moving those in. So we're building with Cloud Code. And then any new projects, any new analysis that we're doing, it's a web page. It's not a spreadsheet. But it's just, you know, just to double click on this, like I know there's like BI, you know, just, you know, there's BI infrastructure, you know, whether it's, Tableau or whatever companies are using to sort of provide sort of FP &A analysis for various types of work streams. Is that just what's going on with BI then? So we were using Tableau and Redash.

7:01Those are kind of two main tools. Redash is really easy for like quick access for SQL queries and dashboarding. Cloudflow can work really well within Redash and can build dashboards, but But the much richer experience is building an HTML building webpage, hosting it on either switchboard or streamlet or Google Apps Script and sharing that with folks. And so I've replaced all of what I use Tableau for in an Apps Script dashboard. So my daily dashboard is going to pull up every day. I built it in with CloudFode into an Apps Script and I don't use Tableau anymore for visuals. Now we still have it, but I've rebuilt all of my needs.

7:41I have a question about that. What does that mean for the role of finance today? And I'm being more abstract. I mean, obviously, there's different functions. So maybe you have a point of view of FP &A versus being what the underlying functions. Like if you're a young finance professional and you're kind of like, you know, listening right now and you're like, oh, I spent years on learning about modeling and so forth. What skills do you think are sort of really foundational for the function? And what do you think is like no longer relevant? Yeah, I don't think that goes away. the foundational skills because i think jeff and his team and when i look at myself and my team we can do the things we can do using cloud code because we know how to build you know we asked about the spreadsheet it's like you can think about how to build things because you've done it in spreadsheet a million times so you know what the scaffolding needs to look like what are the relationships what are the lookups need to look like now you're not building it the same exact way in code you're more or less telling cloud code but you know you know how data needs to come in how it needs to be structured, how you want it to build.

8:50And I think that's one big fallacy that I see everywhere in the news right now. When I look at like some, you know, bunch of companies saying, oh, we're going to, we're going to basically get rid of Excel, or we're going to reinvent something new. It's like, well, you can do that, but somebody still needs to know how to build things. And that only happens if they knew how to build things. Because if you never like went behind the scenes and did something like you're not, you're not really going to know if it's right or wrong, at least initially. And I think you can fix that in the future. But right now, you need to know how to build.

9:21So if you're young, and you're jumping in, I would say you still need the modeling fundamentals, like you still need to think about how to build something. But the key thing I think you do differently now is you got to pick up personally, if not to work a subscription for cloud code or, or codex or whatever it is, and just, and just play around with it as much as you can. I kind of break it down into like there's skills and attributes, like the skills are changing. The attributes have not changed, right? The attributes of being curious, having a high attention to detail and thinking first principles that has not changed.

9:58But that in fact, like those skills and those attributes have become even more important. I think from a technical perspective, it's expanded because you now need to also know how to learn and use cloud code and be super proficient at building skills and checking its work and knowing what questions to ask and how to interact with it. Do you think just finance teams are, when you think about your own, the own org of finance right now, there's all of this sort of conversation about, hey, we can do less, we can do more with less people. I mean, that's always been kind of the goal. And I know that finance particularly, even just, you know, a lot of the legal functions, GNA functions have always been kind of strapped with resources just in general.

10:42Now that you're in this sort of place of doing more, I mean, you just said two weeks to do a board deck. Do you really believe that you can do more with smaller teams or do you feel like the responsibilities of finance is now going to increase because now you have increased productivity because of what you can do with AI? Yeah, so I think what I would say is like the, there's supply and demand. So supply of hours, right, is what the financing provides, right? We're doing, we're doing much work, we're adding value, we're asking questions, we're pushing the business. The demand on the team is, comes from the number of products we have in the surface area, right?

11:20And so while we're, a lot of the repairing processes for us are getting much easier and taking up a lot less time, the surface area continues to expand, right? The demand on the team is growing because product roadmaps are accelerated, right? And so as a result, I think the tension there, what we're seeing is our teams are just able to add value and cover more areas that we previously maybe didn't cover as much as we should have or give it as much support. And so that's what we're seeing right now is that tension of more and more demand coming from the business because of accelerating roadmaps on the product side and accelerating work on the go-to market side.

12:01Yeah, I'd agree with that. I've never been in a situation where somebody has, when I've been in finance, somebody, like an executive, come tell me, hey, the finance team is doing enough analysis, stop. There's always more and more and more we can do. So I don't think it necessarily makes you do more with less people. I think eventually you could kind of get through the curve. You realize, okay, I can do a lot more. Our team's doing a lot more. maybe we do need more people to do even more than we could. Because right now, I think a lot of the work, at least pre-February or onwards, it was a lot of like, let me find this information.

12:36Let's figure out, let's go back, look in the data warehouse, what's going on. We spend hours just trying to figure out to get the information to do what you want to do. But now I think in the future, it's more or less at your fingertips. So you're ready to go when you have a question. So you can answer it much faster and you don't need to spend all that time trying to figure things out. So I think you do more. And I think you also probably have more people doing even more things, depending on, you know, your business. Like our business, like there's a million things we can hit at at any given time.

13:08I don't think I will stop hiring. I think each person will just do way more than they have been producing right now. Do you think like this idea that is being at least, you know, discussed among her, you know, at least in the VC ecosystem of like, oh, well, you know, previously there was a certain amount of ratio logic, you know, if you're just looking at your own, there was always the kind of logic. I don't know, you know, when you described Jeff, the surface area, there's probably some sort of like, you know, a ceiling of like, Hey, once we get to this surface area, we need to go hire another head gallon.

13:40Well, now there's tension to be like, well, now you can have, you've, you've eliminated certain buys of work. So I'm trying to understand that. I know there's always more work to do, but what is the ceiling now of what is possible for a single resource or a single area in finance? And by the way, this kind of applies to other departments too, because I'm sure you're going to engineering and saying, hey, why do you need five more engineers? Isn't, you know, cloud code just making it easier to, you know, do more with a single engineer? Like, how do you reconcile that? And, you know, you guys are doing it across your own personal departments, but you're also advising other, you know, business leaders across the organization to sort of think about this?

14:21Like, how do you all think about that? I think we're in the midst of an insanely fast accelerating curve in a transition period where there's a lot of people who may not get to the point where they are that 10X and near that 10X finance person, right? And those people, I think the difference in impact that folks are gonna, that we're gonna see from across teams is going to widen. And there may be a transition period, right? But would I rather have 20, 10X people than five? For sure, right? Because the business will move faster. The surface area will expand faster. So I think we're in a transition period as we learn what the limits are, and it's frankly changing every day.

15:05Like what someone could do two months ago is very different from what someone can do today. And I think that'll continue to change, and I think it's more about the attributes and the skill sets of the folks you have in your team and whether those are well-suited for the new environment that we're in with all the tools at Elbow. Do you feel like there's any functions that are just going to be in, and we'll talk about finance particularly because you guys are closer to it, that are just going to be completely eliminated? Or just any jobs to be done that you're like, hey, obviously there was a specific function you just mentioned.

15:37Hey, I spent two weeks, which I do every quarter to work on this across five resources to build a board deck. Now I can do that in a matter of hours and minutes as you guys built the infrastructure for it. Are there other functions like that in sort of full resort roles that are just going to be eliminated? I know we're pontificating here about some future state, but how are you seeing it? I think when I look at this, it's the middle layer, the people and the roles that are essentially doing a lot of the scaffolding work right now, coordination work. I don't think you have as many of them anymore.

16:14Like, because you don't need all that, all those people outside in the business trying to connect the dots because you already have that. If you were looking at roles, you know, I would say customer success is one. You know, you have now chat agents, voice agents, we're already seeing that. And I think when you look, talk to investors and board members, you know, one of the key questions they keep asking is like, you know, when are we going to use or, you know, how are we using AI to reduce this so we can increase our gross margins? Like, are you using voice agents? Are you using chat agents, email agents?

16:49So I think that that one is the first one to go. But down the stack, I think across every team, the connectors, the people who are not necessarily doing things in general will be, I think, impacted quite a bit. Because as Jeff said, And somebody producing 2X versus 20X, you're going to take the 20X because they can actually get things done. I think there are. I totally agree. Those coordination roles, I think, are going to change. I think there's also a merging of roles. And I think it's probably something we've seen for the last couple of years, but is now magnified. When you think of an ION role, right, instead of financing, that's something that as companies get bigger and bigger, perhaps in the public company, it's a little different.

17:30But I can speak for ourselves. like we used to think we needed someone who would just do IR, and now I don't think we need that. I think another example of that is business intelligence and kind of data visualization. I think folks who were just building dashboards in Tableau and just writing SQL queries, I think those roles are changing. I think finance and different teams are taking over a lot of those responsibilities. So rather than us being blocked by, you know, ABC team to go build a dashboard or get us a query like that, that's gone. And so I think those roles are merging very much across teams when you look at like data, what we call kind of project management or insights and operations, finance.

18:15I think there's a lot of merging of roles. One of the things that's being discussed right now a lot when we're talking about headcount and, And the adoption of AI is just how token usage is kind of skyrocketing. I mean, you all are doing it. And I imagine there are more extreme use cases across the organization that's also using tokens in a very aggressive way to streamline work. How are you all approaching that sort of tradeoff of – because it's a new type of – and now agents are getting – is having its own life of its own. how are you thinking about this from managing spend, making trade-offs between, you know, token you should spend across different departments, agent, new agents spend, budgeting for, like, how do you plan for this in your planning cycles?

19:06Yeah, I break it down into there's, you know, internal tools and internal filter demand. So the demand that I'm putting and the spend that I have, my team has, and then there's the products that we have that are demanding tokens, right? On the product side and the external facing side, I think the pricing models have to adjust and you've got to incorporate that into your gross margin and your cost to serve. Internally, I think it is a productivity enhancer. I think you have to look at it like an additional people cost, right? It's another cost of your people and you just have to account for that when you look at your total budgets.

19:45Are you just, obviously everybody, departments had software budgets, maybe it got down to an individual level. Is there a rule of thumb that you all have? Or now are you just asking the departments to sort of say, hey, set your... People had always had software budgets. Now there was an AI budget or is all just kind of bundled together? How do you all think about that? Yeah, I think we're definitely moving to a world where there's, in addition to headcount, you have tokens and you have that cost. Similar to software budgets and things for different teams. I think it's pretty similar to that. And I think, you know, we're not in a place where that's really stabilized, where we really know where that's gonna settle.

20:25And so we're doing our best to forecast and plan for that. But right now we're gonna see, you know, a really wide dispersion from what people are at advantage. So there's some average that we're gonna settle into depending on kind of model cost and that kind of thing. But that's what we're seeing. So you're just right now for the case of budgeting and I'd love to hear what you guys are doing here. is you're just sort of looking at historicals, forecasting some sort of usage patterns across different departments and saying, hey, for every headcount, is it just an average across different departments?

20:56Or like, no, we think sales is going to be using more. Obviously, engineering is going to be using the most versus other departments. And you're just kind of blanketing a percent of overall spend, which previously headcount was the largest, now against token use. Yeah. So here's my perspective on this one. So cost is one, right? Token cost is one side of it. The other side I always think about, or any cost in general, it's like, what am I getting for it? How are you increasing the enterprise value? I think a lot of that is unknown, especially for the internal spend. So for me, when I think about spend, it can be department-specific or it can be individual-specific.

21:37One person in a separate department, finance, can be spending probably way more than somebody in engineering. And how do you measure that ROI? I think it's hard right now. It's really hard because it's all anecdotal. It's all person to persons at this stage until we have some kind of measurement system that says you put this many tokens in, this is what you got out, this is how we measure the ROI value. And I don't think we have that yet. So I think until we have a system like that that says what got spent, what was the ROI, I think you basically just say, okay, here's a heuristic dollar amount that I'm going to give you.

22:12Maybe it's 500 bucks a month. Maybe it's a thousand bucks a month for token usage per person in a particular department, or maybe it's more for engineers. And I think you just have that there as a budget because you also need stability in terms of forecasting costs. And then as you see demand go up and down, you adjust that. And I think you have to be able to adjust that almost on a monthly basis or even faster than that. You have to be fluid, I think, in this, I would say, at least for the next six months until we get some kind of scaffolding that can measure ROI. And then we can say, you know what, this person is spending$100 ,000 a month.

22:47That's fine because they're doing everything. They just refactor the whole code base to work way better. We'll spend that money there, but we don't know right now the true ROI, I think, of all the token spend that's going on. Are you guys requesting this, just out of curiosity, when someone's making a request for additional headcount or budget. How does AI fit into the ask here? Is there something you all sort of systemically doing to sort of say, hey, do you have an AI strategy in your department in a role? Or is it just an assumption that that's just what we expect everyone to do now? So at eight, what we're doing, we have AI fluency scores across every person across the whole company on the same grid.

23:31And so that's done monthly. by managers. And what that tells us is which teams are leveraging AI and which team they're not. But I think the teams that are not have a much higher bar to get incremental investment, right? Because the first ask is, okay, how are you leveraging AI and how is your team increasing its impact with AI first? Jeff, how are you measuring that? Yeah, I'm curious about it. Tell me about Footsie. This is really interesting. Tell me how you guys are doing it. So we've got a grading scale across five, and every company, every person in the company is right on that. There's a first cut of that that AI does as an input that the manager then reviewed.

24:15And so the manager has input on that based on the impact that they see. And so right now what we're doing across the company to better demonstrate whatever those impact is, is we have a lot of sharing going on. So every week we have an hour with every team and we're, or at least for our team, the finance team, where we're sharing different projects. If you stand up and you're sharing projects and other people are leveraging your skills that you've built in our marketplace and other people are excited about the learnings that you have, you're clearly, you know, amplifying the work that you're doing and the rest of the company and you're the higher end of that scale.

24:52And so that rating scale is done monthly across our teams, and which gives us real-time input for distribution across individual teams and across the company. And so it's... It's not self-reported. It's like you're actually like analyzing sort of actual usage behavior. Not usage in path. Like what have people actually built, right? And each manager... How do you know, like, you know, you know, a CX person, you know, is building something like, is it, how do you guys, do you guys build some internal tool to do that? I mean, it, it kind of all fast days, right? It goes into our quarterly performance reviews.

25:36So we do performance reviews on a quarterly basis. One of the key measurement measuring points is AI fluency that every manager is expected to have for everyone on their team. Oh, wow. And that's all well done. So, but a manager basically determines that score. but do you have an input to sort of say, hey, there's evidence of that actually happening or is it just sort of like the manager has observed it and they have a deeper understanding through their, you know, sort of managing that individual? It's the same way that we do performance reviews, right? You calibrate, you make sure that, you know, across teams, across managers, we're being consistent and that's brought up.

26:15So the inputs to that is looking at the impact of the work that they're having. you know, and that we talk about that every quarter. Interesting. Are you guys doing anything like that? Not yet. I think we're a little bit earlier stage. So I think this is the thinking that we have to do. So when you ask a question like, you know, when somebody's coming to me for headcount, what do we do now? It's certainly not something that we're deeply asking every role. It's like, what is the individual AI strategy? I think we're thinking more from a company perspective of, okay, what does next year look like?