Welcome to Count Me In. In this exciting episode, we're joined by Nicholas Boucher, founder of the AI Finance Club and an expert in integrating artificial intelligence into finance. We'll dive into how AI is reshaping financial services, from enhancing client experience to improving risk management. Nicholas offers insider tips on how professionals can upskill and leverage AI to stay competitive. Don't miss this opportunity to gain valuable insights into the future of finance with AI.

Let's get started with the conversation with Nicholas right here on Count Me In.

Well, Nicholas, I'm so excited to have you on the Count Me In podcast as we're gonna talk about AI and it's a connection to finance. And I figured we could start off the conversation by talking a little bit about what you've seen, how AI is being leveraged to innovate in the finance sector, and what impact has it had on the services and the the customers that you work with?

Yeah. Thank you, Adam. It's a great question because everybody wants to know, what they don't know. Everybody wants to know, I was giving a training today. Are other finance team more in advance than my team or not?

And what is interesting, in each of the team, the level of understanding of AI is already different. So it's first, to respond to this question, I would say it would be really good that in each of the team, people get the same level of understanding.

Mhmm.

And for this people can, talk about it, can get classes. But, I will say the first thing that changed since 1 year is the fact that now AI is democratized because we got the chance to talk to AI with our own words using a chatbot. And because we can now talk to AI with our own word, and then AI with the use of this chat can translate that, talk to his model, and then reply us back. We everybody knows now that we can do something with it. Now because of this, you can get a lot of knowledge that you didn't get before.

You can get a lot of help on everything we do that is, I will say, white collar work. So everything about writing, everything about also solving problems, you know, like having a framework to solve problems. So this is what you can do today without no implementation of tools. You can get AI to help you do your work better. Then the I will say the the second level, and then we can talk about it after, it's when you are going ready to implement AI also in the current tool you are using.

ERP, your BI tool, also your data. So how you can make AI have access to your data to get answers and insights from your data. That's the second level. And there there's a lot to say about it, but not a lot of companies are, I will say, so far ahead ahead.

Yeah. A lot of companies aren't far ahead. Like, I feel like everybody's dipping their toe into the water, so to speak, and they're they're playing with it, but not everybody's quite there yet because there's a lot of predictive capabilities. There are a lot there's a lot of things that you can do with AI as you mentioned already with the data and stuff like that. And how, you know, how can AI be used to kind of help with, like, things like risk management and stuff like that?

Because I feel like, you know, when you're when you're introducing this new machine, this new learning tool into your system with your data and stuff like that, there's a whole another risk component that's involved there.

Yeah. So if we separate, I will say, like, how it can help with risk management and what are the risks? That, for me, that those are 2 questions.

Of course.

For risk management, what you can do is AI and not only generative AI, so not just producing text, but let's say, machine learning is really good at processing a lot of data and using this power to identify outliers, trends where a human cannot see that because it's too much for the human. So if you take human that can help the model and explain what is a risk. So if we talk about KYC or fraud in banking, for example, And this is already exist since long, long time is how you analyze a lot of amount of data and you notify, some outliers.

Mhmm.

And then from these outliers, do an assessment on what is the risk. Is it like is there a 99% risk that is a fraud, or is it the 10%? And out of that, either you will let the judgment to the human. So you will simplify the work that the human only spend time on everything above 90% of risk rather than doing all of the processing of the data to know, k. If I have a lot of data, then I need to process it.

And then finally, after the processing, I will only look at what are my outliers. So all of this part, you can delegate to machine learning. And also with machine learning, what you can do, if you don't know the rules about what is a problem in my company, you can just feed all of the transaction where you had problems. And the the model will recognize patterns. So, oh, like, each time we paid twice is when the supplier name was really similar, but not exactly the same, where it was a receipt.

And then, you know, like, from a restaurant, you can get 2 types of receipt. So you will notify, like, a lot of patterns where if they are combined, you will get a high chance that there is a risk. And that's where Mhmm. Us as human, we might not know that or we might not identify that because we cannot process as much. And then, like, on the second question on the risks, the we a lot of risk is that for a lot of people, it becomes a black box.

Mhmm. And so if it's a black box, then you cannot delegate the responsibility to a machine if you don't understand it. Like, you will not delegate to a partner, to, a provider some risk if you don't understand how they work. So this is where you need to understand how the model works, And we all need to upscale ourselves on the capabilities, the models. And I see a lot of people are learning really fast and they understand what is possible to do and what not what is not possible to do.

They really are. And I feel like you're either in the boat of I'm jumping in with both feet and trying this out a lot, and I'm upskilling to be better. Or you're like, I don't wanna touch this at all. There's too many risks involved. I don't want any of my personal data and all those things.

How do you kinda bridge that gap? Because I feel like people are in either boat, and they're either not touching it at all or they're jumping with both feet. How do you kinda bridge the gap between the people to make sure that your whole team or the people that you're working with are really getting up to speed?

That's a good question. What I usually use is an analogy with Excel. Mhmm. Imagine maybe, like, when Excel arrived, it arrived the same day I was born almost. So he arrived like 39 years ago, in 85.

I'm sure a lot of our content were skeptical if when you use the formula sum above and you take, like, 100 sale, I'm sure a lot of our content were doing with their own calculator by themselves verifying, well, is Excel really calculating the 100 sales like it should do? And we know that ex actually, after the time that Excel can do calculation of 1 100 sales, and you you will always get a correct answer. But now, Adam, let's imagine you go to your boss and, you say, okay. That's the one the sum of the one other lines. And your boss look at you and say, that's not correct.

And then are you going to say, oh, but it's not me. It's Excel, which is wrong? No. Maybe you did the mistake that in your formula, you forgot one cell.

Mhmm.

Maybe there was a hidden line because we all know this this program with Excel that some hidden lines can can really make a problem in our, calculations. And, actually, Excel was not prime. Your utilization of Excel was the problem. And your boss will not blame Excel, will not also allow you to blame Excel. We'll tell you you are responsible for your work.

So then you have 2 paths. Are you going to still work with paper and do calculations yourself, and then all of your colleagues are going to use Excel? Or are you going to use Excel, but carefully always verifying that the content makes sense, The formula makes sense. Does some make sense? So I will suggest that you choose the the second pay, path.

And with AI, it's the same. Mhmm. You need to be aware of how the tool works. You need to verify, and you need to own it. And there is a lot of way that you can use it without having to give any personal data, any confidential data.

And if you want to work with confidential data, then there are also more and more tools that allow that. Like, I'm not affiliated with Microsoft, but they have Azure where there, it's a cloud solution, and it works like all of the cloud tool you use in, for work. And if your company signed with them, you know, you can, put data inside. If you have the agreement from your company and a clear understanding from the company of what is possible to do or not.

Yeah. Yeah. I feel like it's it's really understanding how to utilize the tools. I really love your analogy that you made with the Excel because, you know, if you forgot to sell or if there's something like that, you're never gonna blame you're never gonna blame your scale. It's gonna become on you.

So it's up to you to kind of get to that place. So upskilling is very important Yep. As we move forward in this in this new this new realm of AI, because machine learning has been around for a long time. This this technology has been around, but it's now more accessible than ever.

Exactly.

And I think that's what's really scaring everybody is that it's accessible to everybody.

Exactly. And, to continue the the analogy, for me, there will be 3 types of people that are using or not using AI. There will be first the people who don't use it, and then they will have a problem in terms of competitive against people who use it. Mhmm. Then we have the 2nd type of people.

They will use it, but they will be really lazy and use it the wrong way. And we'll just do copy and paste and not review. First, there are this could be replaced really soon by agent that can do this connection between, the AI and your own tool. And so gone, you put yourself so much at risk. If you don't review your, the work.

And then there will be the 3rd type of people, the people that understand what you can do with it, that will use their knowledge of the company, their industry, the topic, if you are, if you know well, accounting, or if you are a really good FP and A manager, well, you know where to go if you want to analyze the revenue evolution of your company, and you will get AI to help you to be faster and better at doing it. And then in comparison with everybody else, you will provide more value. You will do your work faster, and you'll be ahead against all of the others who don't use it or use it the wrong way.

Yeah. So I think one thing that we we if we're talking about AI, we have to kind of talk about the whole personalization aspect, that ability to personalize things for your customers, for customer experiences, and and really and really showing that success at the customer facing. Because we've been really discussing how you use it on your own tools and stuff like that. But when we're talking about, you know, personalization, you know, for working with tools for your customers, you know, what are some of the best ways that you've seen people that you've worked with in your organization and how your clients have kind of used AI to help personalize their customer experiences in in a in a good way?

So if you are, are you talking about the, what we can talk about the, if we are talking about finance teams, their own customer are more the, I will say the, the internal customers.

Sure.

And I have a good example of OpenAI. I talk with the manager of OpenAI. We give a webinar together. And, so this finance manager explained me how much work they have because the company is growing so fast. They are hiring so many people.

And like in a lot of companies, people go to finance and they have a lot of questions. What can I expense? How do I book this invoice? All of the questions you can imagine that can come to finance. And so what they did, because they cannot cope with the work with so much demand, they use technology and they use their own technology to create a chatbot.

And this chatbot is using the finance policies to reply to all of their internal customers.

Mhmm.

And they are kind of the first, I will say, the first shield to to answer questions. And they're also available 24 hour, 7 days a week. So it helps also the internal customer to get the response really fast and to treat most of the questions. And if it it's not enough, then people know they can still talk to humans. But like this, both parties, the finance department and the other departments, they get value from it.

And this is something you see everywhere when you are talking about customer support and sales. AI knows how to answer based on some knowledge that you fed to it, and that's where you can gain a lot of value. So I would say this is one example. Second one, you have also in insurance or in banking, all of the credit scoring could be assisted with AI to make it faster as well. Mhmm.

Those are, and it's not new. This is assisted. And, so like this, you don't need a team of, like, credit scores. You have, the machine who is doing most of the work. And, like these clients can get their credit scoring much faster.

I think one thing that I've found as you were talking about that, the importance of feeding in the good data into your whether it's a chatbot, whether it's something that's analyzing and how important it is to have good data as opposed to, you know, that old the old adage garbage in, garbage out. You know, you need to make sure that you're feeding the proper data into your tool, whether it's for a chatbot to give answers to your customers or it's it's something assessing your risk. You have to make sure all your data is right. And so how important it is to to get your data in order before you start implementing these AI tools?

Yeah. So there is so the the quality of data is becoming even more important if you want to have the best insights. Yeah. But the good news is, it's becoming easier to do it. Because if you, for example, if you use a combination of chatGPT and Python, you can really quickly, you go into chat GPT and you explain what is your problem.

Oh, I have a list of 100 line 100,000 lines. And sometimes the I have duplicate for my supplier name or my supplier is twice there or 3 times or 4 times because we have history. And so you can ask, how can I solve that? And if you are not good at Excel and you don't know where to start, you can ask Jigpt, how can I do it in Excel? If you have a lot of data, you can even ask, how can I do that in Python?

Because it's faster and you can process more. And then it will give you data science techniques to clean the data really fast. It will tell you, okay, this is how you identify duplicate. This is how you can code a code that will, remove all of the duplicates, put them in a second sheet to allow you to choose which one you should choose, which line. And really, really fast, like, in a matter of 15 minutes, you can get a database clean on one topic.

You cannot, like, clean all of these topics, but you can, like, if your program is have duplicate in suppliers, you can get a code that will notify all of these duplicates and then we'll let you okay. Suggestions as well on which one to keep and which one to remove. And like this, us as human, we only spend time on reviewing and not on doing the work like manually and, trying to document it. No. Everything is already there for us.

And once, once you have done that, then you can get clean data and work much faster.

Yeah. Yeah. I like that. Cause it's, it's it's interesting because in in back before we had all access to all these technologies, it was a lot more difficult to clean up your data. But there's an example you gave, you're using the AI to help you have the tools because it it really shows how important these AI tools are in your toolbox for your daily job.

Yeah. And a lot of people will tell. Yeah, but how can I do it? You should not upload your confidential data in a tool where it's open. Totally agree.

And I also always say that were in my trainings. Yeah. But, like the method that I teach is you just need to explain anonymously saying, okay, look, I have a column showing supplier names and you just show some examples of lines that are fake. And then you, you show how your data is structured and you explain what is your problem. And then you ask in my own environment, in Excel or in Python, or can I do it to do the work within my own environment?

And, we teach that in one and a half hour course for people who never use Python. At the end, they know how to use Python for this type of topics. And it's crazy to see finance people who never thought they will use Python in their life Now, like just being really excited to see everything they can do between, automation, visualizations, deep analysis, because you can use all, formulas or that are not in Excel. You can use them in Python because Python is much more, complete in terms and more deep in terms of, mathematical calculations.

Mhmm. I like that. And, also, you don't also have to become a complete expert in Python. Exactly, man. You can know just aspects of it to help better your job.

And I think that's something that can be overwhelming for people as they learn new technologies. They're like, oh, I have to become a complete expert. No. You just No. Just know just enough to do what you need to do.

You don't have to know the whole thing. I I think that that's more freeing for people when they realize, oh, I don't have to learn everything.

The the what you need to learn is how to ask the right question.

Yes.

And we are always talking with my wife about that all the time. We see new topics, new, new problems. When you know how to ask the right questions, you move faster. You, adopt changes faster. You understand topics faster and whatever you have in front of you, if you know how to ask the right questions, you can move.

So maybe you could talk a little

bit more about that. How do you start asking the right questions? Maybe are there some tips that you can give for people asking the right

questions? So first, you need to understand what in front of you, how do they deal with questions?

Mhmm.

So usually, in front of a computer was really binary. You needed to enter exactly what the computer expected. So if you are an SAP, you need to know in all of the, the user interface, you need to know exactly what is the, the code you need to enter and what which filter to get data. Now it became with Google, for example, really interesting that you could just put keywords and get a feedback on what is on internet. And I will just search, let's say I have a problem on my how to collect cash.

I would just say a cash collection problem. And then I will get like a lot of knowledge on cash collection problem. But now you don't have in front of you a search engine. What you have in front of you is the model that can generate answers. And then you need to consider that is like an assistant or a teammate and a teammate.

You cannot talk to a teammate. Like you talk to Google. You cannot, like, Adam, if we work together and I ask you by, you know, in our chat, I just tell you cash collection program. You will just reply with a big interrogation mark like, what do you want, man? Like, what should I do with and with, generative AI is the same.

If you do if you put keywords, you will get really, like, broadest words that will not bring you further.

Yeah.

But then what you should do then is okay. Because what you are the model in front of me knows a lot about the topic. Yeah. I need to explain what I need from the model. So I will give context.

I will say, oh, I work for a manufacturing company. I am the CFO. We need to solve our cash creation program. Can you propose an action plan? And now you get really something concrete.

And what you will get is actually an outline of action plan. And there, a lot of people, they will stop again because they say, okay. That's fine. I have a list of topics. Nothing really new because I'm CFO.

I'm experienced. So everything that I see here, like deal with inventory, payment terms, it's something they will not learn much. Yeah. But you get that because you are also limited by the model in front of you. That can only answer one question at a time by limited characters.

So if you know that and you ask, okay, how can I get deeper information from you? Then you will learn that if you ask now in the action plan, action by action, if you go deep, and that's what we call the chain of thoughts Yeah. In front of generic. If you ask, okay. Let's imagine I want to improve the payment terms.

On this specific action, give me more detailed tasks. And then you drill down this way, like 3, 4 times, and you arrive to really concrete action plan that you can use and go straight away and talk with your team to work on it. And that's what changed people from just getting a broad and swear and thinking that AI is not really useful to people who are already, like, 10 steps ahead because they already implemented everything were suggested as a action plan.

Mhmm. I think that's a great explanation, especially when it comes to prompt engineering. It's a different way of asking questions. Like, you can't put, you know, cat plus field plus, you know, happy person, you know, like you would do in Google back in the days, and you would get these great pictures or things you need to get more detailed and say, you know, what type of mindset that you want the thing to have, what what you're looking for, and then you can ask more questions to get more into it. And I I think that's a

really good way of looking at it. The prompt engineering is

is really

just common sense. Yeah. Us and understanding of how the model works. So if you know, okay, when I ask format and I say format a table, you know, that you will always get a table then, you know, you need to use these 2 words together. And if you know, okay, I want you to use to speak like a kid.

You know, like for the middle, each means changing the, the language form. So you need, like, to engineer the way you get the, output by giving some of the parameters, but you can use your own words. But once you learn that, like, okay, what are the parameters I can use to tweak the output in the way that I will get what I want?

Yeah, exactly. And if you want a good laugh, ask your generative AI engine to say, tell me like a pirate. And it'll be quite amusing to read their response if they do it in the wording of a pirate. Children children especially appreciate that one. So if you've never done that, I encourage everybody to try it at least once.

As we're looking ahead, you know, talking about AI, you know, are there any emerging technology or trends that you're seeing that you're kind of excited about that you're saying, hey. This is gonna really shape how we go into the future?

Yeah. The, the first one is the low code AI. So low code AI, you can have it, for example, in, Azure, cognitive services. So it's from Microsoft. It's mini module that are already coded to help you.

For example, one of them is the recognition of invoices. So, like, they read PDF, and they will give you in a Excel table what is inside the invoice. And this is already parameter that you just need to connect. Where are the PDFs? So is it in a folder or are they like coming from some automation from emails?

And then where should you save that in which Excel table? And then you have, you can choose if it's an invoice or if those are receipts. Cause we know it's not the same or is it maybe purchase orders, etcetera, etcetera, or maybe a legal document. And so those exist already in, in the Microsoft environment. And it will help a lot of team that maybe don't have the budget to have, people that build the model for them or don't have the budget to install new tools.

If they have this in the Microsoft environment, they will use the power of AI, which recognizes the invoice. Thanks to, image recognition, which recognizes the words. Thanks to the NLP, so natural language processing. And then we'll we'll we'll transfer this information in a Excel table that you can use after to make your accounting entries or to make some analysis or just maybe to save the information if you are somebody who, wants to have it somewhere saved and easier to use than just go scrolling through hundreds of PDFs.

That sounds like a really exciting tool. I'm excited to get my hands on that as well. That that's really I I I think the more tools, like the like you said, the low code things, the more little things that people that organizations do, that software companies do, I think the more that people are gonna really accept it and really start to use it in a in a better way. Before we end it, I just wanted to give you an opportunity maybe to talk a little bit about your AI Finance Club. I think it's a really cool venture that you're doing, and I I figured, give it opportunity for you to talk a little bit about it for our audience.

Yeah. Well, it's actually interesting in in 1 hour or 2, we have, our monthly master class. And, just to give an example, there'll be a person who implemented, so Azure, what I just talked about. He implemented that already, like, 50 or 100 times at some clients, And it's going to teach us everything about it. And like this, all of the members of the AI Finance Club will learn if, so how Azure work, how cognitive services work, how it can help them.

And my goal by doing that is instead of everybody on their own trying to figure out how to use AI for finance, is to bring all of the motivated and people that are convinced that AI is going to help them and help their finance team to that we all are together to learn together and to combine our efforts. And, basically, me and my team of experts, we do all of the hard work. We research and read every day. We then sensitize that, and we make it in an easy way to learn. And we have master class.

We have video courses every month. We also have we are digesting all of the news because every day there is AI news and, like, it's really hard to keep up with everything. And at the end of the month, I send a summary of the 5 most important news that are relevant for finance because not everything is relevant for finance. And I already show what is relevant for finance. And we also show what are the best AI tools to use and where.

And every month, we go through one finance process. And on this process, we show how you can use AI. So we did it, for example, on revenue forecasting. How can you use AI to help you forecast revenue? And so people can join, and the goal is ready to do the work for everybody so that for them, they can take it, go with their team, and, implement everything they learn with us, and then just elevate their AI maturity in their own organization.

That's great. So I encourage everybody to click the link in the show notes to, check out the AI Finance Club or or connect with Nicholas on LinkedIn. Nicholas, thank you so much for, coming on the podcast today. I really appreciate it.

Yeah. Thank you. I think what you are doing is good because, again, not everybody has the time to figure out where is the best information to learn about everything that is happening in the world of finance. And, by doing that, you help all of the finance people that are listening. So thank you, Adam, for doing this.

This has been Count Me In, IMA's podcast providing you with the latest perspectives of thought leaders from the accounting and finance profession. If you like what you heard and you'd like to be counted in for more relevant accounting and finance education, visit IMA's website at www.ima net.org.