Financial Modeling Will Never be the Same (w/ Tarun Amasa of Endex)

Today we have conversations with some of the world's largest consulting firms, and they're asking what is the world after PowerFront Excel? And that fundamental shift is only possible today because of how good the technology has gone. Totally. Well, you're too young for Clippy, so that was pretty revolutionary for me in the 90s. Now the private equity fund cast is about to begin, so give it up your host, Devon and Jim. Hey, everybody. Welcome back to the fund cast. We have another very good one for you and deep in the weeds. Everybody knows that AI is coming for private equity workflows and banking analysts' jobs, at least if you read the papers. We disagree 100% and think that actually it's going to make people's jobs easier and a lot of them to do more fun stuff and be more productive. But you don't want to care what I think about that. You're going to care what our two guests think about this. So from Parker Gale, the returning and reigning champion Caroline Fipps, as we call her round trip Fipps because when she puts her hands on a deal, there's always a round trip coming. And Caroline's been an associate and a VP with us at Parker Gale, so welcome, Caroline. Thank you, Devon. And we have the CEO and founder of index, Taruna Masa, who we've gotten to know over the last few months because Caroline and I and a lot of people on the team have been demoing every single product you can imagine that sells AI enabled solutions for private equity workflows. And Excel is one of the big things, right? Claude and OpenAI are claiming their products can make Excel sing and do lots of amazing things. And we've actually kind of have stumbled on index and really like it. So that's the answer to the test a little bit. But we're going to tell you why we like it and not just pitch the product. We actually want to go deep into what's going on with private equity in banking and AI. What is changing? What is never going to change and how technology is having a big impact at. So Taruna, welcome to the fund guest first time guest. But I imagine we'll be talking a lot of the next couple of years. Yeah, great to be here. Thanks for the opportunity. So let's get into who you are, Taruna and why you decided to start a company. How did you get here? I'm sure your lifelong dream was to like start an AI Excel business. But given that AI didn't exist when you probably were getting started, like what's the path? Yeah, I had a pretty rude awakening to excel back in high school. Spent a lot of time with state governance actually. I'm working on funding for different nonprofits and education budgets and all of these different budgets were run through Excel files naturally. One of the things at the same time of going through these budgets was I was learning how to code and I felt like the tools for engineers just fundamentally felt so much more ergonomic. Right? Over the last 20 years, the tools and the way that someone has coded has changed as the technology has evolved. But these spreadsheets, some of which that we had to pull up from 2004 or earlier than that, had looked largely the same and rarely did someone ever look into these files or audit some of these financials. I didn't have a driver's license at the time. So my dad drive me to some of these nonprofits to kind of do kind of my own diligence on where they were spending some of these funds. And some of them would report these crazy linem's on cogs in S&A and they'd go and they didn't even have offices. And it was really confusing to see how these types of policies and government initiatives sometimes had great purpose. The actual outcomes were never delivered. And they got me thinking like, were we bottlenecked by the tools and the ability for people to do a real work. And ultimately kind of was the impetus to starting index, which is a plan in words to end Excel work. And which we think is the ability of bringing the tools that exist for software engineering for folks and finance and beyond to really harness the power of this technology. Love it. So can we go back a little bit? Why in high school were you working in government? And how do you, is this like, was it, uh, it's like a hobby or was this something you were hired to do? And how do you get that job as high school student? Yeah. I think one of the biggest things for me was learning really early on about this kind of principle called the Pigmaling effect. And the thesis behind it, it was a study done by Harvard later debunked was that when you had teachers that expected a lot out of students, the students would perform a lot better than teachers who'd expect out of a similar cohort, but with like negative expectations that they wouldn't pass a test. This kind of idea of like, what can external environments do on people's success and how you can push them to do incredible things was kind of a through line. And frankly, the inspiration to start in this company, could you bring together a group of exceptional people to create exceptional outcomes? And through that, one of the levers that existed was kind of school budgets in state fund rate, right? Colorado had this really weird law at the time called table, which prohibited kind of different school districts from reallocating spending toward kind of teacher initiatives or school budgets. And they'd have to go through this arduous process of trying to get bills approved and all of these different kind of red tape mechanisms that were put in purpose to kind of support more observability for taxpayers, but had kind of the reverse effect. And so I ended up writing a bill kind of worked on getting that proposed legislation and then it didn't pass. And instead, a lot of my time like in the government, like as a board member trying to figure out like where could we kind of have different leverage in the funding that we already had? And you'd go through all these Excel files and then there was just these pockets of funding that just weren't ever accounted for. And it was a weird experience as like either a 16 or 17 year old at the time trying to talk to some senators and being like, who is looked at these Excel files? And you just kind of hear a little bit of a shrug a little bit, right? That it just should work and like just hope that it actually ties. But like functionally excel on these tools just don't have the ability to allow you to audit in the same way that you have for a code base or other tools. And that was kind of the impetus of can we make the tools themselves smarter, right? Excel has looked the same way for the last 30 years, right? Every few professions are like that. And this tool is the backbone for so many parts of the economy. And so initially I was like, do I want to go down this government route? And ultimately kind of felt that this entrepreneurship path was more of an expression that kind of have the ability to do it. And when we started the company, the technology wasn't really there yet, right? And there's kind of this uphill battle of these outcomes that you wanted to drive. But that kind of overhang of actually letting people believe that there's a world where we are without Excel or PowerPoint the same way. Today we have conversations with some of the world's largest consulting firms and they're asking what is the world after PowerPoint Excel? Totally. Well, you're too young for Clippy. So that was pretty revolutionary for me in the 90s. But let's do this. Caroline, you have looked at, I want to say, 100 different AI enabled workflow automation tools and plugins for private equity. I mean, and maybe we go back to like last fall. Like what were the things you were seeing? What were the pitches you were hearing? What were people telling you that these products could do? Yeah. So back in the fall, I demoed several products that had kind of the same pitch of covering your entire workflow as a deal team member or investor relations member or sourcing member of a private equity team. And to me, in the end, all these tools seem to go not deep enough and any of your workflow in any of the areas of your workflow. So it was essentially, you know, they would say when a private equity firm sees a new deal, we will ingest the new deal. We'll review it. AI will create a summary of it. And we'll potentially be able to push that into a presentation, you know, in your firm's format and then contract the deal throughout. And maybe there's sprinkles of being able to help you with diligence, connect to the VDR, ask questions and find gaps, which is all helpful. But in the end, I felt there was, there were still a lot of gaps and it wasn't going deep enough in any of the areas. And there were solutions that we were already using with ChatGPT and Claude that kind of covered some of the areas. Then I guess the other topic is if, you know, some of these tools would summarize materials that you would receive and then push that into a presentation for you to present to I.C. But then there's a question of what are the materials actually? What's the purpose of this? If the purpose is to present it to I.C. and talk about whether or not it's a good investment, shouldn't you be the one thinking about what's on the slide? What am I presenting? I need to actually study the company to be able to discuss it. So in the fall, it was hard to find the value in what was being pitched. It was very cool that AI could do all these things. But I think at the end of the day, many of the tools were just scratching the surface of of what we actually do and how we actually use it, that has evolved for sure. - Let's stay on that like last fall timeframe. We'll bring Taroon in here to talk a little bit about it too. So it definitely feels like about nine months ago. So we're recording this in this late spring of 2026 that the jobs either did things that you wanted to do anyway. Like, oh, we'll just summarize this for you. It's like, no, no, actually I need to know it at a molecular level. I don't want it summarized. It's like, and two, it just couldn't finish the last mile. So it was either too high level or it just never finished the job. So no workflow or project or effort was fully finished. So it was like, well, why would I do this if I kind of have to pick it up along the way and help it along? 'Cause now I don't have the context 'cause I'm not in the whole thing closing the loop. So we basically decided like, let's sit back and let the model cook. Like, we'll just use Claude and OpenAI ourselves and get really good at them instead of buying third party products. Taroon, as you're thinking, we can fix this Excel problem and we can give it more power like code. Obviously a place to go is private equity and banking and other financial services where it is literally the code equivalent that people are in all day, everyday making stuff. So let's stay in that fall timeframe like before February where everything changed when the new Claude came out and everybody started using co-work. What were you seeing then and kind of what were the positives and negatives of kind of the pitch and how you guys were getting received? - Yeah, I think as a, as a company, our DNA has always been to under-prompts and over-deliverer. And I think that that has helped a lot as the tools have gone much better and the underlying models have been able to do work and excel in PowerPoint. They weren't able to do nine months ago, right? As we kind of think about the way that we see our lens of the market is to be heavily focused on a very narrow set of very valuable work, right? I think that tools that try to go like Caroline mentions a mile wide and inch deep are probably just going to get replaced by a chat should be T or Claude. Our goal is to be highly focused on a narrow set of things that we think we can do very, very well. And I think it requires you to pair both a team where we're split between folks from private equity and people from engineering. And this kind of hybrid kind of approach to really focus on like how do you do an ARR cube correctly, right? Like what does it look like to do a cohort retention in a way that a private equity firm looks at it, right? That type of nuance kind of exists today as we think about the overhang of relative capability of the underlying LLM model and the application what you use it for. And the term that's canonically used for this is now called harness engineering, right? And Claude code was kind of one of the first like breakthrough harnesses followed by co-work and some of these other coding agent tools and our goal is to build the right harness, which is almost like an Ironman suit for the LLM model to do work in Excel and PowerPoint directly for finance professionals. - For those of you not following along closely enough, it's no longer about the model, it's about the harnesses are more important these days. So let's talk about some use cases. Like let's go deep into a banking or a private or use case because a lot of the people who listen to this are thinking like, okay, that's great, but how do I use it? So Caroline, how do you take somebody else's product and then get up to speed and dive into it? - Yeah, so in the past without AI plugins, what we would do is take a model sent from a banker on a company and we would pull up our ParkerGale templatized LBO and then retrofit it for the information that was provided to build our own model. So we would be looking cross-referencing constantly between the banker model and ours, but ultimately we would want to use our own template and file. This of course is a little bit manual with bringing in new tabs as source tabs, making sure everything's linked correctly. Also time periods are always something you have to update no matter how good the template is. You're not updating the template every month. So you always have to update the time periods. You have to update oftentimes like for the, whatever source file you got, it could be at a great level of detail all the way to the invoice detail or it could be the summary P&L annually. You're doing a lot of manual work to get to what we need as an output. I would say with index, now what you can do is one, you can have index be the doer of that manual work. You can bring in the tabs you need and then have it update the model and or you can have it be the reviewer of your work, which is something I have absolutely loved in the tool is that you can make updates. Often times, now that I have seven years of experience, it's like pulling in the tabs, linking everything up, it's pretty quick. But then having the second set of eyes, not even in another person, but just in the AI tool, to be able to tell you whether or not you linked everything up correctly is so helpful. So that's one use case. - Yeah, I wanna have to run react to that. But here's the thing is what we usually do is with new associates is we try and convince them to print out the model, do it and then print it out and stare at it and then show it to Brian Milligan 'cause he will immediately find the cell, 20 pages deep in the model that looks off to him, right? So now index is now the new Ryan Milligan. He doesn't have to print out the model and like have him stare at it and be like, this cell looks off, which every analyst who's listening to this and every associate's listening knows they've gone into some meeting after spending all night and all week on a model. They think everything's perfect, they've checked everything and the MD looks at it in two seconds and says that looks wrong and then your heart sinks. So, Trune, I know this is a problem you've wrestled with because if you can't do that, you're not gonna get a whole lot of traction with an analyst or associate, right? They're gonna say like, forget it. I could do this better myself. So react to what Caroline said and how you'd started building the model and getting it to a place that it could do those kinds of things. - Yeah, it's a really astute point and I think it's like a large kind of pillar of how we think about the adoption of these tools and finance where the verifiability and being of the catch mistakes is often more valuable than just the generation of materials, right? Where a single mistake is incredibly costly and giving the driver's seat to an AI agent to do the entire model, maybe it's a higher trust barrier than using it to double check your results, right? And I think that framing has been important as we kind of think about how you build products that kind of, you have trust in. And we've seen this kind of story play out already with software engineering, right? So today for engineers, the bottleneck is not generating more code to get created. It is reviewing that code and making sure that there are mistakes. And I'm sure you guys have had experiences of using services like Spotify or AWS, where you're like, why do these tools just feel less reliable than they used to be, right? And that's just a part of kind of the fractalization of kind of the era of AI agents that we're at today. Whether or not good enough to be a full replacement for a human, they're almost there. And those slight cracks are what really like causes petrobation. And so ultimately our goal is like, how do we make communication between humans and humans better but also humans and agents? And this verification of, can you check the last mile has often been a huge lever when people have rolled out. Last fall, Caroline and our whole team would run something through a test and it wouldn't come out right. And they say, I can't trust it. It's gonna take me more time to audit this than it would be to make it. And we don't need help building models. We've got our own templates, everybody's got their own templates. We need help making sure the complexity we've added to a model actually, we didn't miss something. 'Cause we've all had that like heart in your throat moment where you're on the phone of the lender and they point out something busted in the model that actually changed the IRR by 100 basis points and you're just like, oh my God, my life is over as an associate and it's like your life isn't over but it's very embarrassing. So again, that verification. So let's talk a little bit about depth versus breath. 'Cause when we talked in the past, Trune, you talked a little bit about like the LLMs and the big foundational models are trying to optimize for generation, not distillation. And I think that's what you're getting at. So maybe go a little deeper on generation versus distillation. - I think there's two important points on this. The first is that the way that these AI labs make most of their money is on token generation. So they are incentivized, which we like to call big token. For people to spend as much as possible on the generation of outcomes. It's like, can you create a PowerPoint deck or an Excel file or a website? And these agents are now even more expensive, right? Instead of like one term that you might have with Jatch Bt, it's going to take maybe 50 turns, take a few minutes. We've seen our agents run for hours on end without interruption. And that costs more and more, right? And so fundamentally, these labs are incentivized for this generation, right? Where we try to think about as like, how can you distill as much as possible to the key core instinct of what is the underlying set of principles that you need to kind of develop? And maybe is a good segue into where we think about how we've seen enterprise CFOs think about our product, right? Which is like-- like almost like a cost optimizer. Because we're able to use OpenEye and Thropic and Gemini and other open source models, you're able to use the right amount of intelligence for each task, right? So you don't bring hammer for small precision edits, right? And that hammer often costs 100 times more, right? And our goal is how do you kind of think about the right intelligence for the right task? And that's both used for the distillation, but also going to answer the woes of Enterprise CFOs this year, where the budget for what people expect to spend on SaaS tools versus the spend on AI tokens by the end of the year, are going to be several of orders of magnitude different than what an Enterprise CFO might have budgeted. Totally. And we see the same thing, like we build our software companies, we own build on the foundational models, but then we deploy on open source models and small language models, or we're using bedrock and AWS to be able to swap between any model at any time, given the cost or it kind of, as you would say, horses for courses, right? We want this model to do this piece of it, we want this model to do this piece of it 'cause it's better and better price-proof performance. So we're seeing that across the board. Caroline, you had a story recently about a funds flow, you inherited and how you used this product, like kind of, again, what was you've done before, what you do after how did the tool help? Yes, totally. So we close on refinancing on Friday, in this case, the lender was managing the master version of the funds flow, but we always have our own version with some other wire info for closing a deal. So Thursday night before the deal closes, on Friday morning and wires get sent in the millions of dollars to many different parties. Most private equity deal team members are always stressed in that period of time, making sure all the wire information's correct, that every dollar ties out. So I thought to use NDEX on the funds flow. I had built in some checks that were my internal checks on the information I was provided. Many of the cells were hard coded, so I didn't have all the information. Hard coded because the lender didn't, you didn't have access to the information they had behind the scenes, right? They had their own internal version, they gave me a share version, which is totally common. So one of my check cells, as I was going through the logic of, I think these two, and the sources in your says, these two cells should equal these two wires. It wasn't checking, so I confirmed with the lender, there's late Thursday night that that was correct, but I just used NDEX as a second logic check because I was like, I can't see past, you know, the logic that I'm working through. NDEX, what do you think? Does this make sense that are we either excluding this dollar amount that is not checking to zero entirely or double counting it? And it confirmed that logically it makes sense, like my check cell essentially didn't make sense. The flow made sense, so that was an awesome use case. - And it was $85, but if anybody was carrying out, you don't want to show up on the day where you wire money and be $85 short on a wire. - Right. - So another use. So Terune, tell us about some other use cases as you've dug in with clients and things where they've kind of really pushed the limits of the tool and made you guys better and think about product roadmap. - Yeah, especially in private equity, we've seen use cases across portfolio ops, investment team and finance accounting, and across as categories where we've seen are doing things like building a monthly revenue by customer and product line over the last one months, flagging for declining trends and highlighting top contributors for growth. What is affected like net retention, right? Doing some sort of revenue per employee, SGNA as a percentage of revenue had count growth models, right? Those types of use cases. And I think what we've noticed is that there's often times like a creativity overhang of can you do all of these things when you see this kind of open box. And how does it kind of almost guide you to like be a thought partner? And I think that this idea of almost proactively suggesting these workflows is something that we've been really trying to experiment around, right? What we do is we work really closely with clients. Some of the things that we've seen with them is that they want to pull data from all these different systems and kind of finagle them, right? So you have your chronographs, your eye levels, and you want to pull the data from the PDFs and sync them and then use those to create a web dashboard. And it's kind of push and pull across these different systems whether it be an ERP or portfolio monitoring. That's kind of been what we've seen the most advanced firms do. Of usually index almost as a pipe between these different systems of record that exist in an organization. - And true, and in that case, is that creating for creating reporting dashboards in Excel? - Yeah, I can do both, right? Whether it be in Excel or web dashboard, we've kind of seen a mix of those, right? And I think where the world goes is that you'll start to create these more kind of evolving artifacts that are backed by these Excel and PowerPoint files. - Every private equity firm is thinking through like, okay, how do you do portfolio monitoring? You've mentioned some big products out there that the big firms use. It certainly feels like with the standard monthly financials we get from a portfolio company with an MCP server with Excel and a tool like index attached to it, a typical fund with a dozen or two dozen portfolio companies, like that might be good enough. It may be superior to a third party product. - Definitely, right? And I think where the sense of going is, how do you go step further? Can you forward an email with a financials and automatically extracts them? Or doesn't on a cadence or a mind, a company that if they didn't return an offer to the sixth day to kind of bump them, right? And I think this idea of like, almost doing these actions on repeat, where you can kind of teach these more complex workflows is where the world goes, right? And every associate ends up becoming like an orchestrator, right? Being able to make sure that things pass a quality check where taste really matters, but almost can see these pieces of work and manage it like a trained conductor, right? And so it'll be exciting to see how do you go from someone being like, let me get the MCP set up to Open Ampixel and doing the chronograph to like, I have this work that I need to get done. I want you to remember to do it on this cadence. And this is the specific way that we adjust EBITDA for this company. Can you remember that before the specific instance, right? When we look at a software company, we haircut these specific assumptions based on our previous kind of precedent transactions or history that we've had and become this kind of DNA that you can port across Open AI or Anthropic or Gemini instead of being locked down to a single provider. - So for an uncle like me, who doesn't use these tools very often, but benefits from them because Caroline and the team are getting good at them. Can we talk about like specifically how you're interacting with index in Excel? Like I've used Claude for Excel and kind of you've got it along the side and you're asking questions and things. Like how is index in Excel working together? Caroline, maybe from your practical experience and then Sheroon talk about how that's evolved and where it's headed. - Yeah, totally. So yeah, and my practical experience, it is a window that pulls up as an Excel add-in and there's two tabs, one of which is a chat tab and the other is an audit tab. So you can chat with index about your whole workbook. You can also, if you would rather chat about one tab, about a specific block of cells and ask it anything, or you can use the audit function and select tabs in which you would want audited. And then it'll give you a full audit report on the mechanics and logic of the workbook. I have found two things to be very helpful about both of these, the chat and the audit function. One, it'll catch errors, like Tyran mentioned, like it'll catch mechanical errors, logic errors, because it definitely, from my perspective, relative to the cloud, Excel plugin, it does have knowledge of financial formulas and workflows. So you can get from net income to EBITDA or EBITDA to free cash flow. And it'll tell you if there's a sign that's flipped the wrong way in one of those formulas. So that's extremely helpful. I think Tyran, you guys nailed the size of reporting it produces, where it'll give you maybe seven bullets, if it's auditing a workbook, to say, hey, you applied this annual, what was labeled an annual assumption to this whole set of months. Did you want to do that? It's making the end product better, because it's almost like, I have felt like I have another colleague who has a bunch of time and is really engaged and knowledgeable about what I'm doing and is like, yes, I'll sit by you and take a look at this and then tell you what I think, which KG and I would do that all day, if we could, if we had enough time. But yeah, for sure. But it's nice to have a second set of eyes and then you can kind of make, you just make the end work product better. That's been my experience. And so, to room, what's going on? going on behind the covers, like how would you get to that point of like it actually thinks like someone like Caroline rather than just general purpose? Yeah, well, I really appreciate the kind words. I think we've always wanted to double down on getting folks directly from industry to directly affect the behavior of these models. I think today the best way to kind of think about an LLM model that you might get from one of these labs is you have what are called tokens in a context window. The context window is the amount of tokens that you can kind of put in. A token is almost like, I'd say like around like a word or a few characters to make another answer or a call a tool. That context window has expanded a lot over the last year as well, right? So you went from in 2024 maybe able to put in like 4000 tokens, which might be like a few pages to now sending a million tokens, right? Which could be like multiple books. And what these models are under the hood are like better sponges, right? And the larger the sponge and the smarter the sponge, the more it can retain. And the more you can also shift its behavior, right? So it's almost like clay that you want to shape. And because of this context window, what we care a lot about is making the behavior of someone from private equity feel as if it's a colleague, right? And so tactically what that means is if you're someone who works in maybe like insurance or accounting, it might not feel like the same to you. And that's uniquely how we're able to kind of specialize this type of stuff, right? And with that million context, you can kind of give it the guard rail, it's kind of guided. And so all of these labs have massive teams working on agent behavior and model behavior. And like it's almost like when you use like 4.0 people were like, wow, this model was so warm, right? And then subsequent open and model was really cold and robotic. And so our goal is like how do you take these models and almost shape them and raise them like a child or an analyst or an associate to kind of model what we think is gold standard. And anytime a new alum gets released, we're not like, oh, it just scored amazing on these benchmarks. We're just going to put it in. It's like does it do the things that we would expect people would like to dig deeper on that. Do you have, I know you mentioned you have industry folks with industry experience on your team. How do you organize the actual ingestion of industry knowledge into your platform? Yeah, it's almost like studying an alien species a little bit, right? Where this thing is kind of landed on on earth or in her office and so on. And they're like, you guys have to inspect it, you know, which is like, how does it feel? Like there's a series of like kind of vibe checks that you have to do. There's just series of like evaluations that we have on a set of our own Excel files internally. Right. We hear from customers in their feedback where we're like, okay, we're going to roll out this for a small portion of people. How do they think it acts? And together you almost get like almost like a taxonomy, right? Like it's almost like you think about like an evolution chart and you have like these like 20 different variants of a Marmot. Like it's like those are the ways you do it. This person is doing like a co-er retention. We know that this species, whether it be like this model with this reasoning and this prompt will do better. Right. And it's like artisanal, effective. Like, who do we put on this? Right. And it's very similar to I'd say like a best like a world class private equity firm or sports team, right? Where like, how do you put the right person in the right place? Right. You don't want messy being the goalie, right? Something that's on X today is like a good example of that, right? All right. I want to jump into like how portfolio companies would use this because I know a lot of CFOs and finance teams inside portfolio companies will want to know, okay, great that Caroline can audit a model and move faster to give me more things to do when she goes home for the day. How is now the portfolio company betting for him? So let's put our CFO hat on our head of FPNA hat. So like, talk about like a femoral software, ghost agents, like what do these things mean to the CFO of a private equity owned business to you to run? The way that I like to kind of capture it is the prior waves of software. You have to buy, you're kind of choice, your choice to do, to buy a tool that almost fit or build something that's like perfect, right? And the cost to kind of building a tool and such an R&D spend that it didn't really justify unless we're like a large firm wanting to like really double down on it. And we saw this historically, right? I think there's a craze over the last 10 years for every private equity firm to try to build their own CRM, right? Whether like we have our own proprietary deal intelligence and we're going to try to build this thing. And then what ends up happening is the firms have built it, we lose it there. Costs of software is not just creation, it's the maintenance, right? And so what ends up happening is all these firms have built it, everyone's like, okay, now do you have an MCP server for this CRM? Like can you handle like this thing and we don't have an entire team of software engineers focused on the same thing? And so then you have this weird loop of those firms being like, why do we build this ourselves just to just bought something off the show? Now that equation shifts a lot when the cost of execution has shifted, right? Like now you can have a bunch of software agents build these things very quickly. And so now they think the real question is like how do you balance like execution speed with like standardization, right? And so what we've seen is like how do we create like firm wide organization or guardrails of like if we want to do a monthly dashboard, right? This is what it should look like. And we almost see ourselves as this kind of like kind of supporter of interest in like some level of standardization across these flows. Instead of having the Wild West of a bunch of kind of limbs and vertebrae that exists that someone built up that works point and time, but isn't like an institutional piece of intelligence. Yeah. Carolina, do you think about use cases in the portfolio? Like where does your mind immediately go to? Yeah. I think one that automation can help a lot of accounting and finance teams. So like with cloud co-work one thing that's just been a really simple solve for me has been linking it up to my box account where I have hundreds and hundreds of folders from the last five years. And I'm now able to when I know I need to find a document and I know it exists in box, but I don't want to spend 15 or 20 minutes finding it. I ask co-work to find it and it finds it immediately. So that has been really great. So little instances like that I think finance teams end up getting slogged it down by so many requests so often by the auditors, by their PE owners, by vendors, by customers, that it ends up taking up I think probably a lot more of their time than they would like. So I think a lot of the automation tools will help teams to save time and spend more time doing other things. Yeah. Cast just walked through for our annual meeting. Just kind of like how much time a CFO has in a given year to actually do strategic work. And it's little it comes down to like maybe six weeks from after getting done with the audit, getting done with quarterly board reporting, getting done with budgeting process. Monthly reports to the board, quarterly reports to lenders and others. Like there's not a whole lot of time in a middle market company where you have a pretty lean team. So I think our view is this is like a ceiling razor. Like for the best people who want to like really tap into it, like it can really be a game changer in terms of automating things and getting the busy work out of your way. And then see have more time to spend on the strategic stuff. Whereas where does this product come in like that? I think we eventually get to a place where we're doing daily closes. I think we're going to get to daily closes at financial closes if you have the right setup not now, but certainly in the next 12 to 18 months. Definitely. I think that one of the big like macro trends that we've seen is like whenever you use an external consultant or someone to do a QV, right? Like can you like start to use these tools to take off work on both ends? So the entire chunk of work that you give to an external vendor is smaller, right? Like these are pretty like resource intensive workflows. Like if you want to go to, I'm not going to name the names of any of them because some of them are customers, but go to blank and blank for them to do a quality earnings. And you're just like, okay, now that cost hundreds of thousands of dollars to do, how can we kind of take off the forefront and the end of it to organize the books a little bit faster, right? Have them in a methodical way, standardize reporting across all of our portfolio companies and kind of use these agents to kind of structure this messy data, right? And I think this point is like probably the largest macro point, right? Where like previously to use a CRM or any sort of like reporting tool, you have to fit things into like this structure that exists in the world, right? Where it's like needs to fix in this very like column, row style thing where it have to fit a specific drop down and go this way. But the world is messy, but as humans, we're able to give context and agents are able to do that. And I don't know what the form factor for that looks like, but I think we'll be paired alongside a P and L. It's almost like a piece of context to kind of give to it, right? And that level of a 360, which is just not just the numbers, but the backtrack of those of what creates that narrative will be where I think you're going to see a marriage of those two in a way that you weren't able to do 20 years ago with just human resources. Yeah, and I'd say some of our best CFOs are the best people developers in the portfolio because it's like such a clear career track, you know, kind of start like bookkeeper, assistant controller, controller VP of finance, head of FPA, CFO. the best CFOs or even more. able to give their teams more work. Like, hey, try this out. Hey, now you do the clothes today. These tools are just going to make it people be able to move up the stack faster and do less of the minutia and more exception management, rather than put every entry into the GL. Look at every single expense. They're just going to kind of be more again at that orchestration layer rather than the actually typing stuff into a spreadsheet or having to cut and paste things from one thing to the next. I think I do think it's going to make the finance job a lot more fun for people. And back to Caroline's auditing question, like audit the spreadsheet before you send it to the associate because at midnight, she's going to find something. She's going to hit she up at 1230 to make sure it's fixed before 8 o'clock. And it's on the M.D's desk. So there's a lot of audit your work before you send it off as well. Can we jump into demo, like show some stuff to run? And for the audio people, we'll talk through it, but you can also go on YouTube and watch the demo. So I'll be sharing a few of the examples which can do with index. This one of our product lines lives as a plugin in Excel. And we'll be releasing a few more products over the next coming weeks to kind of expand how you interact with our tools. So what it does is you can kind of ask queries about the Excel file, and then we'll be able to build it directly in your firm's formatting. And remember the nuances of your firm, for instance, how you might handle a specific eBudo adjustment, how you guys get to tie to what assumptions are reasonable given the industry or the presence that you've seen, and how you want data to be pulled, right? Like whether you want them to be pulled from public comps or for an internal CRM or from either portfolio companies that you have. And so it'll be able to build those malls and have them formatted professionally in a way that matches an existing standard. So just the way you give a cloud a template for, hey, this is how we do presentations. So you do a-- when you create a PowerPoint, it automatically comes out in the way you want it to look. Right. And with the ability for you to be able to know that you have the best model for the task, right? So for a lot of these PowerPoint tasks, cloud isn't the best anymore, right? And so it's like this ability to have a firm to know which one you're using before which task and almost have like a trade off of like intelligence, speed, and cost to get that done. And so we've made it very optimized to kind of fill out templates, right? So whether that be your firm's LBO template or how you want to roll forward after another quarter and do that in a way that you might expect to be at the parity of a human. And so to leave, sell, notice, and sell comments with a human's name attached and not the agent. So it can be completely plug and play as one of those pieces. It's particularly good at kind of cohort attentionalities, right? So both being able to normalize and catenate and also build the helper columns and create the charts that you need. The biggest nuance in all of these use cases, right, is the last mile. For some firms, how do you handle null values, right? Do you either skip them or do you include them, right? And next we'll ask you these clarifying questions, but also remember that. So there's like this onboarding process that we like to do, which is to figure out the exact nuances of how you do these types of tasks and remember that, right? And I think fundamentally, as a reason why these tools have been ball capped or ball neck in terms of their capability and finances, the ability of understanding when you onboard associate at Parker Dill, that kind of go forward of what they do might be very different than another shop, right? And that ability of how you almost onboard a team member, we think will drive far more economic value than just giving someone an accelerant on their work. Yeah, well, it's like given, you know, an LLM, a very vague prompt, not getting the answer and saying this thing stupid and stopping doing it. If you don't teach it how you do a cohort analysis, how's it going to know how you guys do a cohort analysis versus somebody a generic cohort analysis? Definitely. And one thing that we've optimized it on is to handle some of those more bait queries, right? 'Cause when you think about the use, the ability of these tools, I think there's a huge learning curve, right? Of like people have been so used to a Google-esque interface, where you type in maybe like six words of intent and something comes out. Now you're moving to a world where you're almost like onboarding a task, right? And the ability to do that often is bottlenecked by your ability to think about what to do when you start it. And so our thought is like how do you make it as simple to do as possible, almost like a peer-doing experience. And you'd expect it over time for them if they're a great associate to stop asking the same questions, but maybe ask some newer ones, right? And that's how we kind of think about onboarding these tools. Yeah, and that's the challenge with just a generic LLM is that it's like a great intern that never learns anything from one week to the next. They don't never improves, which is, you know, again, part of the frustration, and I think these tools are being wildly underutilized across our industry because a lot of people tested stuff last fall didn't really work. Everybody kind of gave up on it. Now they're coming back to it a little bit, and it's like, oh, kind of outsource it to the youngest person on the team. They don't have any power to convince anybody to change their mind. They've got inertia, they don't want to do it. So it does take like big change management across an entity to say like, no, we're going to do this. We're going to suffer through the pain of training it and doing it our way. And then we're going to get huge benefits on the other side of it. Definitely. And the sticker shock that enterprises are going after is going to be pretty intense over the next eight months. I think that one of the things that we've seen across enterprises is there's been kind of this token maxing leaderboard, right? Which is like how do you measure which employees are using AI the most, right? And what we found is that these types of goal posts have been almost like, externalities that are often negative, right? And this ability of like, how do you balance how do you use these tools responsibility without having these negative consequences is really key. And there's like a story that's told around this idea of like if you had a pit of cobras and a village, like how do you get rid of them? And the way that this talent tried to get rid of cobras was put together a ransom that if anyone kind of brought back a dead cobra, you get paid. And then eight months later, the talent looked around and the people who ran it, right? The sheriff was like, why do we have more cobras? And turns out the externality was people sort of breeding cobras to return them, right? And so you end up getting to these scenarios where if you try to chase this goal of how do you try to get this specific outcome of AI utilization and just go after things like token usage, you miss the actual implementation utility of it. And what we found is this training and actual onboarding experience is like really important for people to get to a level of kind of confidence of actually using these tools. - Well, for anybody who went to business school, there's a famous piece by Stephen Kerr, which is everybody is read, which is called on the folly of rewarding A while hoping for B, which is, hey, if we token max, we're gonna make so much progress. And it's like, no, actually, constraints drive creativity. So as people come up against constrained budgets for tokens and other things, they'll have to get more creative. They'll have to allocate resources to the biggest return ROI rather than just kind of spend forever. - Yeah, and I think tech companies are gonna be the first of this harbinger, right? Where companies like ours, we spend more on internal token usage than we do on an employee head count, right? And software engineering is the first of that, but that's going to happen for all of the portfolio companies. It's easier for tech companies to kind of think about this kind of ROI on tech spend and tools as always a strictly positive thing. The structure of these budgets for technology and tools has fundamentally not been set up in financial firms for the way it needs to be. It's really soak up the ROI of this. And we've started to see some of these large banks that we work with kind of work directly on how do you train these analysts and associates to use these tools as natively part of the spend. And we'll start to see these kind of building structures that I think will exist where you're gonna start to put like token spend per deal, right? And we're really, really early in this kind of adoption curve of agents and then out of tokens that they are going to use relative to any sort of technology wave in the past. - Yeah, we agree with you on that. So let's finish it off with just maybe a couple predictions here. So Caroline, you first, let's go with kind of a private equity prediction, you bet your bonus on like 12 months from now. Like where do you think how is your life different or how your day to day different or the difference of an associate who's working on a deal team for you 12 months from now? - Well, this might be a recency bias, but I would bet in a year associates and analyst at PE funds are doing a lot less manual logging in the CRM. I would bet that most logging in the CRM is gonna be automated by agents. And I think the driver of that is that we demoed five CRM platforms in the last few months. All five of them have the ability to automate. logging from your email to the serum, which is something that I don't believe existed a year or two ago. So how are you modeling differently a year from now? Like how are our associates? We onboard an associate a year from now. How are you onboarding them differently than you would have before with these tools? Firstly, I would say, index and tools like that are key for being able to onboard an associate and say every time you send me a file, make sure you audit it in index and check that there's no mechanical or link breaks, logic breaks, which I think will also make the associate better. And then there are absolutely things we do all the time, such as rolling forward a model when a banker you're working with, lender, even an auditor asks for the model, you know, updated model six months after you do the deal or the three plus nine budget with actuals and budget. Those types of things, I think you can use tools like index four. So it'll definitely be a part of onboarding to be like you can one, have it be the doer a little bit more and you can have it be the reviewer before you send it to the next person on the chain. And then there's other things like updating for our LP reporting, all quarterly reporting for the portfolio as we mentioned a few times. Current state or current state a year ago is you pull in the reporting from the portfolio company, oftentimes it's reformatted from last quarter for whatever reason. They've pulled all the new tabs, take out all old tabs and update the links and now that can be way more automated with tools like this. And all of that work is not high level work that's making me associates think more cleverly about a deal. So things will change. I just think it's less. It'll probably happen in more of a range depending on how willing or not willing and private equity firm is to adopt the tools. Yeah, I'd also say like the people are saying, oh, we're going to get rid of our associates and our analysts because we don't need them anymore. The tools, what a horrible advertisement for your firm that you value these people so little that you don't think they can be better with tools rather than just replace by tools. But that's my own editorial comment. So Trude, maybe this is a little bit of a glimpse into the product roadmap, but what are you betting your equity on? Literally, you're from now. Where is AI in the investment banking private equity workflow? I think there's a few. One I think that there was this article written about Kirkland, Alice spending 500 million to building their own internal tooling, whether it be that firms actually build their own internal tooling, but have an economic imperative, how do they make their own firm more intelligent than the base models that are dropping open AI? So how do we justify how do we think we're better? That's one. Second, I think that most questions that follow up in a diligence process will be originated from Asians. I think that they will be proactively surfaced and that those might be proactively answered by another Asian on the other end. I think the third is that you're going to see a lot of enterprises cancel their clawed subscription or a chassis subscription just out of fear of how much they spent at the end of the year, just like knee jerk reaction of like, okay, I have no idea how to do this. Or just not everybody gets it. You've got to come and ask, you know, you've got to get permission from the CFO. It's not a funny try. Or you have to get a very strict budget. And the fifth is I think one of the top three or four consulting firms is going to introduce or try to standardize a protocol that isn't an excellent PowerPoint is the way to send a deck or some sort of resource. And almost institutionalize, whether it be an HTML file or something like that, as like the core deliverable mechanism. Yeah, pretty cool. So what about specifically for index? What's on the, as you sit down with people and talk about the product roadmap, what comes next? Again, as an investor, as I look at your product and we use it and we really like it, I'd say, is this a product or a company, right? So I know the product is great. So talk about the company you're building rather than the product you have. Yeah, I think one of the very interesting things about AI companies and the AI native companies today is they're growing a much faster rates than any SaaS companies beforehand. If you look at the path of NZRD and million of revenue for like a Figma or a notion or a Canva or some of these native CRM's, right? Those ramps of revenue of like a triple, triple, double or whatever are upended, right? And I think we are on like an incredibly faster jackaroo of how these things grow. How we see the company growing is that it don't have to move toward more of a teammate, right? We're kind of thinking about how to go from becoming a tool that people collaborate with to things that people delegate with and capturing that economic value accordingly, right? And so I think a lot more ways that people are going to interact with end X are going to be mostly triggered either without a human in the loop or due to like some sort of follow up or over an email or over teams relative to the message that you might prompt directly in these tools. Yeah, great. Well, one, thanks for building the product we love and use. So thank you for that too. As you said, a lot of interesting things today. I think the audience will enjoy the conversation because everybody's in a thick of it. One of the most interesting things you just said, which is the agent will decide what the questions are, the diligence questions are, and another agent will provide the answer, which is like, oh, yeah. Of course that's going to happen because it's a question we would come up with that can be answered. And if the information is just sitting on the other side, why would I wait 72 hours for the banker to ask the management team, the management team to give it to the banker, the banker to give it to me, the agent should just work that out between themselves, certainly for the obvious ones and they may surface non-obvious ones given the context you may provide it. So super cool to run great. Having you on the pod. We'll see you soon. I'm sure we'll be talking to your team soon. So appreciate the effort and we'll schedule this for six months for now and see where the industry is out again. Incredible. Thank you for the kind words and excited to see how you guys adopt these tools across the portfolio. Bye for now. From the heart of Chicago to all over the globe, a couple private equity geniuses, they share what they know. They love to mess with technology where the future is single, swim. It's a private equity fund cast with Devon and Jim. Technology issues, a middle market, PE back companies, they pick up with each other and they don't take themselves too seriously. It's not venture capital. It's private equity. It's the private equity fund cast. Pour yourself a drink and have a seat.