Can a finance team trust an AI answer if the systems underneath it disagree about what revenue, cost or performance actually mean?
In this episode of AI at Work, I speak with Matej Trbara, co-founder of Farseer, about the foundations required to use artificial intelligence in financial planning and analysis. Finance teams are being asked to forecast faster, model economic uncertainty and provide answers while decisions are being discussed. At the same time, many of those teams continue to reconcile data across ERPs, BI products, warehouses, planning systems and large collections of spreadsheets.

Matej begins by defending Excel. He describes it as a programming platform for people who are not programmers, which helps explain why it remains so deeply embedded in finance. A user can start with a blank sheet, add numbers and formulas, and build a model that reflects the way a business operates. The difficulty begins when that flexibility is stretched across an enterprise. Hundreds of contributors may change structures, formulas and definitions before a finance team attempts to combine everything into one report.
This leads to the central argument of the episode. According to Matej, finance often has a data problem before it has an AI problem. Definitions can drift between systems, actuals require reconciliation and different teams may work from different versions of the financial model. Adding AI to that environment can magnify existing inconsistencies rather than correct them.
Farseer’s approach separates the natural-language interface from the calculation engine. Matej says AI should translate a request, such as asking for a sales variance analysis, into machine instructions. The company’s Rama database then performs deterministic calculations against a governed financial model. AI reasons over those results, while the user can trace an answer through each level and return to the original system of record.
We also discuss how this model changes forecasting and scenario planning. Annual or quarterly forecasts can become outdated quickly when tariffs, energy prices, wars and market conditions affect a business. Matej describes a move toward monthly and continuous planning, where new actuals update the model, responsible people receive a notification and a human either approves the result or adjusts the plan. Farseer estimates that AI-supported workflows could reduce the cost of producing a forecast by between ten and one hundred times. That is a company estimate, but it gives finance leaders a concrete claim to test against their own process.
Another practical example comes from board reporting. Matej says some clients now use Farseer AI Analyst during meetings to answer questions that were not anticipated in the prepared deck. Instead of promising a response several days later, finance can question the live model, inspect suspicious results and continue the decision while everyone is in the room.
The opportunity is faster planning and greater access to financial analysis. The challenge is resisting the temptation to let a language model become an ungoverned calculator. Could your finance team explain where every AI-assisted number came from, and would that answer survive the scrutiny of the board or auditor? Share your thoughts with me at Tech Talks Network.
Useful Links
Connect with Matej Trbara
Learn more about Farseer

[00:00:05] What happens when finance teams put AI on top of data that they cannot fully trust? Well, the analysis might arrive faster, but so can the mistakes. So today I'm joined by the co-founder of Farseer. And together we're going to examine how finance leaders can use AI for planning, forecasting and scenario analysis
[00:00:31] without handing calculations to a model that might improvise. He will explain today why Excel still remains an incredibly powerful tool, where it begins to fail at enterprise scale, and why finance often has a data problem before it has an AI problem. And I also want to learn more about Farseer's approach,
[00:00:56] where governed financial model and deterministic calculation engine do the math, while AI acts as the natural language interface. So for anyone who has spent month-end chasing spreadsheets, that conversation might feel a little too familiar. But as you know, I'm a solutions, not problems kind of guy. So let me introduce you to my guest now. So thank you for joining me on the show today.
[00:01:23] Can you tell everyone listening a little about who you are and what you do? First of all, thank you for having me, Neil. So my name is Matej Trbara. I'm co-founder of Farseer. So Farseer is an FP&A platform, FP&A financial planning and analysis. What we like to call it operating system for finance teams. You know, so what does that mean? So finance today, when they work across their process,
[00:01:50] they need to deliver to board management and, you know, budgeting, forecasting, they have to interact with a lot of different systems, you know, ERPs, BI, data warehouses, planning tools, and, you know, obligatory tens or hundreds of Excels. And what ends up happening is that finance people, you know, spend enormous amount of time on manual reconciliatory work,
[00:02:17] you know, which is error prone and, you know, not that fun to do. Finance should be there driving business decisions. So Farseer actually gives finance people one place to interact with the company data and help them significantly in plan analysis reporting. So the objective is, you know, to significantly reduce manual plumbing work and give finance people opportunity to, you know,
[00:02:47] help drive the business decisions rather than doing manual work. And the work that you're doing here is so important now, I would argue even more important than ever. But I'd love to take you back to your origin story there. Was there a particular moment when you realized that spreadsheet-based planning had become a much bigger problem for finance teams than most people admitted? What was it that lit that spark and put you on this path? Yeah, I mean, I would like to take credit that I figured this out.
[00:03:16] But as you know, there's four founders in Farseer. And actually, Matija and Zrinko, by the way, none of us have, you know, a financial background. We're all engineers. But Matija and Zrinko, they were working in a big enterprise telco company in marketing and sales. And obviously, they were involved in, you know, budgeting, forecasting. And to them, that feel like very archaic way of doing things.
[00:03:42] You know, a lot of Excel files, you know, usually before you even finish the budget, you know, actuals would catch up with you. So it seems like a lot of kind of work was futile there. So because we were friends with them, they reached to us and said, like, look, there must be a better way to do it. And by the way, they had an FP&A tool already in the organization, but it served as a kind of glorified data bucket. When everything is said and done and consolidated in Excel, they would just upload the numbers there.
[00:04:11] And some five years ago, they came to visit us in Zagreb and said, like, look, guys, this is a real problem. And it's worth thinking about. And then we spent some time, you know, just talking about it, talking to other, you know, people in enterprise companies, you know, who we knew. Turned out it was a big problem. So that set us on our path, you know, on trying to solve it. A great story. And of course, Excel is something that remains deeply embedded in finance.
[00:04:39] But why has it been so difficult to replace? And what are those warning signs that has shifted from a useful tool to almost a business liability now, especially when everyone's got somebody in a team that's using a spreadsheet that nobody knows about? Yeah. So let me just start by defending Excel first, because no, honestly, I mean, for me, I think Excel is one of the best softwares humanity has ever produced. And I really mean that.
[00:05:07] But why is it good is because it acts as a programming platform for non-programmers. So let me elaborate a little bit on that one. So imagine you have to build a financial model that has a lot of logic in you, that has to reflect how your enterprise does the business. One way would be, you know, getting engineers and consultants and building a custom-made solution. But what Excel provided here to finance people is to actually tell them, like, here's a blank slate.
[00:05:37] Just do it yourselves. And here are just like few rules that you have. These are cells. You can enter numbers. You can create a formula. And with very, very simple, you know, concepts, you can build very, very complex workflows. And that flexibility is perfect for financial modeling. The problem with Excel becomes when you have to scale that on enterprise level. And, you know, when you have tens or even hundreds of people who need to provide inputs for your next budgeting season or forecasting season,
[00:06:05] there usually is a full team within the finance who just orchestrate that process, who need to kind of make sure that, you know, they prepare the Excel files. They fill them with actuals. They put in the right models, right formulas. They send them out. They notify the people. Come on, give me. You have one week to provide the inputs. But then someone changes the structure of Excel. Then they get back hundreds of Excel. They need to compress it all together in one report.
[00:06:33] And that takes a tremendous amount of time. And this is where kind of the Excel breaks down. It's still, you can do it. And a lot of companies, actually, 70 plus percent of enterprise companies still use Excel as a primary way to do budgeting and reporting and analysis. But, you know, it's not the optimal way. And we're approaching, I think, 41 years since Excel first appeared in our lives back in 1985. And fast forward to present day.
[00:07:03] And, of course, AI now promises faster analysis. But it can also produce faster mistakes. In fact, it does it at machine speed. So, what foundations need to be in place before a finance team can trust AI with things like planning and forecasting in an enterprise? Yeah, I mean, to be honest here, the adoption of AI in finance hasn't followed some other departments that you have in engineering, marketing, sales.
[00:07:30] Like, they use AI much, much more than, let's say, finance. But it would be unfair to say that finance has an AI problem. What I think that the problem runs a little bit deeper and finance has actually a data problem. I mentioned earlier, you know, if you want to do a forecasting or budgeting process, you need to interact with a lot of tools before you can even start. So, your ERPs, data warehouses, BI, CRM, HR systems, and, of course, you know, tens or hundreds of spreadsheets.
[00:08:00] And each of that data source, they start with the same, let's say, set of definitions. Like, what's the revenue in our company? Like, there's a specific definition of what revenue is. But over time, those definitions drift. They get changed on the company level, but they might not get updated in every system. And that reconciliatory work starts to create friction. And if you put AI on top of that, it will just amplify that friction instead of kind of solving it. So, AI will not solve your data problem.
[00:08:29] You first have to solve your data problem and then apply AI to it. And this is actually what we do in Farseer. So, you need a way to automatically reconcile data from all those data sources. You need to ensure there's one set of trusted definitions that you use in your business. If it changes, it must change on one place. And you have to know where that place is. You need to have one governed financial model.
[00:08:56] So, you know, everyone looks at the same, not just definitions and data, but the whole model, how the data flows there. And you need a deterministic way to do calculations on your data. Once you have that, and if you put AI now on top of that system, this is where you can unlock a lot of benefits. You know, not just in pure analysis, just analyzing data, but also being proactive and you use AI to control those workflows that you need to handle manually now.
[00:09:27] And I think that's a perfect moment to introduce everyone listening to Farseer's AI analyst that allows people to question live financial models in plain language, while also preserving governance, traceability and accountability, which will be music to the ears of many people in the industry listening there. But tell me a little bit more about that and how it does this. Yeah, so AI analyst is just one part of kind of our AI stack that we use.
[00:09:55] And an important thing here to note is that all our AI offering is AI in Farseer is actually an interface. It's not the calculator. So if you allow AI to do the calculations, you risk hallucinated results and you risk, you know, interpreting different data sets in a different manner. So what we actually do, so underneath the Farseer user interface, you have what we call Farseer Rama database.
[00:10:22] That's our proprietary in-memory calculation engine that we build in-house. It's a real-time, really fast, able to process billions of cells in real time. And it sits on top of that one governed foundational finance model. Now, AI, when you ask it some question, like, can you do a variance analysis in sales or wherever? It actually doesn't, you know, go inside the database, fetch the raw data and do the calculation.
[00:10:52] Rama is actually the one that does the deterministic calculations. And then AI does reasoning on top of that, those deterministically calculated results. This is where you can, this kind of the line where we ensure that trust. And you can then drill down to every result that AI gives you, like, okay, how did you calculate this? Go level deeper, go level deeper.
[00:11:14] And you can go down all the way to the system of record because every that inquiry, it's fetched from a deterministically calculated engine. And AI just interprets your kind of natural language request into machine instructions, which are then executed by our in-memory database. Incredibly cool. If we look, I mean, we're talking in the UK or in Croatia, and we'll have people listening in the US and everyone around the world.
[00:11:42] Economic conditions are changing so quickly right now. So how should businesses maybe rethink forecasting and scenario planning? And just to bring to life what we're talking about here, can you share an example of what better decision making looks like in practice? Yeah, I mean, as you said, like, just looking at, like, the news headline, the wars, the AI, the oil prices, the tariffs, everything, all that affects every business in a way.
[00:12:12] Almost none of the businesses have been immune to that. And to be able to make some business decisions, you need to have up-to-date forecasts and budgets. And the problem here is, as I mentioned earlier, that process of creating a new forecast is really, really expensive. And it takes time, you know, it takes time to coordinate that within the whole team. And the idea is not to do one budget and maybe one forecast a year or maybe quarterly.
[00:12:42] But what we see with clients that implement first year, they do it, you know, at least quarterly, but most of them are transitioning now to do it on a monthly level. And that's even before kind of the AI workflows that we're introducing now. The kind of the ideal would be what we call continuous planning, where every time you get new actuals, your forecasting model gets updated automatically by that deterministic model that you built.
[00:13:09] And then humans in the loop get notified, and then you either tweak it or just accept it. So yeah, we see the tendency of doing forecasts more often, but you need to have a way to reduce the price that drastically. And the workforce kind of AI workforce functionality that we have allows you to actually do that. You would express in natural language, like describe how opening a new forecast works.
[00:13:37] And that would be saved as a kind of workflow. And AI would be able to run that, you know, either on a certain trigger, like actuals are loaded, or, you know, you can trigger it manually. And with that, we estimate that the kind of forecasting process costs will go down by 10 or 100x. And that's then the kind of goal we're looking for.
[00:14:02] So that running a new forecast is almost like insignificant costs. You can do it in one afternoon, and then you just notify responsible people. They either approve or do small changes in the plan and ship them back to you.
[00:14:17] So cost reduction, cost and time reduction, and ensuring that goes without saying, you know, if you have less manual work, there's less chance of something going wrong or introducing an error, which is a great big part of budgeting and forecasting processes. And in terms of an example, we see a couple of clients basically what they would do. They would prepare board decks every month when they speak to their board.
[00:14:43] And there would always be some questions which they haven't prepared because they cannot just recalculate thousands of like KPIs, you know, they need a little bit more time for that. And what they did, they just prepared the kind of executive summary outline. And what they do live is, you know, do you have any questions, board usually has, they would open Farseer AI Analyst Live and just ask those questions. And if something's kind of suspicious, they would ask it to break it down.
[00:15:13] And that discussion then would become really dynamic in a boardroom. And this is what they've been doing for several months now. And it completely kind of changed the way they work there. So they don't have to tell their board, like, give me three days, I'm going to come back with an answer. They can make a decision right there and right then.
[00:15:33] And I think these stories are so important because in particular, the world of finance, people will want to know about return on investment, measurable difference and value that any tech tool will bring. And I had a quick look on your LinkedIn page before we sat down today. And in a former life, I used to be involved with managing the month-end process. I used to dread it every month.
[00:15:56] And for that reason, one of the stories I saw on your LinkedIn page says that you can or you have accelerated a month-end process from 10 days to two, which just completely blew my mind. Are there any other stories or use cases like this or feedback that you have that will just help people listening, the kind of difference that you're making here? Yeah, I mean, I don't know when we made that pause, but today's sound sounds way too much now in the world of AI.
[00:16:25] I think you really can do that much faster. The month close should basically happen instantaneously. As soon as you get actuals, the system runs in the background. It automatically reconciles the numbers and just kind of flags, okay, here are five decisions. I need your input to actually complete this. And this is where we're going with the workflow.
[00:16:48] Because when you load the actuals and then you have to go manually through everything until your checks are zeroed out, this can take a lot of time. And it actually happens. But other examples are just increasing the cadence of forecasts. Instead of doing one budget and one forecast a year, now you're doing rolling forecasts 12 times a year.
[00:17:10] And forecasting for the next 15 months, this is something that almost every client that implemented Farseer is out there. And just the available data from the forecast allows them to make those decisions much faster. And the third kind of example there would be the scenario analysis. I think it's one of the things that everyone kind of talked about, but no one actually did until they get technological opportunity to do it with tools like Farsier.
[00:17:40] And other tools probably do do it also. But if you want to simulate effects of certain key metrics that drive your business and you want to see their influence on your top line or bottom line, again, it's similar to reforcasting. You have to copy the whole model. And depending on which level you change the numbers, you need to back calculate everything. And if that model is spread out in 100 excels, no one's going to do that.
[00:18:04] But if you have one foundational model, one governed model that you can trust and you change the number on one place and then everything recalculates in real time, basically you can do as many simulations as you want. So that's kind of the, I think, best examples of how using a tool like Farsier can help in those processes. Awesome.
[00:18:27] And I'm curious as well, as we look ahead, as Farsier expands from Croatia into the UK and US, any differences that you're seeing in the way finance teams are approaching technology and transformation and the kind of questions and things they're struggling with? Any trends you're seeing there? To be quite honest, like we've been in Adriatic's region for some time until we kind of build our product.
[00:18:52] And we're sure that we know enough that we can do a global expansion now with the help of a series A that we got to several months ago. This is what we're actually doing, like expanding to kind of UK and US our primary goals. And to be quite honest, like it's very, very similar everywhere you go.
[00:19:11] So like finance teams, like in Croatia, same in UK and US, they still struggle with fragmented data, lots of excels, manual reconciliation, and just spending too much time on manual tasks, which are error prone and slow, to be honest. What's the main difference?
[00:19:56] So many great examples and use cases there. But anywhere else you'd like me to point everyone listening today? Yeah, I mean, I think the LinkedIn is a great resource to read about us contact me directly or any of our four founders, we will gladly interact with you. Obviously, farseer.com, you know, as a starting point for sure there. But yeah, LinkedIn, DM me if you have any questions. I try to be very responsive.
[00:20:21] So yeah, looking forward to meet anyone who wants to learn more about this. Awesome. Well, I, for one, have loved hearing about farseer, especially the AI there, the agentic capabilities, how they automate many of the repetitive finance tasks and run continuous scenario testing, deliver proactive insights directly to decision makers. There's so much gold in there. So I urge people listening to check that out.
[00:20:47] But more than anything, thank you for taking the time to sit down with me today and share your story and that origin story too. Really appreciate your time. Yeah, thank you for having me and looking forward for some future opportunities. We have a lot going on in farseer and a few things that we haven't talked about in this podcast should be out in a couple of months. So I think that this should be really interesting to see what we have to offer to market. So looking forward to talk to you again, possibly.
[00:21:15] Oh, you've left us with a teaser. Absolutely. We'll make something happen in the new year. It'd be great to find out more. But thanks again. Thank you, Neil. I think one of the most useful points from today's conversation is also wonderfully simple. Yes, AI can interpret the question, but it's finance that needs a trusted system to calculate the answer.
[00:21:39] So before just adopting an AI analyst, teams need consistent definitions, reconciled data, a governed financial model and traceable route back to the system of record. And these foundations can support faster forecasts, continuous planning and live answers in board meetings.
[00:21:59] But without them, AI could run the risk of simply amplifying the confusion that is already hiding across ERPs, BI tools, warehouses and spreadsheets. So a big thank you to my guests for joining me and explaining how farseer separates reasoning from calculation and so much more. So you can learn more about the company and the links. They'll all be in the show notes.
[00:22:25] Farseer.com is the easy one, but I'll also put links to LinkedIn and all the social channels there over at Tech Talks Network. And if you want to send me an audio message, here's a quick bit of homework for you. Where should your finance team draw that line between AI assistance and deterministic control? Send me an audio message. Let me know your thoughts. I'll be back again real soon with another guest. But thanks for listening as always. Bye for now.

