Name: From traditional CFO to AI CFO: AI roadmap and relevant use cases from practice
Uploaded: 2025-11-20T17:16:16.962Z
Duration: 1 h 6 min 7 s
Description: From traditional CFO to AI CFO: AI roadmap and relevant use cases from practice

Transcript for "From traditional CFO to AI CFO: AI roadmap and relevant use cases from practice": Hello, everybody. Give folks, just another minute or so to hop in, see if, we get some more attendees. Hello, everybody in the chat. Looks like we've got Berlin, London, Ireland, one from Ireland as well, South Africa. Good evening to you coming to us from South Africa. I'm gonna go ahead and share my screen so that we can get started here. And then sure there will be some more, question towards the end. Anytime you wanna just throw something in the q and a chat, go ahead and and, we'll have a slot at the end of the session. Alright. Kelly Nicolas, just making sure you can see my screen, making sure it's popping up for everybody. Awesome. Well, good afternoon everyone. Good morning wherever you are, calling in from, joining the webinar. Feel free to throw that in the chat. Always good to see where folks are, joining the webinar from. I'm, here in The States, so it's the morning here. We're running on coffee and ambition getting the day going. Thank you for joining us. We're delighted to see such strong interest in our session today, from traditional CFO to AI CFO, the roadmap and and relevant, use cases from, practices that we've seen, just from experience here at Agicap and also, from Nicolas experience as well. So I'm gonna start with introducing our speakers. So first, Right. you don't already know him. It's, Nicolas Boucher is one of the leading voices in AI for finance leaders, helping finance professionals understand and implement, AI in their daily work. So he also shares his expertise with, nearly a million followers on LinkedIn. So, I'm sure a couple of those followers are here, watching you, Nicolas, on the call. So look looking forward to, to your expertise today. We also have Kelly Roussel, our generative AI engineer here at Agicap. Kelly works within our lab team on multiple research projects as we, look to test out new AI features with our customer base and our product, just helping us bring more innovation into what AGICAP does. And then I'm Brandon Barnes. I'll be the host and moderator today, based here in the AGICAP office, in The, US here in Austin, Texas. So before we dive into, the rest of the session, just wanna go ahead and quickly, introduce AGICAP, before we go into the use cases in in, business and sharing some best practices, from both what Nicolas and Kelly have seen, in the space, and then we'll finish with a q and a towards the end of the call as well. If you have, any questions, during the session, feel free to throw it in the q and a. And at any time, if you want to, reach out to chat more after the call and get more tailored to your specific use case, feel free to just select the request an AGICAP demo button, and you can fill in your information to chat more, with us after the call. So some background on Agicap. We're a cash management and forecasting platform, platform, giving finance leaders real time visibility over their cash position, automating reporting, and enabling more confident decision making essentially when it comes to your cash and treasury management. So we've been around for about ten years now. Offices, all over the world, several in Europe headquartered in Lyon. And, like I mentioned earlier, I am dialing in from our office here in Austin, Texas. So customers all over the world, significant, funding and experience specific around cash management. And when you go and and look at the, reviews that folks add online for Agicap, it's clear we are the market leader when it comes to treasury management, especially for small, to mid market businesses. So, you'll you'll certainly see more in terms of how we leverage AI in the platform, and how you can potentially leverage some AI in your day to day as well. But the treasury management platform that we offer is essentially an an all in one to meet every cash challenge that you can think of, whether it's daily cash management, viewing your bank accounts and your expected accounts payable, accounts receivable from your ERP, forecasting out whether it's a thirteen week forecast, or a longer term cash forecast as well, as well as handling, some accounts payable processes internally and then accounts receivable analytics. We even, have a a little bit, functionality on bank reconciliation and journal posting as well, making sure we can match your bank transactions in real time with your invoices, from your accounting tool. And the idea is that we are just centralizing all data that's typically scattered between your accounting tools, Excel, individual, manual reports, and and bank exports, so that we can help teams save time, in just putting together a better cash forecast and having better visibility in, their liquidity. We've got experience working with a number of different customers, around the world. Just some few examples here, all the way even to, football clubs, like Watford, that leverage actually have to get a better view of their cash. But many partnerships on the technical side in terms of other accounting tools and integrations as well as private equity funds that, work with our customer base. Some other examples here just in terms of smaller franchisees and family offices as well. But in terms of the main topic today, wanted to get started with a quick question. How do you currently use AI in your finance? Do you use a common chatbot? Do you, have AI integrated into one of your finance tools? Do you have AI developed internally by your company, or you don't use AI yet at all? Go ahead and throw it, throw your answer in the poll, and we'll just kinda get a sense of where we're at today. Alright. Seeing some initial responses come in. Waiting for a few more. Say about a 100 so far. So we're about a quarter of of the folks that are on the call. Interesting. Yeah. Nicolas, what what do you think about that, Abby? Were were you expecting kind of the, the the heavy lean towards folks that are just using the the common chatbots out there. right now? But I didn't I'm impressed that still we have 30% of people that still don't use AI. Because when usually, I train finance teams, if you look one or two years ago, we were around this where two third use it, but one third don't use it. And, right now, I feel like 90% of people use it. Maybe not every day, maybe not really on finance use case, but we it's not a problem of using it any yes or no. Like, people use AI. Like, that's, now we we see that. What we will see today is how you can use it for finance. That's where a lot of people have difficulties, and it's maybe, here in the people that, say that they don't use it, maybe they don't use it specifically for finance. And, the goal today is to change that. Right? That's right. Yeah. I know you've got some, some good use cases, with generative AI. I know that's gonna be the the first topic that we're highlighting today. So, Nicolas, what, what best practices are emerging in this field, and and what are some traditional finance tasks that have become much more efficient thanks to just the adoption of AI more commonly with finance teams? Yeah. And I think we need to close the poll so I can share my screen if, yeah. Yep. Let me stop sharing there. Go ahead. So I want to start with a first use case. Imagine that you are a CFO, and you get again and again, like, for the tenth time this year, you get asked to do a scenario to build something in Excel, because either the investor want another overview or your CEO has another idea, or just basically because it's part of the process to almost every month redo a scenario. And this time, you have actually no time. You you plan something this evening with your wife or you have to take care of your kids, and you don't want to have to spend all evening on this on Excel. So let's see how we can use AI to, to work on this. So imagine so we're going to prompt. I'm a CFO of a manufacturing company. I need to prepare a five year strategic plan, and I'm going to give the assumptions. Then I'm going to give my actuals as a base, and I'm going to upload this in Chargegpt. You can do that in Copilot, in Gemini. What I advise is if you upload your data, you need to have a corporate license. So that's something really important to have in mind. And really important in the prompt, I am saying that we need to use formulas and to make it dynamic and to have an assumption sheet. So like this, we have a model that we can change. Also, really important so I'm uploading the file. Really important, I'm using here the thinking mode. A lot of people, when I train people, they don't even know about this thinking mode. And if you don't use this thinking mode, you have a risk that your output is really bad quality because thinking mode thinks about the task and spend much more time to make you, a model that works. A lot of time, if you try this, if you don't have the thinking mode on, the model will not work. And so it will take, like, around five minutes, which I cut it for us. And then I have a link which I can open, And this link is actually the excel file of the model that we created with the assumptions. Then we have one tab for the first scenario, so the base scenario. And you can see, Brandon, that when I click inside, I have formulas, which is amazing for finance for two reasons: First, I can audit it so I can make sure that the formula is the right one and makes sense. And second, it makes everything dynamic, and I can change after the assumptions, like we will see, and I can change my model. And everything is linked. And I have another one for expansion and for pessimistic. And AI is actually also much better than us. It documents everything, so we save also a lot of time. So imagine in five minutes, I have a model with formulas, and if I want to change one, assumption on The US, for example, let's see if the 11,000,000 revenue we had is also changing. Yes. It's changing. It's taking in in consideration the changes. So this was, like, in five minutes. But then you cannot go in a meeting and present this file. Like, that's not going to work. People are don't understand is tables are for finance. But when you want to go to present to your CEO, you need to create a presentation. You need to have a story behind. So we are going to continue the discussion in the same chat, and this time, I'm using a tool which almost nobody knows about. It's the canvas with a s. And canvas will allow us to build something for our presentation, which will be like a dynamic slide deck, where if my boss is asking me some questions, I can change the scenario in front of my boss, and I will show you how. So this is the prompt that I added inside. And here ChargeGPT in the canvas mode is writing code which is basically building this dynamic slide deck. And after a few minutes, I can click preview to see this slide deck, and Chargegbt build a scenario deck with assumptions with comparison of the revenue, of the gross margin, of the net income, comparing the different scenarios. And when I go to my meeting with my, CEO, and the CEO is asking, okay. I like your scenario, but I want to see what is the impact of changing it to 50% in 2030 Instead of waiting one week to present the result, I can present straight away And so this is really good for ad hoc basis, It's really good to prepare your meeting. What the limitation is is not connected to your data, and, you need to have professional license for this. But imagine how much time you save and also here, how much value you bring because in a meeting, you help to make the decisions happen faster. That's very cool. I I didn't even know that the Canvas feature could do that. Yeah. I learned something today. That is pretty cool. I mean, the generative aspect of of AI opens up enormous potential for finance teams to be able to adjust on the fly, like you were just saying, Nicolas, and and present a better story in those meetings. In terms of what we're doing at at Agicap, we're also integrating this, technology directly into the product. Kelly, could you share your screen and show us how the Agicap AI assistant works? Yes. Sure. So as, as Nicolas just said, the the problem with, using TEGPT is that it's not directly connected to your data. So you would have to copy paste or to import a file with all your data into TEGPT or any other, external bot. So in Azure gap, we've created, an integrated assistant, an AI assistant, to who you can ask any question about your data. So here, you see for the example, I asked it if I will have enough short term cash if I made, $20,000 payment in three days with my bank Bank of America account. And there, the the assistant will retrieve all my data on my Agicap account, so no need to give it the data. No security issue because your data are within Agicap, and no need to prompt engineer your demand to the assistant. It will be really natural. You can ask it in natural language, in any language, you'd like, and he will present you the the plan he has, the action it will take, and, answer you. So here, he tells me, my bank account would be negative if I do that. So I will tell him, I don't want my account to be negative and ask the assistant for advice. And so this assistant is a financial assistant. He it has been prompted to answer those question. And so he can give me advices of you should transfer funds between these banks and these bank accounts, etcetera. So this is, really, interesting for users that don't want to spend a lot of time prompting the assistant and eventually copy pasting the data out of their financial tools. Another usage we have of our AI assistance is to ask directly questions about how Agicap works. For example, imagine your treasury is in a is in a sick leave. So, you need to you need to make a payment on Agicap, but you don't remember how to do, then you can just ask your assistant, and your ask your assistant will guide you throughout all the steps to make your payment through so you're using Agicap. We have a super good support team, but sometimes, if you're it's, in the weekend or in the evening, it can be really useful to have a really quick answer to ask directly our assistance. Very nice. Thank you for sharing that, Kelly. And first part, of course, of this was more around generative AI, but another area where AI brings some real value is automating repetitive financial tasks. Processes like invoice entry, account reconciliation, regular reporting are pretty time consuming and error prone when when you're doing it manually. So, Nicolas, I guess I'll go back to you. What would you advise CFOs, who want to use AI for more of this automation use case? So I've, I've created something now in Copilot because, today, I want to show you that you can use TagPity, Copilot, and Gemini. All of them have kind of the same functionalities. And, imagine a team where every month or every week even, you could have, like, credit card statements, and you need first to clean the data. Then after once you have cleaned the data, you want to categorize the bank statement in order to, after book the bank statement into your accounting system. I know a lot of team, they do that manually. They will have somebody spending half a day on this every week or every month. And so I simulated this workflow in Copilot. So imagine this file. You can see that we have here a file where the columns are not really clean. You have information in the top that is useless. You have columns on the right that you don't need. So before you can use it, you need to delete columns. You need to delete lines. In the bottom, we have, a lot of tabs. Here, we have 30 tabs. Basically, before you can do something, you have to clean and consolidate. So I'm going into Copilot, and I'm going to attach here, something which is on my OneDrive, so the same file. And so I'm explaining, and that's important to be really clear, what is to do. So combine all of the tabs and, keep only b c d, and show me how you consolidated that. Like, that's really important. Ask Always AI. Show me how you did the work, and maybe check on your own work. That's a good way to save time on, auditing the work of AI. And so it's going to analyze, and we have here all of the work. And for I recommend for everybody, if you want to do this type of work, you can use also the analyst here. I didn't do it here just because, I wanted to show you that even in the free version, you could do something almost as good. But if you use the analyst or you use GPT five, the work that we do here will be even better. And so, it's it did now the consolidation, and I will open the file to see if the consolidation is done well. So I will now, have the file in the top, and I will click on it. And when I open it, what do we see? It didn't import the company and the cardholder. So right now, I don't even know from which company or cardholder is each line. So the work is not done properly. And that's often the case what you will have with AI is you try the first time, it doesn't work. And so what do we do when it's like this? We come back. We explain, that you didn't do what I wanted. It it's it's normal. Like, AI never works at 100% at the first time. Like, especially these, these tools because they are so open. They are they are like knife Swiss knife tools. You really need to know how to use them because they can do a lot of different task, but you need to be so good at prompting until you get the task you want. So you have this iterative process where now after three, four times, finally, now you have for each line which car loan name for which company. So I need, like, two, three iterations until I get the result I wanted. But it took me a bit less time than doing it manually. Then my next step is to categorize you can see so I wanted to categorize each of the transactions following these five categories with the account. And so, again, so it did, the work as a preview. Often, AI likes to do that. Like, here's a preview of how it looks like. And so after, I need to ask, okay, please download the file. So again, another step. So it takes a bit of time. And so now I have the file. Normally, they will arrive here. And if I open the file, they will arrive here Now after four or five iterations and I open the file, I have finally the file that I wanted, meaning we have each line, each company, calendar, and the categorization. But one thing where nobody wants to have is to go through this pane that I just did, where I spent twenty minutes back and forth until I got the file I wanted So the magic trick once you did it once and you are happy about the result is called the system prompt And look, everybody, look really carefully at my prompt. I'm going to ask sorry. I'm going to ask, here. Just yeah. So what will be the system prompt that will give me exactly the same output on the first go? And AI is going to detail for me a much better prompt, which next time if I want to do the work, I just have to copy and paste this, give my file, and then it will consolidate and, categorize everything. But if I really want to automate this, I can copy here the the prompt. So I'm going to copy this let's go a bit further so I copy the prompt and then I go to create agent on the left side I configure the agent, so I give a name, I give a description and then in the instruction below I am going to paste the system prompt that we just created together. So let's paste that. And also, let's not forget here to activate the fact that it can create Excel file because if not, it will not read in a create Excel file. And then if I try this So so I can go to the agent, and I can also share it with people from my team if I want. Now if I just upload the file and here, I don't prompt anymore. I just say do your work. Because it has behind the big system prompt, then it will do for us the consolidation. And then after the consolidation, it will categorize as well. So this is what I got after five minutes. I got the file which was consolidated and with the categorization. So this is how you can go from a process which is super manual to use AI and then to also automate within AI. But, again, it's something where, you know, you need to to know how to prompt, and it's still human triggered. If you have no human behind, nothing will happen. You need a human that will go inside and say, do the work and a human to verify that. Yeah. And and one question that I see in the chat, I believe you mentioned it earlier in terms of using the enterprise mode, but David asked about protecting your data when you upload sensitive information into ChattGPT or Gemini or Copilot. Yeah. What would you advise for folks, just to make sure that they're being secure? So, so we have a a part on this a bit later, but, principles is like any cloud tools, like any cloud solutions, you need to check that your vendor has the security standard that fits your business requirements. Most of the time, what you need is at least SOC two type two level of, control, and security, which Checkatpity has for business and enterprise, but not the other one. Like, anything below, it doesn't provide that, And which you also have for, Microsoft Copilot m three sixty five and for Gemini when you buy it with a business license with Workspace, Google Workspace, which, by the way, a lot of people don't know that. But when you buy Google Workspace, you have Gemini included, so you don't need to pay on top. So when you use always those product like Microsoft for your business or Google for your business and you use their AI enterprise service, then you have the same security environment than what you use for your emails, for your Google Doc, or your Word, or your Excel that are online. And one more question, just with the raw data. I see Melba asked, did the raw data that you were uploading in that last example, Nicolas, include categories? Or how did the AI know how to categorize when you were submitting that information? Yeah. So you have two proceed. It didn't include the categories, so AI is doing this. There is two ways. Either it does it deterministicly, meaning if there is a name cloud in the description, I will categorize it as a cloud solution, or you can change it to say it sends to the LLM, and the LLM has to read each line to understand if it's a cloud solution or travel expense or so we have these two ways, either deterministic or probabilistic. Awesome. Kelly, we'll turn it to you now. Could you show us how Agicap supports some of the automation, in terms of, you know, getting rid of those repetitive daily tasks? Yeah. Sure. So, AI will automate a lot of processes. It will accelerate things, and you will tend to to have a less vigilant eye on your data. So the first thing I wanted to show you is the categorization. And so like Nicolas, show you just before that, integrated in Agicap. So here, we have a group of transactions. So this group have been, chosen by AI. And in those transaction, the the AI, color the more relevant words. And by more relevant, I mean the words that are holding the more semantic meaning. And from those words, the AI will define a rule. So this rule is every time you have a transaction that is cash outflow for any bank account, And if the title contains office rent, I suggest those transactions should be categorized in, the category premises. So here, you can eventually, edit any part of the rule, and then you can, validate the the rule. This will categorize all of those transactions at once, but also create a rule for all the future transactions. So all the future transactions that will fit this rule, so having this, specific word in the title being cash outflow or inflow, will be set in this category. So it is a huge time saving for processes that used to be really long and tedious, on Agicap before, and that is really easy and really facilitated by a one time effort now. Then the second thing I wanted to show you, is the payment anomaly detection module. So this is something we have, on beta. And the idea here is to have, an AI that will keep an eye on your payments for you. So, by, anomaly, when we say payment anomaly detection, we mean errors, like misplacing a comma. We mean a fraud because this can happen too. And, eventually, for this is my example in the video, duplicate. Because imagine you have a miscommunication and you happen to make a payment twice to the same beneficiary of the approximately the same amount and with approximately the same level, this is not normal and will probably be a duplicate. So here, you see in the example, I make a payment of 1,000, GBP for my rent. And when I will try to validate my payments, my AI is watching me and is telling me, woah. You have recently paid the same amount to the same beneficiary. This could be a duplicate. It does not prevent me from doing the payment. It is just saying it's your choice, but I suggest you double check this payment. It might be an error of of yours or fraud or something else. So this, is really, it is no no longer generative AI. It's fundamental AI regarding the history of your payment. The AI is constricting a norm and a rage. And if a payment, moves away from this norm, it is considered a normal, and so it will warn you instantly. Cool. Thank you, Kelly. Well, the first two topics that we were digging into, generative AI automation, both very important and just helping to streamline daily processes of of preparing, just cash reports and and visibility into long term, financial statements as well. But, Nicolas, I guess the last piece would be putting a story behind it. You kind of alluded to it a little bit in the generative AI example of of highlighting a report that you could show in Canvas. But do you have some more examples of, how we can leverage AI for, more automation and deeper insights when it comes to presenting a story, reporting on the data, that you're collecting? Yeah. So imagine that you need tomorrow or to tonight to present to your board or to your business partners the results of this year to compare with, last year a bit. And, you are asked first to do the functionalities and then to present the to prepare the presentation. I'll show you how you can do that and also the right way to do it inside, a generative AI tool. And, I will use Gemini to show you, this time, Gemini on this. So imagine we have we are going to use this six step approach, which I see a lot of people, instead of giving a clear instruction, they are just telling to AI, analyze this. They and if you ask AI, analyze this, well, you will not get good quality of analysis because it's like sending a file to a junior and say, analyze this. They will not know really what to analyze, why you need it, what is the context, so don't do that. Instead, follow these five to six steps to these six steps, and I'm going to show you, right now the benefit of using these six steps. So imagine this file where we have a company with different product, different region. We have the revenue and the EBIT for each region and each product. I am in Gemini this time. By the way, Gemini just released yesterday Gemini point three or three point o. So, like, everything is always moving really fast. So I explained I am in FPNA, and I want you to analyze this like an FPN expert and follow this step. So the step one is always I always ask to make sure the data makes sense. I'm an x auditor. You always need to start with this. Then step two, show me the data I used to make sure AI used the right data. Step three, tell me the strategy to analyze the data. Step four, calculate. Step five, show me the visuals. And step six, write the commentaries. So I am using again, like, the reasoning model, so the pro. And then once, the reasoning model is done with all of the steps, we'll see the output. So step one, we say it's verifying that the data is correct. And so it checked and didn't find, any problem. So the first check is done. But sometimes I find really interesting errors or duplicates when I do that. Then the second part is to check that the data we have here is, the same one that we used. And, by the way, with Gemini, what is good, you can always export tables in Google Sheets. I don't know if people knew that, but you can export in Google Sheet. You can even now you can export presentation in Google Slides. That's also, like, new since one week. So I have here the tab. Then now I ask which will be the good strategy. So those are the three strategies for analysis. And finally, I get the calculation, which by the way you can always audit by auditing the Python code which is behind. And once I have all of this, I get, also the visuals. So because, yeah, it didn't give me the visuals. So it's always I show the truth with AI. You don't get 100% of the results straight away. You need sometimes to ask a bit more, but I get the visuals. So once you have done this, now you want to present this result, this analysis. And first, always make sure you understand the analysis, that you have audited the that the analysis is correct before give, showing that to anybody. It's like having a junior doing the work for you. You need to review. You need to own it. So how are we going now to create the commentaries and the reporting with the financial storytelling? So I'm going to ask first, now the analysis is done and I I am okay with the analysis I have reviewed. I'm asking to create the report or the emails that will describe what happened. And once I have this, I can also use Canvas here in Gemini. So you can see we talked twice about Canvas today, once in JGPT, once in Gemini. And when I use Canvas, it makes it a document I can edit manually. If I want here, I can change the word, I can ask Gemini to change some part of the paragraphs. But the best part here is the blue button up right. Because with the blue button up right, I can ask to create an infographic. And what it will do with the infographic, it will transform this in a financial storytelling dashboard, which is much easier to consume. And this is done thanks to HTML code. And once I have this HTML code, I'm going to copy the code. And we are going to see if I, I see what is the the code in a HTML page, how does it look like. So I just check the HTML code, and now I have my dashboard created, which I can either present in a meeting, send by email, or even just make available as a link for all of my team. And this is just so much nicer to consume for my business partners, for my boss. Even for me as a CFO or FP and A manager is as a supporting document, this helps me a lot. And this is also quite impressive when you are in a meeting and you present something like this. Very nice. Certainly tells a much better story than just, an Excel sheet that you would highlight in an executive review. I know, Kelly, many of our customers already rely on on what Agicap has in terms of dashboarding and reporting, but we've seen a growing demand for faster AI powered insights, especially when there are, custom questions that get thrown out. They wanna see something very specific in a report. Can you show us how we're approaching that? Yeah. Sure. So this is our very last AI feature. It is still, in beta for the moment, so not available to everyone, but we hope to release it really soon, and it is about, dashboards. So as you said, we already have some dashboards, but now we will have some new AI dashboards. So here I'm on Agicap. I've created an AI dashboards. And now you see, in addition to the traditional chart counter table and notes you could create in the in the past, you can now create a new type of items that are smart reports. So those smart reports are presented like that. It's just a chat. You can ask in any language, just straight away without having to give, your your data or anything. So here, I asked I would like to visualize the evolution of my cash outflow over the past three months. So it's really general, question, really general request. But, I want to have a first, suggestion from the AI. So here you see the first thing the assistant does is it calls tools to collect my data. So here, for example, it asks for my categories my outflow categories, and then it asked for my cash flows over the past three months on every bank account. And it suggest me something. It's, an histogram, a blue histogram. Well, I have a base to start working on, So I will ask it to add colors, to add my inflows, to set the bars negative, etcetera. So it will collect the data it misses, and it will continue iterating on the graph. So here, I have something that looks a little bit better. I will finally add the blue line to represent the cash balance, and I think my dashboard will be complete. So this feature, I think, is super powerful for people coming, newcomers to Agicap that already have an existing, an exist an existing report they used to and who wants to get it on a dig up. So I just want to, drag your attention to this little open details button because it is really important to me. It is, the content we are still working on. The idea is to give our users the most possible transparency about what is happening under the hood, during the code the code agent is writing. So here, those are logs the code agent itself is writing during the code execution so that you ensure the period is good, the data is good, and it it understand correctly your request because, it has to understand correctly your your request in the order to satisfy you. And so then it is, a classic graph at any order. And what's magical is that it is also possible to create table, with this tool. So, really, the possibilities are almost infinite. That's what I love with this feature is that, it offer a full freedom regarding the form of the graph, the format of the tables, but also a full possibility of calculations. So no more needs to prepare your KPI before going to create your graphs. You can directly ask for any calculation you want to the assistant, and it will create them directly in his code and and display those, on the graphs and table, in Algae Gap. So really super, super cool feature that is coming really soon, in our product. Very nice, Kelly. Yeah. That'll be exciting once that, gets out of beta and is fully fully rolled out. I guess as we approach the end of the webinar, we went through the three main topics of generative AI, leveraging more automation, for finance teams, and then, of course, more detailed reporting and dashboards like we highlighted both with, Nicolas, in, Gemini and Kelly in Agicap. Kelly, I guess I'll go back to you real quick. Just as an AI engineer here at Agicap working daily to bring more AI features and functionality into, cash flow management software, where do you think AI is going to play the biggest difference moving forward for finance professionals and and what are we really looking to continue to implement as we grow as a product. And then we'll go to Nicolas for more of a big picture view after this. So I'll show you first, what we have already in Agicap. So those are our our different modules. So I show you the categorization. I show you the anomaly detection. We also have, some, OCR systems to help collecting data, easily, for example, for invoices. We also have some generation contents, so for example, to automate for an email. We are also working on, scenario generation, etcetera etcetera. So we already have a lot of scope covered by AI, but we are still looking for new, areas. So here is what is on our road map. The idea is to keep working on automating and accelerating the the workflows where you, experts, does not do not bring your value, your expect expert value, but are wasting time. So for example, we want to extend our OCR capabilities to simplify the import of documents, such as debts contracts because we are doing it a lot for now for, received invoices, but not much for contracts. So it is on our roadmap pretty soon. We also want to help you with your forecast because this is something our user have asked a lot. So for example, we will start by working on detecting recurring transactions in the in your history to simplify the creation as recurring, expected the transactions. And finally, you don't see it on Agicap, on your side, but it's working in the background. We are using a lot AI to improve our data ingestion processes and facilitate connections with external sources like ERPs or any other website. So this is a an a really important part in Agicap too. Awesome. Thank you, Kelly. And, Nicolas, I mean, so many people rely on you for your insights, in terms of what's cutting edge with AI for our finance team. So we've got a lot of CFOs on the call, VPs of finance, even folks that are entry level, that are just getting their career started on finance teams. What's a practical practical step that they can take today to better deploy AI in, their their work and and their team's work as well. Yep. So I will make it really practical because I think a lot of consulting companies want you to create, like, another digital transformation project where it lasts months and months and months until you get ready somewhere even if and they just want to charge a lot of hours. So here, we want to be practical and that you have straightaway benefits. So let me show you the road map which I share with the CFOs in the AI finance AI finance club, which I show also when I I get approached by any CFO even if they are from big companies. So you need first to start because everybody today was asking, oh, how can I upload my data in Chargebee, Gemini, Copilot? You need to get to your team and not just two or three manager. Everybody needs to get a corporate license. And don't let people use free AI tools. Like, today, I I talk about it in my newsletter. You need to give them a license, and the choice is actually really easy. If you use Microsoft everywhere, well, stay with Microsoft and go with Copilot because you trust Microsoft, and Copilot is using exactly the same security environment. If you are more a Google shop where you use Google everywhere, use Gemini. It's any way included in your Google Workspace. If you are small, like, you are a fractional CFO or you are a really small team, then you can use ChargeGPT. But ChargeGPT, you have to connect yourself, ChargeGPT, to your own, OneDrive or Google Drive if you want that it has some understanding about your company knowledge. So most of the company I work with, they go with Copilot or Gemini, and that's a good start because after finally people have one tool, get really good at it, and it's connected to their working file. The second part, you saw today that I showed some example of how you can automate your processes. So make sure that you get AI to help you automate your processes so you get benefits straight away. And you can let AI write automation script like VBA, Google script, Python script. It's really not rocket science anymore, and that's a good way to have low hanging fruits and quick wins. So that's, like, what you should do. Then people are asking me what is the first process they should use AI for. If it's not the case, normally, like, since ten years, the payables have been everywhere changed where they use AI to read PDFs and classify the PDFs and book them. But if it's still not the case, make sure you have an AI tool for your payables. Like, it's not normal if today you still have a human looking at an invoice and booking in your ERP manually. Like, that's your first step. After that, like, after this, this first phase, you can look at additional AI tools. Kelly, showed, for example, for treasury that you can use something like HD cap if your pro priority is more maybe on FP and A or on account receivables or payables. You have a lot of AI tools right now, but don't jump on an AI tool if you didn't, look at your use case first. Like, if, for example, you want to automate your revenue recognition using AI, it makes sense if you have one or two person busy with this every month. If it's only two hours of work, don't buy a tool for this. But yeah. So right now and it's moving really fast. But look at this. Look at your constraints. And what is also, something happening is also what is happening with Agicap is a lot of tools that were already good at one process. So for Agicap Treasury, they are including AI in the tools. So you have, like, legacy tools with AI, and that's also something to look at. Is the tool that you are already using, are they also getting AI? And then finally, you saw how my, my process earlier to categorize, I used custom agents. So that was a corporate agent, but in Chargegbt, you have custom GPTs. In Gemini, you have gems. You need to use that. I I train, like, more than 10,000 people on this on AI. And people that are really good at it, they don't even use this. They are really good at prompting. They are really good at, using all of the functionalities, but they almost never use custom GPTs, corporate agents, or Gemini gems. I use that, like, 90% of the time because I created some that know everything about me. And like this, when I prompt them and I use them, they will know the context and will answer 10 times better. And you saw earlier how I was really quickly able to create a custom coupon agent that could replicate the work that I needed half an hour to do for the first time. So don't sleep on that. This is the functionality that, after you, you need to implement. And if you do this already, these five steps, this can be done in ninety days. You don't need a six months transformation plan. Like, you can be super pragmatic and have results straight away with this. And after, you can really be more ambitious and maybe look at, using AI for forecasting. But then it's more like a a bigger transformation because it's a culture change. It means, like, humans are not the first one to forecast. You let first AI to forecast. And that's companies like Coca Cola and Microsoft, they have done that, but you need a bit more, culture change before looking at that. Thank you for that outline, Nicolas. Really appreciate your insight today of just what you've seen from working with so many CFOs in the space and and how they're leveraging AI right now. We have some questions in the chat. If you have any more, throw them in, and we'll use the last few minutes. We even have a couple minutes to go over as well to to answer some additional questions. So wanna make sure that we can, answer any questions that, anyone else has not yet put in. But before we do, just to give folks more time to throw in questions in the chat, if you are interested in learning more about how Agicap can help with centralizing all of your banking data, your transactions from your bank with your expected transactions from your AR and AP in your accounting tool, whether it's NetSuite or Sage or QuickBooks or an industry specific tool as well, that's how we can help provide more visibility. And as I showed earlier, a big reason, why we are the number one cash management and treasury management platform in the market for mid market and small medium businesses is due to how we are leveraging AI in the platform more and more like Kelly showed today. So I put some links in the chat depending on where you are joining from. If you want to request a quick call, certainly no pressure. We can go over the platform in more detail and talk through your current cash management process and visibility into your bank accounts, in a a quick twenty, thirty minute call and then, see if it makes sense to continue the conversation. So, to answer some of the remaining questions that we have in the last few minutes, let's go into the first topics here. Nicolas, I'll throw these to you. Looks like two can combine. You mentioned that AI does not always work the first time and you need to double check. How can you prevent that? How do you prevent, hallucin is it, hallucinadjacent I I. can't say it right now, but you. you know what I'm saying. Hallucinations. There we go. Alistairation. It's good. It's the same word in French. That's why it's easy for me. So, so really important is to be aware that it's possible to have hallucination, and that comes often. And why? The second thing is why? Because generative AI doesn't calculate generative AI generates. And it generates in a probabilistic approach, meaning, like, sometimes it will generate, left, sometimes right, sometimes black, sometimes white. So it's really uncomfortable in finance to know that if you ask a question, you never get this exactly the same answer, especially when us, we rely on accuracy. If you do a calculation, it cannot be only right 99%. It has to be right 100%. So to change this, you need to understand that AI is really good at generating, so generative AI to generate. And so you can let AI generate a formula like we did at the beginning of this, webinar where AI generated an Excel model where you can audit the formulas in in Excel, and you can check that if your input is correct and your formula is correct, your output is correct. Another way is, instead of Excel formulas, which we all understand, is you let AI generate a script either in Python or other, languages. It's it feels a bit more complex, but it's not that hard. Like, I I never called it in my life, and I use now AI to do a lot of automations. And for example, the script that says consolidate this file, the all of these tabs into one is actually, like, five to 10 lines. And you can ask AI to explain you how it works and to audit the work and to make sure the script works. And when it works, it always work the same way. And so it's really important for everybody to understand, rely on AI to generate those calculations or the script, but not to generate something where you cannot audit. And that's really good to to bear that in mind. And on top, if every month you want to productize this work, you cannot always go back to AI. What you do is you ask, show me how to do this in a productive way, in a productized environment, and it will give you, the code. It will give you the steps. It will give you the formulas. And like this, you can have it, and you don't need AI anymore in the future. Awesome. And forgot to mention for folks, if you do have to leave here at the top of the hour, you can always ask any questions, after the call via email, but we are good to stay for a few more minutes to answer some of these questions. I'll take I see two that I can take from the Agicap side. One, Fiona asked, can you include approved purchase orders in cash management? And essentially, Fiona, what AGICAP pulls in would be your expected payables that we would see from your accounting tool, from your ERP. So, that would be projected out. So let's say we have a large thousand dollar payment that needs to go out on the twenty second. We'd be able to forecast that out and see how it impacts your cash flow as well as your bank balances on that day. So that would be part of the process. Even if you haven't actually sent the payment off yet, we can still project, when it will be leaving. So you could potentially delay that, payment, pay it earlier, if it helps with your cash flow. And then Edward asked a question, looking to change from NetSuite to Dynamics Business Central. Wondering if it makes sense to embed Agicap from day one with Dynamics or implement, Dynamics first before looking at Agicap. And, Edward, I I guess it really comes down to what the biggest priority is that you're looking to solve, like, what what NetSuite isn't giving you that you're looking to move to Dynamics for because Agicap, of course, is not going to replace some, true accounting functionality. It's it's strictly a view into your cash, position, your cash forecasting as well. So if you're looking to have better visibility into your bank account balances and project out, like, a thirteen week cash forecast, for example, without using Excel, Agicap can help you get that up and running in anywhere from one to three months. So it's definitely a much quicker implementation than an ERP. So if you wanted to do it first, more of a quick win, go ahead and get that visibility into cache. Whereas if you wait until after implementing Dynamics, that's, I would assume, going to be a much lengthier implementation. So ultimately up to you on, kind of the timing and what you feel is the, biggest priority improving what you have in terms of more ERP and accounting versus getting visibility into, the cash. Nicolas, I will toss another one to you. Someone, Ahmed asked, can we use GPT to do what you did with Gemini earlier? And, I guess, even beyond the chat GPT, Waldemar asked, what about Grok? Like like, all of these examples that you used, could they be replicated with other, LLMs as well? Yeah. So try GPT, if you use Canvas option, you can create this, like the dashboard. That's Gemini is kind of like does it nicer, but, you can you can do it with JGPT. I I saw people asking about cloud, about grok. You need to understand that it's like cars. So you will always have, like, a fancy car that has a nice new option. But, ultimately, there are, like, some cars that are good for a lot of use cases. And if you work for a company, you cannot and you should not let your employees sometimes use your company data with Grok or Cloud or Treddypt or Gemini or Copilot. You should have only one tool. Because if you have many tools, you have analysis paralysis, meaning people will not do anything because they don't know which tool is the best. And it's better you tell them use this tool. This tool is on top connected with our work environment, and then you get really good at using it. And that's why I never talk about grok because grok is never used in a professional environment. It's like people for themselves, for their personal use, they will use it. Same for cloud. Cloud is really good for people that are alone, that write, that maybe for developers that are alone. But in big companies, I never heard one company using or even small and medium size, I never heard one company saying we gave everybody a cloud license. So that's why I don't talk about it because here, out of the 1,000, plus people that wanted to come in this webinar, I don't think one person in their company, got a a cloud license. But I'm sure we have a lot of people with Copilot, Gemini, and some with, JGPT licenses. And one additional follow-up, that I see in the chat, is you said you use thinking mode in chat GPT, but there's also an an agent mode. Could you briefly just touch on the difference between those for folks to better Yeah. So, I can show you quickly. So you have? in ChargeGPT, you have all of this mode. You have the auto mode will basically route your question to either instant or thinking. Instant is really good each time you want to write something fast. You want to have a quick answer. This is basically like the old GPT that we are all all used where you ask something, it goes fast and answer to you, which for a lot of use cases is okay. But each time you think about big problem, you want calculation, you want analysis, you should use thinking. It takes from 20 to ten minutes depending on the complexity of what you ask. But I prefer to take this, grab a coffee, come back, and that the work is really good rather than having to spend and ask and ask questions until I get to what I want. And then the agent mode is here. Agent mode. I just use it today to, to look for some like, if I could buy a company or something like this to find companies. I don't know if it's going to show it. But, basically, what it did, the agent mode, it went online to look at a lot of, web pages to find for details for companies. And that's the when you use agent mode, it can open kind of a new computer. Okay. It's not open. Yeah. So it what it did here for 60 minutes, it did this here. It went and looked for companies that would be interesting for me to buy. And if I will have to do that myself, because it found, like, 50 companies or that I could, go I have to go online and search. This is a good way to save time and to let AI work for you. And it does that as well with Excel. You can give, like, PDF and say, oh, based on this PDF, can you fill up up this Excel file? And it will do it also for you. Awesome. Thank you, Nicolas. Looks like trying to see if there's maybe one more question that we could answer. I see Kelly Roussel asked about asking in Agicap, in any language, like, in when using the AI tool, what limitations do we have from a language input perspective? Our AI assistants are using OpenAI models, which whom we have a a a contract to ensure that all the data we send and all the data we generate, stay as you get is the proprietary of those data. So that's the way we secure, data privacy. But, this means we completely have, the same capabilities as, the last OpenAI models. So, we're really up to date. And if you can talk a language with, Agicap, you can talk, with our assistant in this language. Awesome. Well, folks, don't see any more questions piling in, to the chat. So really appreciate everyone's time, and for all the questions and participation during the webinar today. Kelly, thank you for digging into the Agicap product and how we're implementing more AI. And, Nicolas, we really appreciate your perspective on what you're seeing in the market from working with so many, CFOs, around the world. So thanks everyone for attending today. A recording will be sent off. And if you have any more questions, feel free to reach out to, Agicap directly and happy to go through any additional use cases. But everyone have a great rest of your Thursday. Bye, Thank. you..