Olga Rudakova
Good morning, good afternoon, good evening, ladies and gentlemen. Let me welcome you at our FP&A Trends webinar. My name is Olga Rudakova. I will be your facilitator for the evening. I'm FP&A professional and FP&A Trends ambassador. Today we are lucky to have an interesting topic of FP&A roles reimagined for the agentic AI era. And this topic has attracted almost 450 registrations from around the world. Let me start by presenting the agenda for today's meeting. We will start slowly with defining what AI agent is. Then we will go to agentic AI in FP&A, mindset, culture, and capability. Then we will discuss agentic AI, whether it's for everyone or at least for everyone in FP&A team. Then we'll have a look at agentic AI in finance, unlocking the value for FP&A through technology, through modern platforms. We'll have conclusions, recommendations, and Q&A session. And now I would like to introduce our speakers for the event. So please join me on camera and I will start introducing you one by one. We'll start introductions with you, Olha. Olha is Market Finance Director for Western Europe at TMF Group. So Olha Zotova is an accomplished finance leader with over 10 years of experience in FP&A, strategic planning and business partnering. She specializes in driving data informed decision making, optimizing financial performance, and leading cross functional initiatives. Olha, thank you for joining us tonight.
Olha Zotova
Thank you very much, Olga. Very happy to be here.
Olga Rudakova
Our second speaker for the event will be Harsh Amarasuriya. He's a global head of a FP&A finance transformation data strategy at Clyde & Co. Harsh has over 20 years of finance experience. He has developed extensive experience being a key strategic financial business partner to senior management at multiple multinational organizations. Harsh, thank you for joining us and accepting our invitation again.
Harsh Amarasuriya
It's a pleasure. Good to be here. Thank you.
Olga Rudakova
And our third speaker today will be Dominic Nguyen, FP&A Technology Advisor at Wolters Kluwer CCH Tagetik. Dominic has more than 25 years of experience in corporate performance management. Before joining Wolters Kluwer CCH Tagetik, he built his extensive knowledge of the FP&A industry at Oracle, Workday Adaptive Planning, IBM, and Accenture, working in roles across sales, pre-sales, and consulting for a variety of customers and industries. Dominic, thank you for joining us.
Dominic Nguyen
Welcome, everybody. Thanks. I'm excited to be here.
Olga Rudakova
Let me now briefly introduce FP&A Trends Group for those who don't know us so well. I will present several key numbers. And since we are at FP&A Trends webinar now, let's see that we have actually conducted 185 such webinars since March 2018. We are starting another round of FP&A Trends survey and at the very end I will let you know how you can take part in our survey. We've been conducting those since year 2017 and by now we had over 2,800 respondents taking part. More than that, at the very end, I will share with you how you can connect to us, to our community, using different channels. And let's look at some numbers now. We have over 57,000 members in our FP&A Club LinkedIn group. We are very proud of those numbers, and we are very proud that we can connect with all of you. What we can expect today at the current session, that is one hour event. You will take part in two interactive polls on the webinar subject, so be ready to vote. You can take part in Q&A. You can actually start sending your questions to our panelists right now through chat at the very end. Please don't leave before taking part in a brief survey. We collect feedback. We would really want to know how we can improve further. I would like to thank our sponsor for the event, CCH Tagetik by Wolters Kluwer. Embrace the future of finance with AI-based corporate performance management software. Now a couple of introduction slides from myself. And I would like to speak about LLMs, AI agents, and all other terms connected to AI. Since AI came to our lives so quickly, it's not surprising that we are very confused in how to call things. And we really have a kind of mix of all these notions and definitions. So I think it would be right just to start with defining key terms and make sure that we align on how we use them throughout the session. I will start by LLM, large language model. that we all know them, right? That's ChatGPT we are using or Google Gemini or Anthropic Claude. So think of LLM as a brain that can talk but has no hands. Well, that's the simplest definition I could find. So it's a smart text engine. It reads and writes language very well. It can answer questions, it can summarize information, but it cannot act. It cannot press buttons, send emails, call, well, Take action. So run processes, decide when to act. Automation as such, without any AI involved. Think about it as hands without brain or a conveyor belt. Automation was around long before AI. We've all used robotic process automation. And that's just a set up process without any thinking involved. If X happens, do Y. For example, if new line appears in Excel, send an e-mail. That works. But as soon as input changes or somehow nonstandard, that process might break. AI automation. Think it about hands with brain attached. But it's not an agent yet. It's, let's say, LLM included in automation chain, automation process. It can be used to deal with this variety of inputs. Let's say recognize more invoices in non-standard format to then put it into accounting system. Or it can sort out emails. Let's say many emails arrive to company general e-mail address and LLM can help this automation process to forward them to the right department. Or it can, let's say, do a sentiment analysis. Let's say you own a hotel chain and you get many reviews from your customers. You want to sort them by being positive, neutral, negative, and just create some statistics. So LLM built into your automation process is a good way to do that. But it's not an agent. If you think about AI agent is already a goal driven digital worker. It has a goal and it can decide, and I'm not saying it can think, but it can decide how to reach this goal. So it can break it down into steps and access tools to perform those steps. Obviously not pressing the buttons literally, but it can access it through APIs, through different connectors. and eventually can perform tasks for you. You can think of AI agent as a junior worker who needs lots of instruction but can think. Right tasks for AI agent can be, let's say, prepare competitor analysis. An AI agent will go, break it down into steps, research, summarize, analyze, maybe visualize. So many, many steps can be involved. Having defined all that, now it's time to think, are we ready for that? Are we ready for AI agents, junior workers in our FP&A teams? And who else and Olha Zotova can answer this question? Olha will be speaking about agentic AI in FP&A, mindset, culture and capability. Olha, let me pass this imaginable microphone to you and leave the stage to you. Enjoy
Olha Zotova
Thank you very much, Olga. I come from TMF Group, a global partner that provides accounting, tax, payroll and entity management services. And today I want to talk to you, are we ready actually for the agentic area of P&A? What does it mean for us and how to get to that step? Because I believe there is a lot of excitement around this topic, but also there is a lot of fear, what is going to happen to our roles, what's going to happen to our jobs, what the future holds for us. And quite often, I hear from many companies and many organizations saying, we are not ready for agentic AI because our data is not ready. And yes, data is important. But is data the only prerequisite that is required in order to start using Agentic AI? I don't think so. I think you need to have at least three things in order to make sure that you are ready for that stage. Data is very important indeed. Your data needs to be clean, but your data do not need to be perfect. You need to have consistent definitions. You need to have preferably single source inputs. And what's more important that the data that you have needs to have minimum level of manual manipulation. Because typically, AI is not good when you need to manipulate a lot of data manually. The other pillar, which I believe is not the least important, but even more important, the data, is actually culture and mindset. Are we ready as humans Are we ready as people actually to the next stage of using agentic eyes? And that comes to a lot of many sub steps that I'll describe and show to you later on. But you need to remember that culture and mindset and our people needs to be ready for that role. And the third pillar that I show is the system thinking, which means you need to be able to build your processes. You need to know how the processes are built from A to Z, Preferably, you need to have documentation actually of those processes. So that means that you need to know the clear steps and the exception that needs to be taken. And I want to highlight here that to my beliefs, culture and mindset are as important as data and processes you have in the company. And while you're preparing your data, maybe for the next era, it's important to take care also about your teams and to make sure that your teams are ready for that step. So it shouldn't be like one step is happening under another. I'm afraid you need to work on all three together to make it happen and to make it a success for you. So today I want us also to focus a bit on the culture and mindset and what that means. Because in reality, there is a lot of fear and a lot of scariness going on around saying, are we going to lose our jobs? At the end of the day, we're not going to change what we do. We're going to change how we do it. But that still is quite scary, right? So how do we do it as leadership and what do we need to do as companies in order to make it happen? First of all, it always starts from a culture. The company and your team needs to have the right culture. What I mean by that, first of all, it's psychological safety. People should not fear that they will be punished in case they fail. Because guess what? With agentic AI, you will need to try to test. You're going to fail, then you're going to try again, then you're going to improve. It's a process. You're not going to get there in one day. So people should feel encouraged that they need to try to try to try and eventually get there. The second step, besides of culture, it's about mindset. We find as people are quite often coming from historical looking, backward thinking, and now we really need to focus into the forward-looking thinking, right? We need to make sure that we change the ways how we work, and that's completely okay. But don't forget that we are actually not losing the control over it. On the opposite, we are getting more and more ownership because every agentic AI, every agent that you're going to create in your company will need to take an owner, a person who will schedule rules, cartrails, we can say, for this agent and who will ensure that it works properly, exactly as expected. And who is in a better place than us to tell how the work actually needs to be done. And of course, we need to see AI as an enabler. It's not a threat. It will help us. It will not stop us. The third important pillar, besides of culture and mindset, we need to have capabilities. And that means, first of all, we need to make sure that our teams have time to experiment, right? Quite often finance teams are so much overloaded with day-to-day work that they simply don't have capacity to try something new. I guess each of us can remember at least one time in our life when that happens to us, right? So we have to ensure that people have capabilities, time capabilities. What does it mean that maybe we can have to make part of the work that they do. And it doesn't mean that we need to do it with AI right now. We can do it in a very old ways. But the point is that people need to have time, right? Yes, processes are very important. So we need to have a roadmap of processes. The processes need to be documented. Ideally, it's important. And we need to have cross-functional collaboration. Because let's be honest, we are not going to do it in finance on our own. We will need data specialists. We will need IT, we will need sponsors to this project to make it work. So it's not only up to us, it's a lot about collaborative environment. And then the last pillar is talent. And when I think about it, right, what are the chances we're going to go to the market right now and find people who already have experience working with Agentic AI, right? I guess we all understand that it's quite minimal. And probably we don't need it. because we can grow that internally. The most important point is that we know the business, we know how the things needs to be done, and if we have the right culture, minds, and capabilities, we can also grow our own talent that is required. And speaking of this hybrid model, by hybrid, because it's a combination between people and agents, it's a completely new model that we are going into, so where to start? Few simple steps. What do we start from? First of all, create capacity for your team. It's very important. Secondly, identify the areas that you can give to agentic AI. And look, Agentic AI will not fix problems for you. So if you have an issue that you don't know how to solve right now, that is not the right area to give to AI. The right would be where you have a very clear process and you can describe all the steps and all exceptions that you have to this process. Assign an owner to that task, to that process, and to that AI, to that agent that will be built eventually, right? And please, Do not forget to celebrate wins. This is super, super important because again, we will sometimes fail, we will recover again, we will fail, again recover, we will learn from it, and eventually we will get to the point where we will thrive. But celebrating early wins is very, very important. So to me, when I ask myself, what is the biggest challenge? Is the biggest challenge technology? Now technology is actually on the market, the biggest challenge I believe is resistant to change. And it's up to us to embrace the challenge. Thank you, Olga. Returning back to you.
Olga Rudakova
Thank you for a nice presentation. And that's really indeed an inspiring ending call to action, not to be afraid of change and getting there through this iterative process. Thank you for that. If you have any questions to Olha here, just don't hesitate to send them through the chat. We'll try to answer most of the questions at the end and some in writing after the session. Or what will happen now? And now I will actually ask you a question. It's time for our first poll to the audience. So what is currently the biggest barrier to adopting AI agents in FP&A? So let's launch this poll. And I'm really curious to see how you will answer. So the question is, what is currently the biggest barrier to adopting AI agents in FP&A? Data quality and structure, FP&A skills and capability gaps, culture and mindset, technology limitations. I'm really interested to know how much Olha's presentation affected you. Obviously you have your experience and challenges you personally face in your organisations. So let's wait for majority to vote. And you're an amazing group today. You vote very fast. Most of you have voted already. So let me end this poll and share the results with everyone. We have a winner, but not in a big majority. So data quality and structure gained 35% in the poll. Second place, 27%, goes to culture and mindset. And then almost even results, 19-18% go to FP&A skills and capability and technology limitations. Olha, what do you think? Are you surprised?
Olha Zotova
No, I'm actually not surprised at all. That's what I see on the market and that's what I have experienced as well. We still struggle with data quality and structure, right? Even though we have been talking about it for years now, but that is still the case. And more assistance, more ERPs we have, I guess we even more complicate the situation. Happy to see that there's also culture and mindset. Happy and not happy at the same moment because happy to the fact that we do realize that this is a challenge. It's important because realization of that fact means that's the first step to resolving it, right? Not happy, but understandable. It's also the challenge for us because for us being probably conservative from a finance nature, right, that for us it's difficult sometimes to embrace a change. But I'm happy to see that we're all very much aligned in the process. And again, I think if we're all focused on data quality and culture and mindset, we're really going to make a huge step ahead in all our things.
Olga Rudakova
Perfect. Perfect. And let's move on to our discussion. So let me invite-- close the poll, and let me invite Harsh and Dominic to join our discussion. So we have discussed that, obviously, AI agents are coming into our teams. And even though we face some barriers, it's almost inevitable. So which FP&A skill do you think? or which have been skills are becoming obsolete and which are becoming non-negotiable today with the presence of AI agents. Harsh, let me start with you. What do you think?
Harsh Amarasuriya
Yeah, I think from an obsolete perspective, I don't Firmly believe that any of the skills that people currently have will be truly obsolete. But I think the data wrangling perspective, lots of our teams spend lots of time getting data together, loading them up into different databases, loading them up into different Excel spreadsheets, things like that. will become a lot easier. But to Olha's point, you've got to have an owner of those agents to make sure that it's doing the right thing and stuff like that. So again, that skill will be useful, but as to how much time FP&A people spend doing it, I think we'll get a lot less. But the non-negotiable skill that all FP&A people will need to-- they already have it, but they'll need to have more is the storytelling, the interpretation, what does this mean, what does the numbers mean, is becoming more and more important because you can get a lot more data a lot quicker with the agents. Then you've got to interpret it and sort of make a decision based on it.
Olga Rudakova
Right. Right. Thank you for this answer, Dominic. What do you think? Are there any skills that are becoming obsolete today and other that are becoming non-negotiable?
Dominic Nguyen
Good question. Just kind of building on my colleagues here. I kind of take a look at it from a kind of 80-20 rule. So I know we're spending a lot of time in history. with data. So I think things that are repetitive, like auto mapping, volumes of data have typically taken a long time and we spend less time on analysis. But with the advent of the data agents, I think we're going to be able to flip that and do the routine tasks. And we're never going to replace a 20 year veteran in FP&A space that has got a lot of experience in analysis and strategic thinking. I think we're going to be able to switch that pendulum a little bit to spend more time on that. And that's not.
Olga Rudakova
Great answers. Thank you for that, Dominic. Back to you, Olha. What do you think? Which are skills are becoming obsolete and which are becoming non-negotiable?
Olha Zotova
I completely agree with what Dominic and Harsh just mentioned. This is a transformation process using Agentic AI. It will not happen overnight. It will take time. So I don't think any of the skills will become obsolete immediately. Maybe within time, data analytics will become less important than it is now. But I definitely think what everyone can benefit from is from investment in soft skills. It's what Harsh mentioned storytelling, but also general communication. Because building AIs will require a lot of cross-functional communication. And the way how we communicate will change a lot for us how successful the project will be. So this is something I would definitely recommend investing in.
Olga Rudakova
Right. Thank you for the great answers. And by the way, so many questions are coming up your way. So brace for Q&A at the end. Meanwhile, let's move on to the second presentation of the day. Harsh Amarasuriya will be speaking about agentic AI, is it for everyone? So let me pass this virtual microphone To you, Harsh, and I'll disappear myself. So enjoy.
Harsh Amarasuriya
Thank you. And yeah, following on from all as a discussion, I think it's a timely sort of set way to actually ask the question, you know. well, what do we actually use these agents for and where do we, you know, where can we use their help with? So these are a couple of real examples that we are beginning to use in my own organization. I'll sort of briefly explain some of them, but, you know, I appreciate that all of you people, all of the audience wouldn't necessarily understand my industry, I'm in legal, but I think that the premise of using bots in these sort of ways is probably equally applicable in whatever industry you actually are in. I think bots are becoming a lot more used. I mean, there's a bot on this presentation. I'm sure there is lots of bots being used in our own organizations and things like Copilot and other software is having their own bots integrated into it. But I think what agents are enabling us to do is to customize those bots for different functions as well. And what I call out here is the FP&A chatbot that we've implemented at my organization, which essentially is sort of a frequently asked question automation. You know, there's a lot of questions that get asked every month. I'm sure all of us recognize that. You know, the answers can be easily got by navigating the dashboards and understanding what the numbers mean. But every month, you have different people asking the same sort of questions. So within my team, we've developed an FP&A chatbot, which is working quite well, which sort of links up different data sources, different dashboards, has some degree of automation. But I must admit, we are at the infancy. So to Ola's point, it's also changing people's mindsets about being confident, asking a bot as opposed to asking an individual, which they've done all their lives. But that's a very good example, I think, of making it easier for everyone, making it more self-serve for the end consumer. making it less repetitive for the producer, who in this case is an FP&A individual. We've also used it quite a lot across updating our disparate system architecture, depending on which business you are in that would resonate with yourself where you might have a mix between old systems and new systems and all of them don't talk to each other. And historically, even currently, we have lots of teams that essentially are moving data from one place to the other, being the gatekeeper and ensuring that systems work and they're updated on time and things like that. The example I've highlighted here is in our CRM systems, which is our customer relationship management systems, where different regions use different pieces of software. Some of them actually relate to the same client. So we tried to, historically, we had individuals extracting all that data and making sure that, if it relates to one client, it's all put in one place and, summarized for the right individuals and things like that, which was quite manual. But now bots are enabling us to do that at speed, you know, rapidly. Now I must highlight that it's probably not the best use of a bot and I'll explain why in a couple of slides time, but it is a quite a useful use of the technology to solve an immediate problem. The third one I've highlighted there is summarization. I think depending on which industry you're in, you're probably using a mix between bots and automation and technologies that do this. But the summarization capabilities in a lot of softwares, including things like Teams for meetings and everything else, is becoming quite a useful time saver. And I think a very good, efficiency enhancer as well. If I just move on to, and by the way, this is just an example. I'm not trying to sort of harp on any of the technologies on this slide, but effectively what I was trying to show here is the rapid growth in technologies, which then can be linked together using bots and other things. to actually do a multitude of tasks. And even if you look at something like OpenAI, which I'm sure everyone is familiar with, it's actually not just one technology, but it sort of tithers together lots of different things. including a thing called Harvey, which is out here. And it's quite big in the legal industry. So Harvey as a product is being used in legal, but that becomes a part of a bigger ecosystem when it goes into OpenAI and it's using OpenAI platforms and things like that. Now, this sort of back end is never shown to the end user, but that's what AI agents and multitude of technologies is enabling to harness at speed and that increases your capability. And finally, I'd just like to leave the audience with a couple of takeaways. Having sort of said that AI agents are That's very useful and it is, and it's enabling lots of things to happen. I think it's really important to actually understand a couple of things before you go down the route of, are we going to deploy an agent and what software we're going to use and what technologies we're going to use is actually initially to ask the question, what are you trying to solve? By the way, this is not new. I mean, I think those of us who've been doing projects and things like that, that's the first question you should be asking, even if you are not deploying agents but trying to do something else, right? So I think you've got to be very specific about what pain point you're trying to solve and what impact that will have on the business. before you decide on what type of technology is going to help you solve that. And agentic AI is 1 enabler and one type of technology that you can deploy. It's also quite important, I think, to Ola's point and to reiterate what Ola was talking about, is to understand the process and is it riddled with anomalies, exceptions? Is it quite standard? Do you really need business process engineering first? Or could you be using standard processes, but just do things quicker through agentic AI? I think a lot of our firms are sort of in the middle of that spectrum, you know, we do have processes that are quite mature and they're quite standard. We also have lots of processes that are very dependent on individuals. They have morphed into various non-consistent workflows. And I think we need to solve those before we deploy something like AI agents. And the third point is we got to be quite clear on what determines what the measures of success are for AI implementation. And what I mean by that is it's quite easy to deploy an AI agent. It does what it's asked to do, and then the whole team moves on to the next task and you forget the agent continues to do it. But if you haven't defined what defines success, and it could be error rates, you know, processing volumes, customer, value, how many customers you're getting. If it's something like a survey, how many surveys you're sending out a day or things like that. If you don't continue to monitor those things, I think you risk not knowing when agents are not doing what was originally intended or missing a step. And because you don't necessarily monitor it as closely as you may have, you tend to forget about it, and you're sort of spending your time on something else. So I think if you keep these three things in mind, then AI agents are quite useful, and they are becoming more and more useful. Back to you.
Olga Rudakova
Thank you. Thank you so much for your presentation, Harsh, for sharing practical examples with us. If you have any questions to Harsh, don't hesitate to ask them. And it's time for another poll for all of us. So we now will be speaking about which of FP&A activity is most likely to be delegated to AI agents first in your organization. And actually now we'll define areas, but I see that many questions that come our way asking for more and more practical examples. So we'll start with just defining in which areas we are most likely to use AI agents first. Is it data validation and reconciliation? Is it forecasting and scenario modeling? Is it management reporting and commentary? Is it business partner Q&A insights on demand, such kind of bots we discussed? Let's wait for most of you to take part in the poll to vote and see Anna that's amazingly how results are changing so data validation and reconciliation or forecasting and scenario modeling, management reporting and commentary, business partner Q&A insights on demand. Let's see. who will win, majority have voted. And actually, we have quite interesting results already. So let me share them with everyone. We see that first place goes to data validation and reconciliation, 46%. Then forecasting and scenario manage-- oh, no, that's management reporting and commentary, 23%. Right after that, forecasting and scenario management, 20%, and business partner Q&A or insights on demand, 11%. Harsh, are you surprised by the results or that's exactly what you expected?
Harsh Amarasuriya
No, I can't say I'm surprised. And I think it reflects the fact that most people recognize that where things are repetitive and data validation generally, once the process is defined, becomes quite repetitive. And to some extent, even creating your management reports and your standard commentaries and stuff like that can be based on rules. AI agents are quite good at doing those sort of things. The other two on here, forecasting and scenario modeling as well as business partnering, needs more human intervention and thinking. So yeah, I think it's quite representative.
Olga Rudakova
Super fully agree. Let us move on to another discussion between our panelists. And now we will discuss another aspect of implementing AI agents for FP&A. Where do you see the biggest benefit from AI agents implementation and why? So now we'll have this open question. And actually, now I really wanted to start with Dominic. Dominic, where do you see the biggest benefit from AI agents implementation and why, obviously, in FP&A.
Dominic Nguyen
Again, that's a great question. So as I think about this, I think the biggest benefit, in my opinion, as we build on the data is the volume of data. I think historically, small data sets are manageable, but as organizations grow, make acquisitions, and kind of expand their business, the volume of data is beneficial if you can use an AI agent because that processing time is going to be quicker, so that way you spend more time on other tasks.
Olga Rudakova
Right. Let's move on to you, Olha. What do you think? Where do you see the biggest benefit from AI agents?
Olha Zotova
Yeah, on top of what Dominic just mentioned, I see it also as a digital worker who can work 24/7, so do not require vacations, do not require sleep or eat, right? So we can actually give it a lot of work to be done that is repetitive. And as Dominic mentioned, for us to focus on something that is really exciting that AI cannot do.
Olga Rudakova
Perfect. Harsh, back to you. I know you gave us already many examples, but if you just say where exactly you see the biggest benefit from AI agents and why?
Harsh Amarasuriya
I think just to extend on what both Dominic and Olha was talking about, but I think granularity is the other one that AI agents are enabling us to do. So historically, because human beings used to do a lot of these tasks, you would if you had the top 20 things, you could only do the top 10, or you'll have some sort of a threshold. I think AI agents are now sort of making thresholds irrelevant, right? You can run through the entire data set, however small the anomaly, it'll sort of pick it up. So I think there's a lot of benefits in being able to do any task to any level of granularity, which wasn't in a manual world, always a reality without an army of people.
Olga Rudakova
Wow, that's great. We can actually redefine the processes, not only do them more effectively, we can set up completely different goals for our processes now. Thank you for all your answers. And we are approaching our last presentation for the session. Dominic Nguyen will be speaking about agentic AI in finance, unlocking the value for FP&A. Dominic, let me pass this virtual microphone to you. I'll disappear myself and let you take the stage.
Dominic Nguyen
Thanks, Olga. Yeah, so just kind of build on my colleagues. I'm just going to focus a little bit on some of the key value propositions, technology, but here's a quick agenda. I'll probably start off talking a little bit about the trends and kind of why we're here, some of those challenges, and kind of map that into the overall solution use cases and benefits. Very high level. I mean, why should we care about this? I think we're past this stage, in my opinion, of, hey, this is a new technology. It's here to stay. And I think that's the important aspect. Typically, with any new technology, we have two camps, right? Number one, we've got the early adopters and we've got the laggards. Well, you know, why should we care about that now? Well, to be in this technology advancement world, in my opinion, right? Getting that competitive edge is the key. A lot of organizations that are early adopters kind of yield benefits for the organization. And I think everybody is looking. There isn't a week that goes by that I have to talk a little bit about AI in some form of capacity. So I think that the message here is that it's here to stay. It's very important and very strategic, mostly to gain that competitive edge. Some of those challenges that we face and why that AI can help, I mean, obviously these are very strategic high level areas from keeping pace or increased data volume. But I think the ultimate efficiency around being able to transform those financial processes that you have, you obviously could start small and grow over time, but gaining those efficiencies, saving time, saving money, and ultimately a lot of benefits in terms of being able to have enterprise why planning quality is accuracy, but all of these challenges face growing demands in our CFO landscape. And I think we all understand that. So now just leveraging technology to kind of keep pace and kind of helping with some of these challenges that we have is very important. As we go through this journey together, I mean, there's two kind of key areas, right? Objective AI, Gen. AI, I think that's a growing area that's obviously going to get a lot better. But, you know, having those agents that are doing routine tasks right from the data transformation to actually having, you know, natural language to generate something to act kind of like a human is kind of where we're headed. So to be able to mimic us, learn from us and go through this journey together, ultimately all the way down from transformation all the way through to data analysis. So Very, very important in my opinion. So what are some of the key functions or areas here with that? I mean, obviously we. have this growing area to have things like, you know, on the fly data visualizations. I think that's a big area and maybe create a dashboard for us on the fly through our natural language. The reason why this is becoming very popular because finance organizations and people are very little tech, right? So it's not a high learning curve from IT where you got to learn some kind of script, right? We want to make it simpler for people. So speaking to that, using natural language and kind of having it do some of the things that traditionally in the past require a little bit more skills is becoming easier for us in the long term. So authentic AI for insights, right? Just some of the areas where the data agents can, you know, help obviously explore data, analyze information using our natural language text or voice, which is becoming the most popular area in my opinion, right? Kind of giving us insights. And I think it's we're moving towards that evolution, just helping us, you know, understand the data, helping us with commentary or recommendations is, I think, going to be a growing big area, just kind of helping us make those decisions. Doesn't mean that we're going to get to replace our experience, but it's kind of leveraging, augmenting that through some of these agents that we have. So And as well as kind of pushing it towards more the reporting analytics side, I spent some time talking to a lot of prospects over the last few weeks. I think a big area for them is executive reporting, right? That's always #1, right? Being able to do things in advance, collaborate. And still leveraging some of the common technologies that we have today, Excel, right? Kind of even though we have, you know, agentic AI and agents, you know, embedding that in Excel has become extremely popular. So give people the comfort of Excel, but the power behind some of these AI agents that will streamline our processes significantly. Quick use case here, just kind of leveraging A, where we can, like I mentioned, use natural language to build a quick dashboard and I think it's cool to see that, you know, these use cases are allowing people to do things that typically in the past has taken a lot of time and effort. Right? A simple dashboard is part of the executive reporting part of that. insights into the business so you can kind of quickly see that, you know, we're able to do a lot of this stuff. And here's a video kind of in action, you know, taking a data set that is common that we can leverage and just either typing or just, you know, speaking to the agent and it's going to be able to quickly build that insight of the dashboard. So in seconds, right? So that's the key, being able to adjust that, build quick dashboards or reports, spend more time on the analysis component that we all realize is very valuable. As we build down this path, there are also kind of some of the blocking and tackling tasks where the objective AI things are very valuable, like AI anomaly detection, right? So be able to see those inliers or outliers in finance, you know, with the large data sets that help us identify variances or anomalies, right? You can also customize and set some of those key thresholds But in essence, this is a big growing area that allows us to analyze and detect information. Secondly, also predictive planning is growing as well. Obviously, planning is a big area for FP&A. Traditionally, it's driver based, it's bottoms up, top level, but the ability to kind of predict and leverage some of those drivers, whether it's internal or external, to kind of generate a forecast quicker to derive those predictions. And one of the big areas with the AI agents is to be able to learn some of this stuff and be able to help us go through and reiterate this process very, very quickly. So I think that's a big area of predictive planning. And kind of building a use case here, we have a top automotive company that leverage the AI agent for auto mapping. I talked about being able to do so through team tasks, but also leverage the predictive planning capabilities to help drive sales revenue. Ultimately, intelligent analytics, leveraging these data agents will help speed up the process, right? give us more quick data insights, better reporting and dashboards. But overall, this organization was able to realize extreme value in terms of being able to improve those capabilities through the AI agents. And lastly, to kind of sum this up, I think we're all here to learn. I think we're all here to improve our lives. But ultimately, the key value is higher productivity, right? We're enabling time and effort that is leveraging technology to streamline that through objective Gen. AI agents and all the capabilities that are available to us. That way we can then focus on what's more important, right? Be able to kind of broaden our skill sets, be able to save time across a work week, be able to spend more balanced life. But overall, I think we're excited to be here. And I think all of us are as well as we embark on this journey together.
Olga Rudakova
Thank you. Thank you, Dominic, for a nice presentation. And thank you for sharing several use cases because people are asking a lot about practical use cases of AI agents. And I actually plan to come back in Q&A to that. So now let's jump right to Q&A session because we have received many, many questions today. That's amazing session in that sense. Heads up a bit. There are many questions asking about use cases for AI agents. So I will at the very end ask all of you about your favorite AI agent use case so that you already have a plan to build in your team. But meanwhile, let's go through several. open questions before we get to use cases. Just wanted to get you prepared. So first question now addressed directly to you, Olha, came during your presentation. So based on your experience, what are good areas in FP&A to start implementing AI agent? Please provide a couple of examples.
Olha Zotova
Very good question because I guess we are all wondering where to start. I don't think there is one rule that will be applicable to each and every company because we are so specific, but there are general rules that you should look out for. The task need to be repetitive. You need to know the process from A to Z, the end process, and you need to know the exceptions about it. Typical examples would be some reconciliation files, merging of different reports for one management report, providing comments into the board presentation for the performance of the company, but just don't forget to check it always, right? But those would be typical things actually, then you can watch out for. But in general, what I always recommend, do a step back, look into your processes, look what takes most of the time. Is the task repetitive? Yes, great. Do you know how the task is performed from A to Z? Great, yes, okay. Shall we try to move it to Agentic AI? Is it simple enough to try it and that if we fail, nothing merger is going to happen? Yes, great. You have your user case that you can try it.
Olga Rudakova
Perfect. Thank you. Thank you for this answer. Harsh, I have a question that addressed you a period during your presentation. Just please keep answer slightly short so that we can speak about your favorite use cases of agents. So Harsh, can you share how much time and trials it took you to implement your agents?
Harsh Amarasuriya
The first one took quite a long time. I must admit, I mean, we didn't know what we were doing either. So you generally will need some help by external support or something like that. But once we got to the second and third iteration, we were quite quick. So some of my team members, about two or three people who are now quite good at smaller sort of automations and using bots, managed to get a few few extensions of our chatbot up and ready in two or three weeks after testing and things like that. So the first one took us several months.
Olga Rudakova
Sure, perfect answer. And I have actually a related question which I would like to address to you, Dominic. So once data and processes are established and in place, how do you move forward with AI agents? Do you create AI agents in-house with IT or you and data specialist support or it's best to have off the shelf solutions providers? What do you think?
Dominic Nguyen
Oh, great question to kind of balance between, you know, internal versus external. I think as AI is a very new area. It's always great to leverage experts and their experience initially. And I think the best practice is always to kind of be a hybrid approach because it's going to be. leveraging experts externally, but as we learn, kind of train the trainer approach, obviously all organizations want to keep it in-house over the long run, but I think it also depends on your organization's size and resource availability. But I think the trend is to kind of be able to manage this over time internally, but I recommend starting with experts at the beginning.
Olga Rudakova
Amazing. Thank you for all these senses and for keeping them short. So now I'll ask each one of you just describe very specific but shortly your use case for AI agent, what it can do already does for you and your team. Olha, let's start again with you.
Olha Zotova
We're a bit too early, actually, in my company for the agentic AI. This moment of time, we are really focusing on getting the data ready And the team's ready because we're very global. We are present in 87 countries. So for us, standardization is a big thing. And this is something we started to focus two years ago, and we plan to finalize the journey now. I personally use a lot LLM every day, starting from, hey, here is the original file. Here are the comments I made. Here is a new file. Can you do the comments absolutely in the same format? Works great. all the emails and all that kind of stuff. Actually, I became personally much, much more productive. And I do hope that we will be looking to agentic AIs actually in the future in my company too.
Olga Rudakova
What you are describing already can be transferred into a agentic AI mode. Let's say you are in a car late for the meeting. You can say, hey, through whatever communicator you use, WhatsApp, Telegram, hey, what's the meeting I have first? Can you please message participants that I'm 5 minutes late? Using the personal assistant already with the voice control and you're getting there. Harsh, what about you? What is your favorite so far use case for AI agent in FP&A?
Harsh Amarasuriya
I mean, like Olha, I also use LLMs and stuff like that for my personal use. And they've become a very useful assistant to do various things, summarizing, presentation, enhancements, things like that. But I think from a work, pure volume perspective, we've also used it for, things like getting statements out, setting out things that people used to do individually. We've now sort of created, I mean, we have lots of people, like Olha said, around the world. Every month we have to e-mail lots of different statements, all of those things, which have now become almost automated, right? And there's a whole series of agents who sort of take care of that. which is very productive because even the human beings used to do it, there was no real value add that they could add to it. So it's a good use case for them.
Olga Rudakova
Indeed. And in a law company, obviously, AI agent can do whatever junior lawyer would do, like prepare the first draft of any contract.
Harsh Amarasuriya
Okay.
Olga Rudakova
The one to see. Exactly. And it's already, if you give it hands connected to tools that can pull in data. Dominic, what's your favorite AI use case? One practical example.
Dominic Nguyen
I'll answer it in two ways, professionally and personally. So professionally, I have found the data validation area with the agents being able to, you can talk to it and say, hey, did my data load correctly after an argument just task of all the data and it comes back and it helps you identify variances, anomalies with the data. I think it's been fun to kind of like manage that and kind of see how that is in action because I think it's extremely valuable from a data validation perspective. Personally, I think with FP&A and we're traditionally used to very routine, mundane tasks, it's really invigorated a passion. I think AI has kind of leveraged this new technology for finance to have fun, right? So kind of not just go to work, but I think building on that passion and fun that all of us in finance want. We don't want to be doing the same thing over and over, and I think that's the exciting part.
Olga Rudakova
Amazing. Thank you for all the answers. I really enjoyed the session. I would like to thank our sponsor, CCH Tagetik, for making this event possible. Embrace the future of finance with AI-based corporate performance management software. And I would like to ask all of you to join us if we are not connected yet and to follow FP&A Trends through any of the channels that works best for you. So choose the one. And please don't leave yet without filling in very short feedback form that helps us improve. Thank you all for being here. Hugest, the biggest thank you to our participants for sharing your wisdom, for sharing your practical examples with us. Have a good day, have a good evening, and see you at the next FP&A Friends webinar. Goodbye, everyone.
