AI, Decision Culture, and the Future of Enterprise Finance with Matija Nakic

Welcome to the Future Finance Show, where we talk about treasury management on Mars. One of the things you mentioned with AI recently in a post, kind of elaborate on that. What do you mean by the right questions and why is it so important?

It can give you a wrong answer with high confidence, but it can also give you a great answer to the wrong question, which is worse than useless. It's just plain dangerous. We have customers that use Farseer AI in an amazingly clever way to really get true business benefits.

Today's guest is Matija Nakic, CEO and co-founder of Farseer, an enterprise-grade financial planning and performance management platform built for organizations that need to plan, forecast, and collaborate with greater speed and precision.

So we've built our own AI modeler internally, and our productivity in the professional services team is way above 10x. How this works now instead of, you know, having a big project team, you basically have 1 or 2 people that are architects that can kind of navigate the modeler. It's governed, it's audited. You can see every change, whether it's done by a human or an AI. And this really gives people confidence. We see Farseer as the single source of truth, regardless if the UI is Farseer AI or Claude or ChatGPT. We really see before the end of the year having at least 1 or 2 enterprise customers customers that will, with oversight, build their whole solution end-to-end. And this is something that we're very, very excited about.

Future Finance is brought to you by qflow.ai, the strategic finance platform solving the toughest part of planning and analysis: B2B revenue. Align sales, marketing, and finance Seamlessly speed up decision-making and lock in accountability with qflow.ai.

Welcome to Future Finance. I am Glenn Hopper along with my co-host Paul Barnhurst. And Paul, I'm pretty excited today. We actually have a guest, so people aren't just going to listen to us ramble on for an hour. So we'll have someone who actually knows what they're talking about.

But doesn't everybody like to listen to us ramble?

We like to think so anyway. Tell us about our guest. What are we doing today?

Well, we're really excited to have a guest with us today. We have Matija Nakic. Welcome to the show.

Thanks guys. Thanks for having me.

Really excited to have you. So I'll give a little backstory and I'll go through a bio. We, we met each other almost 5 years ago now. It would have been before I had even started my business. I think it was late '21 when you were first looking for funding and you had just started.

Yeah.

And so one of the first FP&A tools I saw, and I've followed her journey, it's been exciting to see. So really excited to have you on the podcast. I know we've been able to chat several times through the years. And so a little bit about her background and then we'll, uh, we'll jump into things. So after she graduated in computer engineering, completed an MBA program to combine her technical and business knowledge. Throughout her career, she's worked at several different companies. Including Span, Five, HT, Aeronet, and others. From the very beginning of her career, she was big in B2B software for large enterprises. That's where she found her passion was, and she's managed several different teams to deliver enterprise solutions and products. She's gone through all the different phases. She's been the developer, the product director, and now she gets to be CEO and co-founder of So Farseer is a fast-growing AI-native SaaS for business modeling, planning, and analysis platform. And today she's working to lead the company toward its stated goal of disrupting the planning software field and providing financial professionals a true solution to their everyday business problems. Farseer's long-term vision is to empower business users to build any type of enterprise software on Farseer. Which we'll get into a little more later. But again, welcome to the show.

Thanks, Paul. It sounds better when my marketing team writes it, but, but it's all true. I have this feeling that everything led me to Farseer, you know, which is a bit of a romantic way to look at your career, but—

Isn't it always fun to read something when marketing writes it versus how you would write it?

And you're like, yeah, I like this girl. Sounds cool. Yeah.

It definitely, like, wait, is that really me? That's what I usually think when I see something that marketing wrote.

And then there's the error when people use clichés, like the LinkedIn, this stuff you see all over LinkedIn. A good friend of mine was just promoted and he sent me the promotion announcement and I was giving him a hard time about how that was worded. One of the things it said was delivered consistent results across a complex and dynamic operating environment. And I was making fun of him because I thought some copywriter had done it. But he's like, I had to write that myself. So maybe it, I think it is better to let marketing handle that, especially when self-promotion can be a tough thing to do.

That's funny. All right, well, great. So I want to start with this question, just kind of level set for our audience. Obviously, Farseer is a planning software, primarily financial, but does other stuff. And why did you start You know, a tool in this space, very fragmented. There's lots of options out there. So why did you think the market needed another tool?

We didn't think about the market at the time. We thought about our own issues though, truly, because Zrinko, who's one of the co-founders, and I worked together in a telco environment and we're both engineers and we were flabbergasted by the way that functions or doesn't function. So it started as a hobby almost. And Matej and Luka joined as the other 2 co-founding members, great software engineers, great people. And then we started thinking about what this could be, you know, and researching, do others' company, do other companies have the same problem? Is there any existing solutions that, that's when we got to the market, but not before writing a lot of code. You know, so that was, that was, that was kind of the first thing that, that's always the first thing that engineers do. And basically when I joined the team, it was like, okay guys, let's see, you know, where this is all going, what we could be, whose problems can we solve? Let's start with that. And curiously enough, we ended up immediately in the enterprise segment, which is not a typical startup path. Usually startups sell to other startups and it kind of grows from there. And we ran into these very data-heavy, very complex planning environments from day one. And that really shapes the company and the product.

I agree with you. That's not the usual path, like you mentioned, you kind of start small.

Definitely not. But as we understood the market better and understood the problems better, We immediately knew that this is an engineering problem. Sure, you know, there are nice ways to build a dashboard, probably nicer than it is in Excel today, but in the end of the day, in an enterprise with so many people doing planning, reporting, simulations, with so many data sources, with so many outside data sources that people are starting to add to, to Farseer lately to, to improve their forecasting. Uh, it's just a performance problem. So we started building our own database and calculation engine, if not from the first week, then from the second week. So 5+ years now into the journey, and we're actually now kind of re— uh, getting the dividends from, from that early decision.

I remember seeing that the first time you had rebuilt it. I think we talked like it was a Friday or something, right? The engineers are celebrating. We've made this much of an improvement in the database as one of your first early rebuilds. So I know you've been working on that for a long time.

True. Still are. Still are.

I'm sure. It's a never-ending thing.

Never-ending story. Yeah.

So I am super interested in the idea of letting business users build any type of enterprise software on Farseer because no one would ever pay me to sit in front of a blinking cursor and write code. But I've been, for my whole career, I've been a citizen developer. And whenever IT would tell me no, Or I would be in on the dev side. If I was stuck in a backlog, I would build it myself and run, uh, you know, run it on my own laptop or whatever to try to get around it. And I'm thinking now with generative AI, so many people vibe coding out there and vibe coding will, you know, you can build anything you want without knowing much about it, but man, the dangers of it, of not understanding the schema, having heard of a semantic layer. Not knowing what OAuth is or role-based access or where, or even making your GitHub repository public instead of private. People, things people would do if you're just, if you don't have a programming background. But I think about all the limitations of the software that's out there and all the workarounds that people have had to have around that. And I'm wondering, and maybe this, maybe there's 2 sides to the question, one internally maybe, but with the rise of generative AI and vibe coding, the way that you can build finance software, or on the flip side of it, how the users might be trying to build software? And is there anything you're doing differently today, just sort of in that vibe coding world? And I'm thinking with what your vision is here, that this is probably a really good moat differentiator for you guys, that if people are building in your platform rather than just out on GitHub and hosting something on Vercel or whatever they're doing with their vibe coding.

I love the intro. For 15+ years, we've been promised that, you know, people can build their own enterprise software. In the end, all of those platforms, they have a very high barrier of entry, and there are some people like you, but I would say there are not enough of them. And, you know, for decades now, you have this crazy situation where you have a group of developers, a group of business people, and a lot of money gets burned, you know, between these two not communicating properly and not understanding each other properly, regardless of if it's in-house people building something for the business or outside vendor that comes in and tries to replicate the processes and the models of the business side. So we noticed a couple years back that people started hacking Farseer and building things that were not FP&A, really like, like, uh, car fleet management solutions, you know, database of all the marketing projects, all the investments, ROI, things like that. And, uh, you know, ESG planning, ESG reporting, different things. Because as engineers, we haven't built a platform which lets you log in and immediately you have your kind of building blocks of how to build to P&L. We built a platform that has these distilled building blocks, actually, you know, so to some people that could understand it and see it, this meant, oh, wow. So this is, you know, a robust tool that I can use to basically model everything. And that's what they did. So this really informed our vision and we worked a lot on the UX to make this as more, uh, as easy as possible for, for our customers. But then you get the LLM story and this is where things start exploding. So we've built our own AI modeler internally and our, we call it Mindseer and our productivity in the professional services team is way above 10x. So we're seeing crazy things happening here to the point where we're reorganizing the way that our team works completely. So how this works now, instead of having a big project team for an enterprise with 6, 7 different roles, a project manager, you basically have 1 or 2 people that are architects that are highly experienced that can kind of navigate the modeler and build the backbone of the model and then just kind of observe and teach the customer to fish rather than handing them, putting the fish on their plate. And this has been a crazy ride. And this is just like a couple of months back that this started happening. So this is extremely exciting. And we really see before the end of the year having at least 1 or 2 customers, enterprise customers that will, with oversight, build their whole solution end-to-end. And this is something that we're very, very excited about.

Wow.

And I imagine you're also seeing your speed to deploy gone through the roof as well, right?

Crazy.

Wild times. And I think about, I argue with Claude every day about the way it writes copy, but I don't have that issue with code. And really, these are large language models, so you would think that the language part would be the best, but the advances in code, and I'm always doing bake-offs between Codex and Claude Code and seeing, you know, just giving it because it's so cheap now and so quick to push stuff to production. Again, these are obviously anything I push to a client will go through a QC and people with a little bit more skills than I do. So I'm not pushing, you know, unsafe AI slop out there, but it is, it's amazing just for whether it's building a little app to track your workouts or something that is actually a full-blown You know, client dashboard that has a chatbot that you can talk to your data with the semantic layer and all that. I mean, things that we're deploying today relative to how long it used to take us to roll products out, it's pretty amazing. And I know you have a much higher bar working with enterprise clients though, but you also have the architects and the skillset in there to do it. And it'd be interesting to see. It feels like You know, if you're on an assembly line factory, everything just got 10x, just flying down the, uh, down the production line.

I would love to show you this. It's, it's really wild. It's really wild because we did, we invested a lot of effort in, in this, you know, layer of, of knowledge that, that goes above the, the LLM and the results are staggering. But I wanted to add a few more things. So we are, we, we already had JavaScript. As, as, you know, part of the platform where you can just write code and this is the, the most widely used language out there and you can automate stuff. And this paired with AI is then, you know, a superpower. And we're also going to add SQL. So we're working on that as well. And our plan is opening Farseer by the end of the year for highly technical users so they can, you know, really get their hands on it and play with it because we believe in builders. And this is something that the FP&A or the APM category of tools forgot somehow. You know, IBM had a really good community and was quite a technical product, but there was no spiritual successor, I would say, of this approach until we came along. So, so this is the path moving forward.

Ever feel like your go-to-market teams and finance speak different languages? This misalignment is a breeding ground for failure. Impairing the predictive power of forecasts and delaying decisions that drive efficient growth. It's not for lack of trying, but getting all the data in one place doesn't mean you've gotten everyone on the same page. Meet Qflow.ai, the strategic finance platform purpose-built to solve the toughest part of planning and analysis: B2B revenue. QFO quickly integrates key data from your go-to-market stack and accounting platform, then handles all the data prep and normalization under the hood. It automatically assembles your go-to-market stacks, makes segmented scenario planning a breeze, and closes the planning loop. Create airtight alignment, improve decision latency, And ensure accountability across the team. Exciting. It's, it's amazing how AI has changed the way we work in so many ways. But, you know, I think everybody's trying to get the most out of AI. Everybody is struggling with figuring it out. And one of the things you mentioned with AI recently in a post, and I'd love for you just to elaborate, share your thoughts on it. You put out a post where you said one of the most important, or it's more important than ever for people to ask the right questions. Kind of elaborate on that. What do you mean by the right questions and why is it so important?

AI is tricky. It can give you a wrong answer with high confidence, but it can also give you a great answer to the wrong question, which is worse than useless. It's just plain dangerous. And it's just a multiplying force. You know, like the, I think I even use a Bible quote there, but I have the feeling that, you know, people that are smart were smart because they learned how to ask the right questions and, you know, be intellectually honest and, you know, not let go until they get something that clicks, that makes sense. And here it's true, but it's like 10x or 100x because you have this multiplier in your hands. And it's the same, like if you're not drilling down and if you don't have real-world experience, it can be very dangerous. So that's, let's say, from the perspective of a person using AI. I mean, you mentioned copy, Glenn. I'm also very annoyed by it. I don't want to get AI slop sent over to me. You have to be the editor. And being the editor is the hardest work. All the best writers are like very, very good editors of their own work. So this is, this is something that really matters. And it's the same with code. It's actually, you know, the editing is what really counts. But if we talk from an organizational level, we're seeing the same pattern. We have customers that use Farseer AI in an amazingly clever way. To really get true business benefits. I said that finance should move from precision culture to decision culture. And these are exactly the teams that have decision culture where they have, you know, Farsir in front of the board during the board meetings and asking really right questions. And that was fascinating for me to see because we participated on one of those. And then you have some teams that are really like, basic level, kind of still learning how to use it and not getting so much value out of it. And of course, we try to help them as much as possible to get to the right questions. So yeah, I guess right questions are foundational now.

Thinking about it in the board meeting, that feels like the Star Trek computer where you just like, you have this omniscient, this all-knowing computer out there and you're asking it. And I know exactly what you mean though, because it's like when you create a project and you give it the context and you put the files in, or if you have it connected to the right system, you know, in the context of that, it's going to be pretty good at answering questions, but also the underlying memory and that just whether it's Anthropic or OpenAI or any of them that are holding and are building on you, that context is growing too. So I think sometimes People forget, like you just start a new chat, they forget that they're coming in and the AI is basically like that movie Memento where it doesn't remember all that. And they'll ask a question, but maybe from its memory, it's picking up some tiny bits of context so it can give you a really wrong answer. So having an environment like Farseer where it has that, again, the semantic layer and the data and everything behind it where it knows within guardrails, okay, we're answering the question based on this is significant.

And it's governed, right? It's governed, it's audited, it's, you can see every change, whether it's done by a human or an AI, and this really gives people confidence because director of FP&A in one large enterprise company told me recently, like, the business users are driving us crazy. They got completely open access to ChatGPT and Claude and She said, now they're throwing some Excels in there and they're asking me, oh, my EBITDA is like lower than you told. She said, we have like 18 definitions of EBITDA, you know, but their Claude doesn't have that. We really strive to have the right context there. And then the questions are making much more sense.

Yeah, that reminds me of the era of when everybody wanted a self-serve data mart and the idea everybody can come just Pull the data that they want, but if they're not, if they're not following the same data dictionary that, like you said, well, sales wants to define the pipe. Yeah. Yep.

Good old data dictionary.

The dictionaries, data foundation. Haven't we been hearing this for 20 years now?

Yeah. And somehow people still, it's still not sexy. Nobody wants to do it, but it's the found. And now it's just like, if you don't have your data right and you're trying to throw AI on top of it, good luck.

Same principles still apply. You know, the laws of software. there haven't changed. It's always going to be garbage in, garbage out. AI is just amplifying the garbage out and it's giving it to everybody as opposed to one or two data analysts. So the problem is becoming just heavier, you know?

You mentioned before, I want to talk about the 3 levels of AI use. And I always, I think I see this too when I do a lot of training for teams because before we do any kind of implementation, we like to get that Well, the data foundation, but then the sort of enablement and training foundation. And I think about, so generative AI has already hit the peak of inflated expectations, gone through the sort of trough of disillusionment. I would argue that it's sort of coming up into the, not sort of, I don't know, I haven't seen the latest Gartner, but I would imagine they have it positioned in the plateau of productivity or whatever.

Productivity.

Yeah.

Yeah.

I would imagine that generative AI is generally sitting there right now, but There's the hype cycle, but there's also the adoption curve. And I, you know, early on I thought people were the adoption, I thought it mirrored the hype cycle. But really what I'm seeing is the people who rode the front end of that wave going back to 3.5, you know, ChatGPT 3.5 and through that, that was really the early adopters, the bleeding edge and all that. And now where I would think we're in laggards, I think that it's the central mass of adoption is coming through right now. So if you spend all day using these tools, you think about it as if like, that's just, everybody knows that. But when you talk to users and how they're actually using it at work, this goes with what you said about the 3 layers of AI in finance. And the first one being, you know, the copy paste layer where you're talking to the chatbot and having those interactions just kind of Ask, answer, take that, copy it, paste it somewhere else or whatever they're doing with it. And then some are automating processes, but that third layer, and I think this goes along with sort of the Farseer mission too, is creating that decision culture. And you referenced it in the, in your last answer as well. So if we could dig in a little bit on that, maybe talking about the layers and then really what that peak layer of it, of having AI as part of a decision culture What does that look like and how does finance create one?

Well, the first layer, like you said, I think majority of people in companies are still here, which is throwing Excels, building some models, doing some analysis, kind of ad hoc work. And this is fine. It brings some value, but, you know, on the company level, I'm not sure how much, you know, the CFO or the board can feel some value from this. Sometimes it even gives you more work. than you used to have, right? In certain situations. And then you have the second layer, which is a bit more advanced where there's more orchestrated approach in a company for AI, usually, you know, using somebody from the Big Four, some vendors taking a specific process and just running it more efficiently, cash collection or whatever. And I think this is very, very valuable. And this is, a small fragment of the companies, but the most advanced ones are certainly already here and they're not even bragging about it, you know? So that's one thing to know as well. They want to gain competitive advantage, of course. And then the ultimate layer in my book of AI would be really getting to decision culture where you have a single source of truth and AI on top of it. And we see Farseer as the single source of truth. Regardless if the UI is Farseer AI or Claude or ChatGPT, because we're opening towards this as well, or in Excel, and we think users should pick their environment and work where they feel comfortable working most. So we even have Teams and Slack as where our AI chatbot approaches the user or gets answers from them. So this decision culture basically requires the old school stuff, you know, clean data, data dictionary, connected data sources that are continuously updated with only the changes, business logics, which is basically context, human judgment in the end of the day, because it's impossible to just run things, you know, and have AI run things. So it's kind of human in the loop. And there are 2 levels here. One level is, like I mentioned, the minds here, the builder, the modeler that helps our customers build any kind of a model really quickly, different kinds of reports, different kinds of planning sheets and so on. But the other part that we've built is called workflows and workflows essentially enable our customers to automate standard set of steps so they can run it more efficiently and faster. And this very often includes a human in the loop. So for example, a flash report, some sort of deck that needs to be handed to the board once per week or once per month, regular forecasting cycles where something comes in and something changes. And then it needs to go through a certain set of people to do kind of a high-level forecast. But also it enables a much more frequent forecasting based on a change. We all know about cocoa prices and things like this, but these are real-world scenarios where our customers in the CPG had this drama and it took them weeks to replan. Maybe it even took 1 or 2 weeks for somebody to even react and say, hey guys, you know, I don't think our plan is valid anymore. So you can also do these ad hoc things much, much faster because the whole cycle or almost the whole cycle is automated and it just pings a human when it needs an answer or a confirmation. And we are still early on in these workflows and we're testing it with, I think, 5 or 6 different customers at this point, but We already see the great benefits that they're getting from this. So the decision layer would, in our eyes, look like that.

Yeah, it's such a fascinating time because I think if you've been working with the tools and the models for a couple of years, you can sort of see that. But if you're just now jumping on the AI train, that sounds, it still sounds like science fiction, but with the longer running agentic tasks. And also, like you said, building frameworks, whether it's an agent harness or whatever, where you have the guardrails and you have it where this piece is deterministic, AI is working here, it's repeatable, it's auditable through the whole thing. Like, that 2 years ago would've sounded like sci-fi, but we are really starting to see that out there now. And I know with the work that Farseer is doing, that's gotta be your— you see that future and are locked in on it and moving towards it.

Absolutely.

It's crazy how quick it's moving and what AI can do right now. It just keeps getting better and better. And I think, like you said, Glenn, maybe we're— we need to all watch Space Odyssey again or something. Yeah.

I don't know if we want HAL, but I don't know, I might take HAL over some of the recent Claude outputs I've gotten. So I don't— just don't turn off the life support.

Only a few nerds will understand that are listening.

I could agree with Glenn. I would take hell. Recently the copy is really bugging me because the language is not human at all. Okay. I'm not a native English speaker, but—

But come on, it's not X, it's Y.

Yeah. Yeah. Yeah. Yeah.

There, there are a lot of things that I just read it. I'm like, I can't put that out in the, that's not going into the world.

Yeah.

That's getting a rewrite before anyone sees that one.

So being the editor, that's our job now. Ruthless editor of AI slop.

A lot of truth to that because it can produce so much so fast. Also, you guys will laugh about this. We'll get back to the question. I was preparing for a different podcast interview I had to do, and Copilot had my transcript, so I just asked it to summarize and come up with the questions I should ask. And, you know, normally it will get— it's knows I do about, you know, 10 to 12 questions on this podcast. It came up with 2,500 words. It was like 8 pages long. There was literally like 100 questions and I just responded with, and how does this fit in? And I mean, I gave some snarky response and it was like, yeah, I way overdid it. And I'm like, you think?

The fact that we're talking with AI like it's a person, that's, that's also really, really fun. You think, you know, that, that, that comment is like, you think?

I know. And I love how people, like some people I know, they'll refer to Claude as a man or others will call it a woman or, you know, it's, we give it gender sometimes. It's really interesting to watch.

Yeah. I call it Uncle Claude.

Here, here we, we, here we have this Italian variant, Claudio. I don't know why, but it's, it's sticking in Farseer, maybe because we're across Italy here. So.

Yeah, you're pretty close there. All right. So I want to ask an Excel question because, you know, FP&A software and FP&A wouldn't be good without mentioning Excel. We hear from a lot of vendors. We hear a lot of this, hey, you need to get off Excel. Excel is dead. Excel's a terrible tool to use. It's error-prone. So I'd love your take. When is Excel the right tool? Because I think we all know it's not going away, and nor do I think it should. There's always a place for a spreadsheet. But what's your take? When should you be using Excel? When should you be looking at a purpose-built tool? Maybe just share some of your thoughts there.

I think this is sheer laziness, you know, to say Excel bad, software good is bad marketing. And it's not telling the truth. I mean, you know this, and we've talked about this a lot. We always saw Excel as a source of inspiration because it is inspiring. And it still to this day, I think it's the best tool we as, you know, tech people have given to humanity. But of course it has its limitations. So it really, the answer is it depends. I mean, Excel is great. Excel is great if you are doing rapid modeling, you know, a new idea, some commission structure, new pricing model, you want to test it fast. It's unbeatable for speed in modeling. In isolated analysis, one person checking things, great tool. When in general, when there's low collaboration burden. I would say, you know, it's really great. And we know that there are problems when multiple teams are touching the same numbers, versioning chaos, hidden formulas, things that are a great audit risk. So also, you know, when you need real data inside and you need it refreshed quickly, that also becomes a problem. There's a lot of problems with like large amounts of data. And then you have all these crazy techniques that people use called Excel cutting and so on. Distributed ownership. So, you know, we always say a lot of people planning, a lot of spreadsheets, probably there's a pain point. And also for industries that are very sensitive in terms of planning, like we see this a lot in pharma and CPG companies that have their own products. If you plan poorly, you Really lose cash. So you either don't sell as much as you could because your demand forecasting is off, or you have way too much stock and you're burning money, basically. Yeah, I would say Excel is great if it's working for you as your, you know, one-off, one-person Excel jockey, please continue. And in general, you know, there are just companies that don't have such a complex planning environment. And we had inbound leads coming to us where our sales team would say like, no, please keep doing it in Google Sheets. This is great. Like you really don't need us. This is structured, nice, looks great. There's like 5 of you. It works perfectly. But, you know, in the enterprise, it's pretty chaotic and it's very dangerous. But I would never say something like Excel has these billion-dollar errors. You can make billion-dollar errors in the planning Solution too, you know. So, so I think that also needs to be said.

Yeah, I'm trying to say is the errors are a human error. Now, could you say the software could be designed differently to limit them? Sure. Like having a database on the back end and some things you define might help the chance of that error, but it's still a human error. It's not the tool's fault.

Yeah, still can be an error.

I'm with you. I, I hate when people Use that as the reason you should buy a tool. Well, Excel makes a lot of mistakes.

It's lazy marketing. It's lazy marketing, basically. But in the AI context, it's kind of playing in our benefit, especially in the enterprise segment, precisely because of the things that we've mentioned, you know, the data dictionary, the context, the stamps, who changed what. And where, you know, was it human? Was it AI? Was it some demand forecasting model that did something? You really need this audit trail to understand what's happening to your numbers and you need context as well, right? It's not just stamping things. It's also context from the people that have to explain why they did something. And then if you have all that in one place, it becomes much easier to make decisions.

Agreed. All right, so we're gonna move into our fun little section we do at the end. We're right toward the top of our time, so we'll, we'll go through this pretty quick. But how it works is we give your bio, the questions we came up with, access to the internet, and we ask AI to come up with 25 kind of quirky, fun, personal questions about you. And we use different models. This one was Fable 5.1. Okay. And then You get to pick. So there's 2 options. I take one approach for one question. Glenn takes another. You can keep a human in the loop and pick a number between 1 and 25, and I'll ask that question. Or I can use the random number generator to pick a number.

7.

All right. She didn't even hesitate. So it says, you keep your hobbies as a reset with no deadlines and no KPIs. So what is the hobby And how bad are you at it? I don't know how I came up with that question, but that's today's question.

Okay. I go to the gym just to stay healthy and fit. I wouldn't even call it a hobby. I think it's just kind of a support system to everything else that I have to do in a day. But I never ever, so I have this app, but I never ever put in the kilograms or anything like that because I think I don't want to see data ever, you know, if I'm outside of the And I, and I, and I work by feeling and it's, I'm also crazy like that in running. So, you know how people have these fancy Garmin watches and this and that. And I was like, I go out without a phone. I don't have a smartwatch and I run until my teeth or my gums start tingling. And then I know that I had like, that I'm drained, that I'm like just really tired and I can go home. So I guess, uh, well, I've never done that one.

I love that.

Uh, okay. So weird, but I, so I am obsessed with Strava and tracking everything, but the talking about your gums tingling, this has started in recent years for me. If I do a long bike ride or it's really in the heat, more long bike ride or run, my lips go numb. And I asked my doctor about it. He's like, you sound crazy. I've never heard that in my life.

No, it's like, watch it.

Now I'm hearing it from you.

So it's like, okay, that actually is a thing. Okay. I just use that. No, I have no smartwatch and I don't bring a phone.

So.

All right.

Well, that's, I'm glad I found a kindred spirit with the same mouth going numb thing when after a workout. So it makes me feel less—

You go running together till you're numb? You're telling me you're numb?

Go running till your gums are numb. Try saying that 3 times.

No, I butchered it when I said it.

All right. So I take the human out of the loop in my selection and I figure AI wrote the question. So then I'll ask a different AI and today it was Gemini. I've fed it the questions and your bio and I say, which of these is the most interesting? So on mine, it came back with, you, oh, this is a pretty good question. You once wrote that back in your telecom days, you had strong contempt for financial planning. What was the moment that turned that contempt into a company?

That's a good question. I think the contempt turned it into a company. There was no box in between. It was like, this is horrible. How is it possible that people still live in this 2-dimensional array? And it was crazy because we would get to March and we still didn't have a plan for that running year. So it was complete psychosis, really. And, you know, as part of this managerial team, I was the captive of this Excel hell that we were living in. And it was super inefficient. You know, nobody knew anymore after 6, 7 cycles who said what, where's which version, if the marketing and sales actually synchronized. Most of the times they didn't. So it would really put company into jeopardy because we didn't really coordinate things properly. So yeah, I guess that's the answer. There wasn't much in between.

You kept company, CVC.

Yeah. Yeah. The, the, the thing in between was, was the, you know, the, the question, there must be a way that this can be better, you know, and let's try to build that.

Great. Well, Maria, thank you so much for joining us. It's always a pleasure to chat with you. Uh, been a huge fan of Farseer. I love the area you're in, being there in Croatia, beautiful country, and keep building. Thank you for carving out 45 minutes for us. We really appreciated the chat. So thank you.

Thanks, Glenn. Thanks, Paul. It's been a pleasure as always.

podcast platform of choice. And may your robot overlords be with you.