Olga Rudakova
Good morning, good afternoon, good evening, ladies and gentlemen. My name is Olga Rudakova and I'm happy to welcome you at our FP&A Trends webinar. Today we are going to speak about the power of driver-based planning in the age of AI. So let's have a look what we have an agenda for today. We'll definitely cover driver-based planning and AI key concepts and key statistics. We will have a practical case study on moving from annual budgeting to monthly driver-based plans. We will speak about FP&A strategic velocity from Excel drivers to agentic planning. And we will cover practical adoption of AI in driver-based planning. We will have conclusions, recommendations and Q&A session. It is a pleasure for me to introduce our speakers for the event. Please join me on camera and I will start my introductions with Mike Uvarov. Mike is the head of FP&A at Viskase Companies. He's based in Chicago in the United States. He's the head of FP&A with a focus on margin analysis and FP&A tool implementation across the organization. Mike, thank you so much for joining us today.
Mike Uvarov
Happy to join and honored to join this webinar.
Olga Rudakova
Our second speaker today will be Vignesh Dumonceau, CFO at WAGO. He's based in Switzerland, and Vignesh has a proven track record leading financial transformation, M&A integration, post-merger integration, and redesigning finance processes across complex multi-entity structures. Vignesh, thank you so much for accepting our invitation and joining us today.
Vignesh Dumonceau
Thanks a lot, Olga. Nice being here. Thanks for having me.
Olga Rudakova
And let me introduce our next speaker, Graham Hunter. Graham is an AI and ML solution architect at Wolters Kluwer CCH Tagetik. Graham has a background in the corporate performance management, space plans innovation, product management, finance transformation consulting, and AI strategy roles. Graham, thank you so much for being here today.
Graham Hunter
Thank you very much. Welcome. Looking forward to the conversation.
Olga Rudakova
I'm looking forward to at the very beginning. Let me introduce FP&A Trends groups through key numbers, through statistics. We are at FP&A Trends webinar and we had 191 of such events until now. Today I will kick off by showing some key statistics from our fresh FP&A Trend Survey 2026 that will be available to you at the beginning of July. And I would like to say that we had 3,361 survey respondents until now. And if you join our FP&A Club LinkedIn group, you will be among another 60,000 members. So if you are not, please join our big community. What we can expect today at this session, that's one hour webinar. You will take part in two interactive polls on the webinar subject. You can ask questions and you can actually start right now through the chat box at the very end. Don't leave without taking part in a brief survey. That's indeed brief and it helps us a lot improve our webinars. I would like to thank our global sponsor CCH Tagetik by Wolters Kluwer for making this event possible. Embrace the future of finance with AI-based corporate performance management software. I will kick off with a couple of slides on some key statistics on driver-based planning and AI. And I would like to start with one maybe seemingly sad number. Only 1% of organizations use dynamic driver-based models supported by AI that automatically adjust drivers in real time. Well, it might seem like a low number. But if you think of that, that is possible. 1% of organizations actually already use it, not only driver-based models, but supported by AI and that adjust drivers in real time. Otherwise, driver-based planning adoptions is obviously much broader. Another 18% use fully driver-based models. Another 41 use some partially driver-based models. But that ideal state is also possible. Why I'm speaking about the statistics? Because driver-based models correlate with another amazing things in your FP&A teams. First of all, companies that use driver-based models rate their forecasts as great or good in 77% of cases. compared just to 35 for those who don't use driver-based planning. That's a huge difference. And actually, driver-based models correlate with things such as having optimized FP&A teams or percentage of organizations where decisions are made using data or amounts of time FP&A spends on high value activities. So we do see the benefits. Let's jump to AI. I have specifically pulled out statistics and use of generative AI. That's because we talk about it a lot. Well, the first insight is most of us don't use generative AI at all in FP&A so far. Many plan to, but out of usage, the most companies use it in 15% of cases for communication and decision support. That's what I see in practice as well, that work with text, with documents, with presentation, with summaries is already at quite good level. But when we look at process automation or for advanced insights, adoption here is very, very low. So that's where we can try to get to. and try to extract really strong insights using AI. Another relevant number, because we'll be speaking about driver-based planning, not only generative AI, but machine learning. And 12% of organizations use machine learning and another 17 plan to adopt it within next. six months. Let's look at impact of AI. Both of them, generative AI, machine learning, to, again, key parameters of FP&A teams. And we see that forecast quality, 64% for greater good forecast for companies that use AI and only certified for those who don't. Optimized teams, again, huge difference and it correlates as well with decisions based on data or amount spent on high value activities. What happens if we use driver-based models and AI at the same time? That is something that we will learn very soon from our speakers at the event and we will start with Mike. Mike Uvarov will be speaking about moving from annual budgeting to monthly driver-based plants. Mike, let me pass this imaginative microphone to you and this stage to you. So enjoy.
Mike Uvarov
Hello, everyone. First, let me start with a brief intro about the company. It was established 100 years ago in Chicago and makes cellulose and plastic casings for hot dog and Sausages and it has a 400 million in revenue and 2000 employees worldwide. There was some a few external shocks that we've experienced recently, a new competitor, supply chain disruptions and an M&A transaction and going from private to public. So the company embarked on the manufacturing footprint optimization initiative rigorous margin analysis by product and customer, and an analytical infrastructure update with the new EPM tool implementation. So the FP&A team was tasked with the monthly three statement forecasting with focus on cash, scenario analysis with what if, and an extensive communication with stakeholders with the board, the bank, the auditor, the M&A partner and internal stakeholders. So the initial situation was in our annual budgeting planning with the mid-year revision. And where we want to be, our desired state was a monthly rolling forecast. And here are the major building blocks that help us cross this gap. First, we started with existing basic drivers, such as volume and headcount. And we have two ERPs, so Excel isn't enough. So we needed a tool. And prerequisites for an EPM tool implementations are clean metadata, your accounts, cost centers, product hierarchy, transaction quality, and consistent data extraction and transformation. Interestingly, the same prerequisites are a foundation of an AI use in FP&A. So the same building blocks, essentially. And so next steps were to find relationships between drivers in the P&L buckets. And this is where we needed help of AI. And it did help, and I'll go into this. We also rebuild our P&L to show those buckets and to make respective departments responsible for the input. So on the left-hand side, you can see our P&L bucket. So this is our new income statement looks like. It was split into meaningful components, each of those with a specific driver. So this looks very much like a fishbone. diagram and some call it a Christmas tree diagram because it's vertical instead of being horizontal. And it shows for each P&L item a department, responsible department. For example, for volume, product volume of product, it's an SNLP department. And for plant controllers, it would be responsible for an absorption forecasting. Then it's a department and a driver. What drivers did AI help us find? Number one was a demand seasonality by product and customer. Number two is machine hours, manufacturing machine hours for absorption dollars and the waste percentage projection for production variances. Those additional forecast drivers helped us to improve accuracy and speed and also elevate the variance analysis. If you have all of the details and then you have a P&L structured like this, it's much easier to go and see what has happened, what are the reasons, and then And this would create a feedback loop for the future improvement of the forecast. So the takeaways, look for non-financial data for drivers such as machine hours or waste percentages and so on. And don't be afraid to modify your P&L buckets. in forecast with those newly found drivers for those P&L buckets. This creates those feedback loops and reinforces success.
Olga Rudakova
Thank you. Thank you so much, Mike, for bringing this case study and for sharing valuable tips, looking for non-financial drivers. And by the way, it's really nicely that you brought in that driver-based logic doesn't help us only to look forward and to plan or forecast, but actually to look back for variance analysis, absolutely same logic, understanding root causes through drivers. Thank you for that. If you have any questions to Mike, you can ask them right now through our chat box. Meanwhile, I'll have a question for you. I have a polling question prepared for all our participants. And we will be speaking about basis of your current forecasting process. So let me launch our first poll for today. And the question goes like that. What is the main basis of your current forecasting process? Is it annual budget with periodic updates, rolling forecast, driver based forecast or AI and machine learning enhanced predictive forecast? I'm very curious to see the results. I already see percentages running up. So what will that be? What is the basis of your current forecasting process? Annual budget, rolling forecast, driver based forecast or artificial intelligence and machine learning based option? Thank you for voting so quickly. We have majority of participants voted already. So let me close this and share results with everyone. We definitely have a winner, and that's annual budget with periodic updates. Second comes rolling forecast. Then driver-based forecast with 10% and zero for AI machine learning enhanced predictive forecasting. Or maybe we need a bit of more time to wake you up. I'm pretty sure that someone in the audience has optional before just didn't vote. Mike, what do you think? Are you surprised by the results?
Mike Uvarov
Predictably, the option #1 is where the majority of companies still are. And I'm super happy to see 10% of people selected option #3. That gives me a lot of confidence that it's possible.
Olga Rudakova
Exactly. 10% is a good number. Thank you, everyone. Let me close this poll and move on to the discussion between all our panelists. So Vignesh, Graham, please join us on camera. And what I would like to discuss with all of you is about this shift from annual budgeting to more dynamic ways. So what is the most important shift required to move from annual budgeting towards a more dynamic, driver-based planning process. And I will start with you, Gignesh here. What do you think? What's the most important to do?
Vignesh Dumonceau
Yes, thanks, Olga. So the most important aspect is what we want to achieve from scenario A to B. So if we were using a typical annual budget, it has its pros and cons. I mean, at some point of time, the budget gets old, the assumptions move, and so it's less relevant. So if it's a more dynamic and driver-based planning process, it's more lively and you would roll it out. But importantly, I think this is an alignment that is required with the key stakeholders to show the fact that being more lively, more transparent and more proactive, we can help them take faster decisions and therefore improve the numbers. So that alignment, I think, is extremely important.
Olga Rudakova
Right. Thank you so much for this answer. And I will move on to you, Graham. What do you think? What is the most important shift we need to do to move from annual budgeting to driver-based planning?
Graham Hunter
I would wholeheartedly agree with Vignesh. I think he said it well. Having been an architect on a few dozen driver-based planning implementations, I can say very clearly that when that alignment is not there, every single time it results in cost and time overruns. So having that alignment between the stakeholders is critical.
Olga Rudakova
Yeah, it looks like alignment becoming one of key words for today's session. Let's see how we go. Meanwhile, let me come back to you, Mike. From your previous experience, what was the most important shift for you to move from annual budgeting toward a more dynamic driver-based planning?
Mike Uvarov
Sometimes this comes from an external need to be faster, to be more reliable and quicker and more precise. Sometimes the world makes you look for those additional methods like the driver-based planning.
Olga Rudakova
Right. So external pressure or sometimes internal motivation. Wow, great. Thank you all for your answers. And that brings us to the second presentation of the day. Vignesh Dumonceau will be speaking about FP&A strategic velocity from Excel drivers to agenting planning. So Vignesh, let me pass this imaginative microphone and the whole stage to you, the stage.
Vignesh Dumonceau
Thanks a lot, Olga, so and welcome everyone. Happy to be here. So the theme we are going to move on is what is the next step after the drivers. Okay, so we'll take a small step back and look at what is the main reason why we do controlling since we've done controlling for a long time. It's really to read behind the numbers So to give some sense to what we see from a reporting standpoint. And then once we've given that sense and the reason, then we want to act. We want to take decisions. So we want to steer the discussion and through our ability to discuss with the business, influence decisions so we can have better numbers coming towards us. So it's not only looking back and then commenting a narrative of what happened, but being more an actor and an architect of what we want to achieve in the future. That's been always there. I think this is nothing new. There were different methods to it. So in the past, I mean, if I take a small step back and take one example, if you would use sales forecasting, you could say, look, I sold 10 million per month, so 120 million. Next year, I'll do a small price increase. I'll be around 130 million. I'll adjust a bit for seasonality or for other judgmental elements. So that's one element. A lot of companies do that, but we realize that we can do it better if we go on a driver base. So you would then say your sales is actually made of volume and price. So that's these other drivers, might be other. And then you start to investigate and understand how your volume is behaving by geography, what's your pricing positioning and so forth. So that's a little bit more complex. So it deepens the causal analysis and allows you to go one step further. Now as we move from the mandate that we receive, we move more from a classical backward looking reporting function into an orchestration engine where we have these discussions And we want to build those drivers and make them more lively. And that's where AI comes in, make them more proactive versus doing it on an Excel or on a reactive basis. So if we look at the maturity aspects, we have in our FP&A and forecasting and controlling, those elements of process, technology, and people that make all the forecast and the predictions and the analysis. And then the maturity is really whether we are backwards looking or whether we are able to influence and project the future and move the scenarios as we go forwards. So the more autonomous monitoring we have in our processes, the more we get support from drivers that are already given to us And the more our people become strategic. And instead of being in a typical data crunching position with limited dialogue, we have more discussions with our key stakeholders and more involved in the decisions the business makes. Now, if we look at one example, we've taken this example, you have three typical drivers in any business. You have the volumes, the headcounts, the machine hours. So the way we would do it, if you would do it without AI, you would probably put them on an Excel and then monitor those drivers separately. And then at some point of time, you would consolidate them and try to make sense. Now, this is possible. It's better than not having those. when the complexity justifies the accuracy of the forecast and the depth of the analysis. But further to that, if we would imagine having agents, AI agents, for example, who would continuously monitor the volume, have an impact on the capacity, and then based on the capacity, probably also be able to discuss with the headcount which is required in the right phasing of the headcounts. and then have the ability to train the accounts when we are in downtime on the machine hours. And when we integrate all those elements, probably with an impact on inventory, for example, not to impact the sales volume, it becomes a little bit more easier to build scenarios and have dialogues and discussions versus managing those in separate spreadsheets. So typically, this is the message, legacy run rate is a little bit, more based on Excel. There's no real causal link. And then when we have a structural shift, we have a lag. We don't always, we are not running ahead. Driver-based orchestration would be able to, easier said than done, and also we have to do it step-wise, but if we program it well, it should give us, or at least that should be the promise, an ability to have more causalities mapped and should anticipate structural issues. Now, as we move forward, if we take one example, I just took the example of absorption. You know, this is a typical scenario. We've discussed it also earlier where your volume drops and at some point of time you have higher fixed cost that impacts also profitability, profitable series, for example. And so the ability to maintain the rates properly, predict those fixed cost allocations, maybe to isolate them and to manage capacity is done through Excel. We can do that, but it's easier done if we have a integrated movement of AI based drivers that can predict the trends. So if we say, okay, we run at a volume 20% down, if we keep on this trend, then this is what's going to do to our profitability, to our overall cash burn and so forth. So having those elements predicting is much more powerful for us. So what breaks at scale is that at some point of time, doing all those analysis is very time intensive. One topic that we have which is scarce for us in our teams is time. Continuously scale it. We don't have always the right feedback loops. So what we would have in the agentic shift is precisely that, is the ability, let's imagine that each of those drivers are agents. So individual agents monitoring and they can talk to each other. Now we can now build a dialogue and we can talk in an orchestration loop where we get scenarios, we say, okay, your volume is dropping, you have a risk of underabsorption, and that could impact negatively on the pricing. Now, what are the actions that you can take? So you can either move some of those forecasts, demand forecasts forwards, so you can manage inventory if that's an option. You can, if it's structural, start engaging in a pricing discussion to cover some of your fixed costs. and so forth. So these scenarios, they become much more fluid and easier to implement versus having Excel based. So there should be really a discussion, and what would be the scenarios? This is what I project. These are the segments on which I can redeploy. And that discussion should be facilitated by agentic AI business partner type of simulation. So typically, I would say there are three benefits to it. One, we should be faster. So if we would move our data on Excel and Reactualize, there's a lag. If you have it automatically monitored, it's easier. It should be integrated. So we have a better understanding between the different drivers, and there is one coherent scenario. And then we should also be able to have a better forecasting accuracy. So at the end, that's what we want to achieve. We want to still, through those tools, we want to be able to enhance and support the judgment of all controllers. So a couple of numbers on what we think, and in my experience in different areas, These are targets. I mean, depending on your maturity scale or where you are, this might change. It's dependent on the environment, of course. But these are the three, I would say, cardinal values that we are trying to achieve. We want a better accuracy. We want less bias. So if you do this or if I do this, it should not be totally different. It should be more or less with the same parameters giving the same type of forecast. And then the driver should enhance them and drive a better stability. An example on the maturity model, you know, you traditional Excel, some kind of automation, and then the agenting AI would be able to continuously, autonomously refresh and provide scenarios. So at the end, you know, that's a little bit my conclusion here. If I look back a bit of the reporting and the way our FP&A teams and the function has been evolving, we were more towards number crunching, producing reports, putting some commentary. It grew from traditional accounting into analytical accounting. And then we moved into a driver based where we had the ability to break it down further and we entered a bit closer to the business. and the variable and the causality. Now I think we reach now, and you see it with Copilot or other any type of AI-based language, you have a dialogue which is possible through the agents. So the agents would be able to monitor those drivers, propose your scenarios, and then you have a discussion and validate those models. So our teams are more moving into those type of areas where they validate models, orchestrate, and make recommendations. So the AI enriched BP is more a dialogue, which I would call the new reporting.
Olga Rudakova
Thank you. Thank you so much, Vignesh, for your presentation. And I really like how you, among all amazing examples you brought, I really like how you made this connection between driver based planning and FP&A integration and driver based planning and business partnering, because driver based planning allows us to understand business much deeper than we could before, and that exactly enhances this integration and this dialogue. Thank you for that great presentation. If you have any questions to Vignesh, don't hesitate. Send them now. By the way, thank you so much for all these questions that are coming already in our direction. So it's important for us to know that you are listening, you are curious. Meanwhile, let's jump into the second polling question, and what I've prepared for you is really about understanding how planning drivers are managed in your organization. So let me launch this poll and see how you will vote. So how are planning drivers primarily managed in your organization today? Manually in spreadsheets, partially automated planning systems, centrally governed through driver models or AI supported or agent assisted. Yeah, that let's see. Let's see where we can get if we get some percentages in top categories. So manually in spreadsheets, partially automated in planning systems, centrally governed through driver models or AI supported or agent assisted. Let me wait for the majority to vote. Thank you so much. You're voting quite quickly so I can close this poll and share the results with everyone. We do have a winner. Manually in spreadsheets got 51%. Right after partially automated and planning systems 46%. Centrally governed through driver models 4%. Wow, we have some numbers there. Vignesh, what do you think? Because last answer, AI support didn't get any votes, zero. So what do you think? Are you surprised by the results?
Vignesh Dumonceau
Well, that's a great polling and I'm very excited to see those results. No, I'm not surprised. I think we still have in a lot of organizations a very traditional FP&A, which is slowly moving into automation. AI is new to a lot of us and so I'm not surprised that this is still a new area where we want to venture step by step. But yeah, that absolutely reflects what I had in mind. Yes.
Olga Rudakova
Great. Great. Thank you for your comments. Thank you all for your answers. We are waiting for more questions to be, and meanwhile, let's have a short discussion between all our participants because obviously when we speak about adoption of driver models of AI or both of them, it's a lot about dialogue. So how should finance leaders explain the value of AI-enhanced driver models to stakeholders who care more about outcomes than methodology? And I will start here with you, Graham. How do you think finance leaders should do that?
Graham Hunter
Excellent question. The crux of it comes from the driver-based models are where finance speaks to the business units in their own language. And it's really important that they speak in the outcomes that are there. And the number one lever that finance has is the budget. So the driver-based planning implementation will generally facilitate that conversation, say, we have the same objectives, business growth. Let's align on your outcomes and we'll align our budget practices and our investment to your outcomes. And that's a great conversation starter. and beyond.
Olga Rudakova
Right. Thank you so much for this answer. And I'll move on to you, Mike. What do you think? How should finance leaders, how do you as a finance leader explain the value of AI-enhanced driver models?
Mike Uvarov
The obvious benefits are speed and accuracy, but the new reporting and forecasting method should speak for itself. What we did, we ran our old traditional reporting and forecasting method in parallel with the new one. And yeah, after a few months, we see that which one is a winner. So yeah, this was our method and it works.
Olga Rudakova
Right. That's pretty convincing way of the transition of adoption. Back to you, Vignesh. What are you saying? How should finance leaders explain the value of AI-enhanced driver models?
Vignesh Dumonceau
I think I join also what Graham said earlier. For me, if I put it simplistically, AI remains a tool. We had different times, we had different enhancements in our tools. It's better than riding a horse than walking, and it's better having a car than riding a horse. So every time you get more capabilities, but then the question is always a cost-benefit trade-off, and then are you going to achieve the additional power through this implementation? And I think we will. I mean, I have a personal conviction that it will help, And it will help to enrich also the value activities that our teams provide. But we need to demonstrate that this is going to positively influence the leaders in their deliverables, in the results that they provide to the business. And that will come with time. Yes.
Olga Rudakova
Right. Right. Thank you for this outlook. And I would like to move on to Graham's presentation. So Graham Hunter will be speaking about practical adoption of AI in driver-based planning. So Graham, let me pass microphone to you. Let me leave the stage to you. That is yours.
Graham Hunter
Excellent. Thank you, Olga. It's important to understand that. I'm trying to advance the slide. There we are. It's important to understand that these are transformation projects. Even 10 years ago, without AI, driver-based planning involved a significant amount of corporate transformation, whether large or small. And AI introduces some new vectors into that, making it more complex. It's very easy for us to look at the right side of this, look at all of the new things that we can do. But countering that, are the other factors that have to be acknowledged as well and managed within not only your finance organization, but across the business partners that you need to work with to enable full driver-based planning. The enablement gap, I think, is probably underscored by the last survey as well in that there are rarely corporate resources around to help you understand what is possible with agentic compute and with AI in general. Having this as part of your plan is very important. Looking at some of the benefits that we can have as well, it's pretty obvious. I mean, accountability is a big part of driver-based planning and what AI can drive in your organization across the entire company. Of course, the business is going to say, well, how much change are you going to introduce with this? And that ends up being a political game. Transparency, it's often, especially with sales teams, do we have to give finance all of our data? As you go through and you start managing what data is appropriate to bring over into finance systems for prediction, that adds complexity. Who pays for the system integration? There will be time and money. Whose budget does that come out of? And then agility is what I think every finance team is really looking for in the end. And there's the even more politics introduced by that. So where to start? This is one of the biggest things that we get asked on a regular basis. It's not wise to start to boil the ocean, of course. Here are three quotes, an ancient Italian proverb. You know, the enemy of the best is the enemy of good is perfect. I think a lot of people have heard that. I found a quote recently from King Lear 400 years ago that says something very similar. And what's funny is I've actually been told to stop saying this. We don't know what we don't know. That is the truth in 2026. So where to start? So I have two recipes for success to propose. The first one is coming from the software development world and especially the world of startups. And when you're a new company and you're really trying to launch something, what a lot of people do is to define what is the minimum viable product. That's not the product that you're going to launch with your version one. It's not even a beta. It's what product can we assemble that we can put it in the hands of people in our target market and deliver measurable value. And then you iterate from there based on that basic feedback. And I would suggest that with driver-based planning, especially with AI added, this is a very valuable mindset to have as you approach a very large, potentially a very large scope. The second strategy we can have here is crawl, walk, run. And I've seen a few companies do this well. And they have PowerPoint templates that are very similar to what you're seeing now. Three columns and the crawl obviously is where you put the most detail. This is where you're trying to go in the next three to six months. What are your immediate objectives in front of you for the first stage? The walk column is purposely roughly half the size of that. This is where are we going next? And then run is on the right side and obviously one to two sentences. And it's important to have the cultural discipline there because inevitably there will be someone who wants to expand more on the run and will drag the box down and make it larger. And it's important to say, no, stop, run is one to two sentences, you're getting too far ahead and resize it back up. The other thing to note about this is that once you're in that 90-day time frame and you've done things, you start all over again. What's your next crawl? What's your next walk and what's your run? And this will adapt based on the lessons learned. This is a great way to get started quickly and not wait for the big perfect plan. Prioritization. This is a matter of perspective, and I'm sure everyone on the call will have a different way to prioritize. From a business perspective, there could be problems with your sales pipeline. You could see that commodity pricing, which can be very variable these days. That might be something you want to get control of first. From a finance perspective, there may be a finance process that you want to really zero in on. Or the human perspective of the three broad risks and the three broad benefits that we listed earlier. one or two of those might be most important for you as well. So try and think of this as you prioritize where to start. There are multiple angles that you can use for this. I have some examples here of three different levels, what you might aim for. One of the easiest that you could go for is just revenue projections and forecasting your end of year income statement. Use your existing GL summaries monthly summaries are fine. You don't need to load transactions in. And run a basic machine learning prediction using a linear regression model. Add some input variables. Hey, we think that revenues will increase 5%. We think that this OPEX category, we can control that and reduce it by 5%. And start just running some basic models and drive what you think your end of your income statement is going to be. I've run this a couple times just to start as a pilot, and funnily enough, I've actually had these come within 2 to 3% of what actually happens. You can't do this over two or three years. There's just not enough data to do accurate predictions, and you do need to shift into a more driver-based planning to get longer range or more accuracy than that. Walk, integrated revenue planning. This is another very logical second step when you're in the walk phase or in other words, you've gone through your first crawl and you're ready to take the next steps. This then becomes your crawl, of course. The CRM systems generally host a wealth of data for predicting revenue and you can start using that data. And then on top of that, working with your revenue operations teams usually, you can go into more advanced algorithms that will only help stage validation. A sales rep's forecasted something at an 80% probability of close. Does an AI algorithm, an agent, come in and actually validate that, yes, this is 80%, or hold on a second, you haven't identified the right buyer You haven't even given them a quote. That big deal is really a 60. And I think you can tell that inaccurate stage validation alone would give you a far more accurate forecast of your revenue. Add into that, so you have an end of year revenue forecast, add in what you did in the crawl stage, just your OPEX expenses, and you're going to drive an even more accurate and more up-to-date and more real-time end of year income statement. And then a reference architecture. I mean, this could take one to two years easily to even approach. It's not something that you're going to get out of the box. Again, it's a transformation, both of the company, the culture, and the technology. But in this, and I've had the pleasure of delivering a few of these systems for large companies, It's worth it. It's worth it to move to an area where you're pulling in external data, you have unified business units, finance is able to inject variables and do what if analysis on this, and your transactional systems are aligned. The benefits at the bottom, I'll read them because it's worth reinforcing. A predictable revenue stream, a predictable spend, there's no surprises jumping out at you from Coupa. planning agility, the ability to go and when something happens in your ecosystem, you feel in control. And really to underscore it all, finance can become strategic in these and not just the people at the end who report the numbers or run the annual budget. Thank you very much. I look forward to your questions.
Olga Rudakova
Thank you very much. And we actually look forward to your questions and answers. We have already so many questions in our chat. Thank you so much, Graham, for amazing examples and for introducing this framework, crawl, walk, run. It's obviously suitable not only for driver-based planning, but for any change you're making. And let's jump right into Q&A, because I really would like to answer at least some of the questions. If your question won't be answered, please wait for a written answer afterwards. And let's start. And I will start with you, Mike. I have a question that is specifically addressed to you and came right after your presentation. So it goes like that. What do you use as a measure for success and base for bonus if you don't have annual plan or annual operating plan?
Mike Uvarov
Measure of success. I think this is what Graham, I piggyback on those. Just set yourself this crawl targets in the beginning. that you feel really good with the quick wins. This is really important to that reinforcement of that to the project to move forward. When you see those first wins, even the small ones, something you cleaned up your structure of this and that, and then you plugged it in and it works. That makes that Success is not at the end of the project. So it needs to be this consecutive steps of successes. And then what was this, the bonus? The second part of the question, I didn't quite understand.
Olga Rudakova
Usually bonuses are connected to meeting annual plan targets. And so if you move away from that driver based plan, so how do you decide on bonuses?
Mike Uvarov
Yeah, so then the bonus structure may need to be updated. There are various schools of thoughts of how to do this, but maybe it's a different webinar about that.
Olga Rudakova
Right, Let's move on to you, Vignesh. I have a whole bunch of questions for you, but I picked quite practical ones related to driver-based planning. So do you use a limited number of drivers for the whole P&L, or do you have drivers for each P&L position?
Vignesh Dumonceau
That's a great question. Actually, I would say it's really a combination. my approach, but then every controller CFO is a bit different, but I try to have limited drivers, but I try to really make them selective and connect them and make sure that they are simple enough so people can understand and that there is a real connection. Now, drivers are a little bit like allocations. We can increase the number of drivers and increase the complexity, but I have a tendency to go more on a fewer number of drivers, but more powerful and connected.
Olga Rudakova
Yeah, exactly. That's always a trade-off between simplicity and transparency and slightly better accuracy but much higher complexity. And basically that's obviously up to particular controller to set it up in a good balance. Graham, another technical question for you. really try to select the most practical ones and how to. So how much analytical expertise does if FP&A realistically need to build and validate driver-based models when relationships aren't obvious? And to what extent can AI handle this versus requiring dedicated data science support? Well, in a simple words, can average if FP&A team build this model with AI without super expertise or we still need those data scientists?
Graham Hunter
I'm going to say a very unpopular opinion that data scientists don't know how to speak to finance and rarely do they understand how to speak to people in the business. I'm sorry if you're a data scientist joining, but I would encourage you to learn how to speak business terms and understand what finance does. I was, as a quick tangent, I was very lucky to start my career as a statistician. And then I saw Arbor Essbase in the 90s and I pivoted immediately into the enterprise performance management world where I spent my time. So I do have that background and that was something that I've seen again and again is that when you do engage data science, if you have a team at your company, You can expect long cycles that talk to them about crawl, walk, run is usually something they don't like. They want full design, full disclosure up front. I would suggest that, to get right to the heart of the question, I would suggest that you either train someone on your existing finance team in this, and there are so many courses out there on And you don't have to go into machine learning right away, just statistical algorithms. AWS and Azure and Google Cloud all have courses and basic certifications that people in finance should have no problem adopting or perhaps hire someone that you know isn't a data scientist, or if they are a data scientist that clearly understands the finance role. And then from there, you should be able to do, I mean, your crawl, Again, I'm sorry if you're a data scientist. The crawl probably shouldn't involve the data science team. It should be the finance team getting their feet wet with these new technologies. And then once you've gone through that first phase, you can have a more intelligent discussion with data science if you feel that that's the right time to bring them in. I hope that answers your question.
Olga Rudakova
I really hope so as well. And thank you for this answer. Well, that's another indication that finance needs slightly learn the language of data science. Data science needs to learn the language of the business and eventually we can all talk to each other. Thank you for that comment. And I have a practical question which I would like to address to all of you. And that is how to start. I will start with you, Vignesh, but let me read this question first. So where shall an FP&A team with no use of AI so far currently relying on EPM system and Excel should start to introduce AI? What part of FP&A would be easier to adopt AI, reporting, planning, modeling, analysis, and so on? So how to start with AI if you are not using it yet, but you have EPM and Excel? Vignesh, what do you think?
Vignesh Dumonceau
So that's a great question. And I think that's what a lot of FP&A teams in their day-to-day life go through. So I would say that, AI, look at your company policy. Sometimes you have an internal AI tool which has been developed because in FP&A we share confidential data in terms of business decisions. So you have to look a little bit at your data policy. If you have an internal AI tool, or you have a license that is available, then my first step would be more on the reporting side. So try to use your reporting and then you can, as simple as copy paste some of your Excels on the tool and start a dialogue and try to build a model. If you don't have that and you have to move to an external AI tool, please check with your company policy. Ask them if you can share the data of your internal policy there. And you could, with some context, probably do the same thing. That would be the crawl, as Graham said, the first step. Just to get yourself familiarized in terms of what would be the response of the tool and what it would give you. Then the next step would be try to learn a little bit about agents and agent TKI, and then try to build for yourself an agent that can do a forecast for you, for example, where you provide base data, you give him a model, and you ask him to give you, based on the driver, a sample forecast where you can start to have a dialogue. That would be the second level. And the third level is you have multiple agents, but that would be another step. But the first step would be just have the tool, put some base data inside, and started a discussion. That's how I would do it.
Olga Rudakova
Exactly. Just start with experimenting. Mike, let me move on to you. Well, Vignesh gave very elaborate answer, but if you have something to add, what would you be? How can we start implementing AI in FP&A?
Mike Uvarov
Yeah, if last year was a Wild West for AI, everyone or not, some people started experimenting and uploading data into GPT or Claude. As Vignesh said, companies caught up with their policies. And I would really suggest if you have an EPM tool, look at the latest releases. Chances are there is an AI element started to getting built into those tools.
Olga Rudakova
Right, actually starting with you already have exploring what you already have in your existing tools. Amazing. Graham, what do you think? How can we start implementing AI in FP&A?
Graham Hunter
I think that the most obvious gap at the start is prompting skills, how to write effective prompts to get what you need. That generally is the first barrier that people in finance have to effectively analyzing, effectively moving into the agentic world. I completely agree with what Vignesh is saying here. I think this is just a wrinkle on it. If you have access to M365 Copilot, It's not the best AI, but it's definitely safe and it's immediately connected into your corporate data. I'm kind of waiting for someone to come along with a much better business AI, but until then, M365 co-pilots well worth the $30 a month. Start a little working group, tiger team, whatever you'd like to call it, and identify your champions and make sure they know that it's important to help the people that are falling behind. There is a way in Copilot to share prompts. Microsoft conveniently hides it, find it, and teach everybody how to share prompts across their Copilots. And yeah, I mean, that's really what I would suggest is the first step is once you learn how to prompt and write bigger and longer and better prompts, the rest will start to flow. It's amazing how quickly people, once they pick up on it, how quickly it proceeds into the more advanced use cases. But without those basics, people will stall. They will sit there staring at their prompt window asking, What do I ask AI to do?
Olga Rudakova
Amazing. I'm so glad that, Graham, that you mentioned that, exactly. Prompt engineering then gets you to context engineering, to proper instructions, to skills, to plugins. So eventually you get all of that, but from a good, well-written prompt. Thank you so much for all your answers. We'll try to answer the remaining questions in writing. Just please be patient with us. Again, I would like to thank our sponsor for making this event possible, CCH Tagetik by Wolters Kluwer. I would like to let you know how you can connect with us with this FP&A Trends global community using various channels. So choose the one you use yourself most frequently. I would like to ask you to fill in our short survey at the end to help us improve. Your feedback is very, very important. And most importantly, let me thank our speakers, our panelists today for sharing your wisdom, for sharing your practical experience in case studies with us. Thank you for that. And big thank you to the audience for joining us live today. Have a good day, have a good evening, and see you at the next FP&A webinar. Goodbye.
Graham Hunter
Thank you very much. Thank you everyone.
