There may be errors in spelling, grammar, and accuracy in this machine-generated transcript.
Blake Oliver: If you'd like to earn CPE credit for listening to this episode, visit earmark Cpcomm. Download the app. Take a short quiz and get your CPE certificate. Continuing education has never been so easy. And now on to the episode.
Blake Oliver: Hello everyone, and welcome back to earmark. I'm Blake Oliver, and today I'm chatting with Nichoas Boucher who is the number one finance thought [00:00:30] leader on LinkedIn. Is that right? A million followers on LinkedIn, Nicolas. Uh, quite an accomplishment.
Nicolas Boucher: Yeah, it's crazy to think about that. Uh, like, I would never have thought that I could reach something like this 2 or 3 years ago when I started. But, um, what is good is I always believe technology is a good way to scale and help a lot of people. So on this reach one of my goals, which was to help more than 1 million people. And that's what I'm doing every day on are LinkedIn.
Blake Oliver: And [00:01:00] you're specializing in helping people with artificial intelligence understand what AI is going to do to accounting and finance. And you've put out some great videos showing like practical examples of how to use AI. And this is exciting to me because I read about it all the time. I use it myself, but when I talk to accountants, I find that maybe 20% have used AI tools. And so there's [00:01:30] a lot of talk, but it doesn't seem like, you know, we're actually, as a profession, using it all that much. Is that your experience as well?
Nicolas Boucher: So what was interesting is when I was asking in my trainings where people are in, in, in terms of usage, one year ago, we were there like only 20% use it, 3,040% used once, but don't use it now we are more at 50% use it then you have still [00:02:00] 10 to 2020% that don't really use it. But the adoption is really increasing and each 3 to 6 months there is, I would say, a new phase of adoption. So for example, two years ago you have almost nobody was using it like was the start of ChatGPT. You had only the the technologists, the people that love technology will will use it, but they will use it just to write emails. Then six months after, you will have [00:02:30] 20, 30% of the people that will start using it, using it for emails, but then the technologies will start using it, for example, for financial analysis. And every six months you see something like this. And right now the first movers, they are not even only in Egypt anymore. They are creating their own agents using this low code platform or even with coding with Python. And that's what is interesting to keep up because you cannot just [00:03:00] stop learning. You have to learn and continue progressing. And the goal is to also bring everybody with you because you can. The goal is that everybody adopt technology to be more efficient.
Blake Oliver: Okay, so what can we do with AI? Like right now if I am working in accounting and finance, other than writing emails, written communication, which I think we're all pretty familiar with at this point, right? Write me a poem, write me an email, draft me a memo. Like what can we use it for in our actual [00:03:30] jobs as accountants?
Nicolas Boucher: Yeah. So I came up with some examples because I think it's important to to show it. I think one of the new type of business that is more and more there is the SaaS business and SaaS business has a specific metric to watch is, for example, the retention rate. How much of your client are you keeping over time? Because it's a lot of subscription business. And the thing is, if you want to do that [00:04:00] in Excel, if you want to do, for example, a cohort analysis is really long and really hard. So, um, let me show you, I will screen share my screen. So this for example, if you want to do that in Excel, this cohort analysis retention rate. So it's a heatmap for those people that only listen to us. It's a heatmap where on the left side you see the different types of cohorts. So when the clients came so which months they started [00:04:30] to be a client for us. And we can see that, for example, for the first one the clients coming in January 2022, we can see that over time, after 22 months to 23 months, only 40% stay with us.
Nicolas Boucher: So that's what is telling us here in this heatmap. And now if you never have done that, you will need to go online. You will need to find templates [00:05:00] to build this. Or you can use the most efficient technology. Right now you can go and use ChatGPT to do that. So let me show you how. So the basis is a table where we have dates, customer ID, product and invoices. If this table I'm going to use it and later to upload inside ChatGPT then. [00:05:30] So here I am using a ChatGPT version where my data is protected because I'm paying for it. I have a teams account, so I opt out for ChatGPT training on my data and I will just prompt and say can you do a cohort analysis visually? The word visually is super important here and I upload the file with this prompt and you will see that the GPT is first understanding what is inside the file. [00:06:00] It uses actually Python for that and then it will build the same type of graph. Wow.
Blake Oliver: So you've uploaded the Excel file which had four columns. If. Can we go back and look at that. Yeah. It, it, it had the date customer ID product and the invoice amount. And so now you've just uploaded that you've given a simple instruction to ChatGPT to do a cohort analysis visually [00:06:30] by month on retention rate. And and it builds it.
Nicolas Boucher: Exactly. And really like.
Blake Oliver: This color coding, all this conditional formatting that would normally be like quite tedious.
Nicolas Boucher: Yeah. So you have like a lot of intermediate phases to build that in Excel you need first to Do to have formulas to calculate your cohorts. You need then to calculate the retention rate. And then you need to apply conditional [00:07:00] formatting which is also with variables really hard to do. And um, that is like an example when I show that in my training, people stop thinking that it's just a nice toy to play with for emails. They start to see the potential because they know how much time you need to build this.
Blake Oliver: Quantify this for me, Nicolas. If you had to build this chart in Excel from scratch, how long would it take?
Nicolas Boucher: I [00:07:30] think one afternoon, 3 to 4 hours.
Blake Oliver: 3 to 4 hours. And we just did it in a matter of minutes.
Nicolas Boucher: And I will say like 3 to 4 hours. You will need to have done it once or twice before to know how to do it. I think if you never did it, you will probably need one day because you will have so many trial and errors. Yeah. So yeah, if you have the template, I will say like it's just updating the figures every month. It's less, um, you save less time, but still, you even like updating a template. You are never that fast.
Blake Oliver: Okay, [00:08:00] so we see a preview here in the ChatGPT interface and you're using the Foro model, like can I, is this just an image. Can I get this spreadsheet now.
Nicolas Boucher: So yeah this is only an image. And if you want you can ask because here's Python who did the image. If you want you can also ask to have those results in a CSV file or Excel file. So you just need to ask and to prompt please [00:08:30] extract that in an Excel file. But you will not get because this type of graph doesn't exist in Excel. That's not a function of Excel heatmap. So you will just get a picture. Um, either you can export this picture in a PowerPoint or PDF to reuse it, or either if you just want for each of the cohort by month Payments the data. You can have the CSV file that will give you this information.
Blake Oliver: So how do you when you look at this, validate that this [00:09:00] is correct? Because that's one of my big concerns when I when when you ask I to do some analysis, like how do you how do you know it didn't hallucinate.
Nicolas Boucher: This is a really good question. So actually this raises a lot of questions after because at first there's the wow effect where like wow, really cool. It can do that. And then once we have digested that there are actually four questions that come. The first one is confidentiality [00:09:30] of data. I could upload my data inside ChatGPT because I have a pro or a teams account, and my business allows that. But there are many other business that are not yet there. Uh, maybe, uh, the employees, Is the. They have not the authorization to use ChatGPT for confidential data so they are stuck. So that's the first problem. The second [00:10:00] one you just raise it auditability. How are you going to understand that and to make sure that this calculation is correct. But also how is your manager going to be able to prove that? How is your, um, your colleague if he's taking over your job, the internal auditor, the external auditor? How are they going to be able to prove check that what you have done is correct. Then third, large data is actually not possible yet in ChatGPT [00:10:30] or the other llms, because they cannot process that much data at once. And then fourth, when we think about AI, we always think about saving time. But imagine having to do this type of analysis. So a heatmap where you need every day for ten companies to go and chat with ChatGPT to get the output. That's actually not really time, um, time efficient.
Blake Oliver: Right?
Nicolas Boucher: And this is why we have these four problems [00:11:00] to solve this.
Blake Oliver: I was going to say the other issue with this is with the scalability, right? Is is like you might get a slightly different version of that chart every time you ask for it. Yep.
Nicolas Boucher: And actually that's a that's a good point too.
Blake Oliver: So, so so you've saved you know, it maybe it took it would take you 3 to 4 hours to build the heat map in a tool. But once you build it you can reuse it over and over again. You have auditability because you created the formulas, you created the conditional [00:11:30] formatting. And and then every time you update the data, that doesn't take very long, right? So have we really saved that much time with ChatGPT in this example?
Nicolas Boucher: And that's why there's actually a secret formula to solve this Where we are going to replicate what ChatGPT does, but in our own environment. So here again we have this file and you can see what [00:12:00] I do is I go back to the chat where I was, and you could think that maybe the first time I just uploaded fake data. So no problem about the confidentiality of data. But now what I'm going to say is I'm going to ask ChatGPT to give me the code that was there to create this table. And once I have this code, [00:12:30] so you can see here that we'll get all of the code for um, for Python. I can copy the code, I can go to something called Google Colab and I can paste the code inside. I can then upload my file into Google Colab and Google Colab is inside the Google environment. So if you are already using Google Sheet, Google Drive, Gmail, you are able to use this because [00:13:00] it's inside Google. And then I just run the code on my file. And now, Blake, what are you going to see?
Blake Oliver: And now we've got the chart. We've got the heatmap. Okay. That's amazing. So now we've solved that problem. We're not having to go into ChatGPT and ask and prompt it over and over again. We've basically built an app, yes, a small app, a mini app inside of a colab. And [00:13:30] you can just reuse it over and over again on new sets of data.
Nicolas Boucher: Exactly. So now it solves the confidentiality of data because you are not in ChatGPT, you are inside your Google environment. And for those who don't use Google but use Microsoft. You have the same type of tool for Microsoft called Visual Studio, where you can also run this mini script that creates the heatmap on your Excel file. And the second part we said [00:14:00] was Auditability. Here I can see the source. So the Excel file I can see the code which is just some formulas saying okay, based on this columns create the heatmap based on these conditions. And it's fixed. It's not random. It's not like a black box. You can see all of it. You can read all of it. So it's auditable and traceable. And then you can like you say you can scale it. Because now every day I just need to play the [00:14:30] the button. And I will get the heatmap updated based on my new data. Or if I want to run this on 100 files, I just need to change a mini part of the code saying. Produce the heat maps, but on this folder for all of the files, and then it will do like one after the other. All of the heat maps.
Blake Oliver: And how would you figure out how to how would you figure out how to make that change to the Python code? Because I don't know Python. Yeah.
Nicolas Boucher: So and that's where um, that's [00:15:00] where like we can stop now the screen sharing. Um, so that's where what is good is I am also not a coder. When I did that, I just asked ChatGPT the code to do it on my file and I just need to explain the file. If now I say now, you see I have this file, but 100 times in this folder called um, SaaS analysis for all companies. And if I explain correctly with my words, now do [00:15:30] this analysis on all of the files inside the folder. Then it will tweak slightly the code. And then I can go back in Google Colab and change the code and say, okay, now I use this code instead of the other one, and then I will see the code running and having the results of all of the heat maps that I need.
Blake Oliver: So that's amazing. Okay, so so basically using this method we can generate Python code to create visualizations of data from CSV [00:16:00] files, Excel files. And I imagine we could we could do all sorts of different types of visualizations. What what other kinds of visualizations have you created other than heat maps?
Nicolas Boucher: Yeah. So one thing that I learned through ChatGPT and now we don't even need Python. I will show you something also with Excel. That's the first time I saw it. I was telling to myself, this is really a big change. This [00:16:30] is going to change all our life for all of us in finance. We cannot get rid of this anymore. I remember having said that to my wife in August last year when it happened. So let me show you the screen again and show you my example. So imagine here a headcount analysis where you have um, so you have here the name of the employees, the department, the position, the salaries [00:17:00] and the increase decided for the year. So here we'll go back again to ChatGPT. I will upload the file and I will ask ChatGPT to do the analysis on it. So this time I'm not really saying what I want as analysis. I'm just saying you are a McKinsey consultant for once that we can have them for free. And, um.
Blake Oliver: Mckinsey is in trouble.
Nicolas Boucher: And I'm asking like, okay, we have a board. [00:17:30] Please draft the visualization for me. And you can see what is going to happen, because now I'm explaining a bit what is happening behind with ChatGPT first. Actually, ChatGPT cannot read Excel files, but ChatGPT can tell to Python. Hey Python, you are my friend. Please analyze the file for me. Tell me what are in the first five lines. Then python reply back with a text and ChatGPT understand text. [00:18:00] And now ChatGPT knows that in this file there are five columns with name, department, position, salary and increase. And based on these five columns, ChatGPT proposes analysis. And then GPT talks back again to Python and asks, hey my friend, can you now do these graphs for me? And Python will reply back with the graphs on the data. And this is really important to understand that is not ChatGPT making the graph itself [00:18:30] is ChatGPT. Being a mini coder that knows how to ask Python to do the calculation so it's not a black box. Also because you can read the code, you can understand how it is built and it's in front of our eyes. It's just instead of clicking in a lot of buttons everywhere in Excel is just a code, because it's much easier for a model to code rather than to manipulate an Excel file.
Blake Oliver: So the charts we're looking [00:19:00] at here are bar charts. We see average salary by department. We've got administration, logistics, maintenance, etc. um, comparing those, we've got average salary by position with nice color coding for each position. And we can see that management uh, does really well. Management has the highest average salary of course. Uh, there was one uh, I think if you scroll back a little bit, there was the first one. What was the [00:19:30] first one? Uh, departmental head count distribution. So production 600 head there. Number of employees by department. Fascinating. So we didn't even have to ask. I to make a specific chart. We just said do some visual analysis of this data, this headcount data this and and it came up with three possible charts.
Nicolas Boucher: Exactly. And so here a lot of people [00:20:00] will tell me this is like baby graphs here. Like it's really easy stuff to do. Why do I need AI for this.
Blake Oliver: Right.
Nicolas Boucher: And then I was okay I do agree it's a bit too easy. So then let's go to another example where actually now I want to see the salary distribution by department. And again ChatGPT will ask Python. And now we will have a graph that I remember the first time when I was in front of my computer here, [00:20:30] and I saw the output of the analysis of salaries for this file. And I saw this bar with like, it's like it's called a box plot, but it's like a candle bar with on the left side the lower salary and the right side the higher salary in the middle, the average. I was like, wow. Like, this is actually the best way to show a distribution of salary. And after 15 years of finance, I never used that, but that was the best way [00:21:00] to use it. And now I have an AI tool that shows that to me, I was just like blown away to to see that what is possible to do.
Blake Oliver: This is a really great chart. Okay. So box plot. So I've seen these before, but I didn't know what to call them. So we're seeing like we have a line and each end of the line is like the highest and lowest salary.
Nicolas Boucher: Yeah. It's like the person, you know, like the percentile like the, the last 10% and 90%. And then you have dots on the outside of this range that are exceptions, outliers.
Blake Oliver: Oh the outliers. [00:21:30] Got it. So we can see an administration. The range is like, you know, below 4000. Right. There's sort of like that box. The box represents what, like the 80% of the salaries or something like that. It's like most of the salaries fall within 2000 to. It looks like a little over 3000.
Nicolas Boucher: Yeah. And we can see our management [00:22:00] here that is paying very well that all of the outliers all of the dots. So that's a good way. You know if you go to a to a meeting, you can show the behavior of all of the departments against each other. You can also show what are the outliers. And you can debunk a lot of I will say, like a lot of legends about salaries of departments, because here on one graph you have all of the information you need and they are visual to [00:22:30] people. It's not a table where it's really hard to compare.
Blake Oliver: Yeah, this is where averages would not be helpful at all. And this is this is really interesting information. The the outliers really skew the average a lot higher. Exactly. So so is this something that ChatGPT came up with as a suggestion or did you have to ask for this?
Nicolas Boucher: Um, no. So really the what happened the story is, was in August, not [00:23:00] even August last year, um, was in August 2023 when it came up with the possibility to read Excel files. I prepared a training and I uploaded that, and I asked for more creative ways to analyze the salary distribution or the salaries. I just say salaries. And it came out with like 3 or 4 different files. And the last one was this box plot. And I remember, like I was here in this office and [00:23:30] just like stormed out of my room, went to my wife and said, you need to see that. Like you need to see what? Chatgpt just did in front of me. Uh, it just did a graph that, I don't know, box plot. I mean, I know what it looks like, but I don't know, you could use it for finance. And, um, then the fact that me, after 15 years, I could learn something from AI and that it was actually the best way to show it. I was really challenged and a bit, uh, first a bit disturbed, [00:24:00] but really, actually really excited because I thought I need to, to show that to people. And that's what I did. And in my trainings, I always show this example because I think on top of teaching that I can do that. I teach to a lot of people that are running analysis on expenses, salaries, revenue, that it's a good way to show distribution between brands, between countries, between departments. And people are like, oh, I could use [00:24:30] that also for my own analysis.
Blake Oliver: Yeah, AI is a very powerful tool for helping me think of things that I'm not thinking of, because it's been trained on every possible chart that's out there. So if I don't know what a box plot is, I can't ask for it. I've never thought to use one, but I can suggest it. And like you said, it made for suggestions for you and one of them was brilliant. And [00:25:00] that's the beauty of this is you can you can ask for as many charts as you want, and maybe one of them will be genius. It will be perfect for your situation. So it's not that the AI has to be perfect or come up with great answers all the time. It just has to give you something you hadn't been thinking of.
Nicolas Boucher: I heard something the other day where in another podcast where somebody said, if AI is like having discovered a new continent with [00:25:30] 10 billion genius on this that are willing to work for free. And it's really that like you have like this free resource, uh, and uh, and you know, here it's not the end of the story. You want to see the end of the story on this.
Blake Oliver: Graph, please. Let's go.
Nicolas Boucher: Well, again, the problem is a lot of people are asking that's fun to have that in ChatGPT. But you just showed me something on headcounts and salaries. Never in [00:26:00] my life my company will allow me to put that in ChatGPT. I'm like, okay, let me show you then. How do you apply this inside your own environment? So now instead of using ChatGPT to do this graph, I'm just asking to ChatGPT can I do that in Excel? Because that's where I want to do my analysis. And then I ask to give me a step by step approach on how to do this in Excel, [00:26:30] and ChatGPT tells me the four steps to perform that. And I mean, you can read on the screen that it says step by step. It tells me how to create it, how to customize it. And the most important here is that you just need to go in Excel. You select your two columns Department and salary. You click on insert and you choose the graph in statistics that is called Box and Whispers. And now you get the [00:27:00] same type of graph. I think here is a bit 90 degree inverted, but it's the same graph with the outliers with the range of salary. And now you are in your own environment. You can every month run the same analysis. You don't need ChatGPT anymore and you have no problem of confidentiality of data.
Blake Oliver: Amazing. Wow. So okay, I am convinced 100% of the [00:27:30] incredible value of AI for analysis and especially visual analysis. What else are we using AI for in finance or accounting?
Nicolas Boucher: I can, um, let me give you another example of maybe one tool that or two tools that I really like. Um, let me because I created this, this market map here. So there's a lot [00:28:00] of tools on AI for finance. And you can check that on my LinkedIn. It's like top 100 AI finance tools. But there are some tools that I really like.
Blake Oliver: So when you say tools, we're talking about like wrappers for AI tools that plug into these llms that we can then use for our work because like that's been, I think the main thing stopping us from using AI in our actual accounting finance [00:28:30] work is that it's hard to integrate our data, the tools we use with AI. Copy paste a lot of times you can't get it into a format that you can use reliably in in ChatGPT or in Claude.
Nicolas Boucher: Yeah, exactly. So the the limit is now those tools are not integrated. But there are people like here's concourse.io and it's a French guy who is running the company, but he's in the US. So that's why I have [00:29:00] affinity for this tool. But actually I just show it because I really like what they did. I have no partnership with them and what they have is they have connection with QuickBooks online. So here their data is connected with QuickBooks online, and they will just ask to create a monthly report showing the difference between the actuals and the last month. So first dare to ask if it's okay [00:29:30] to follow the structure of having first executive summary, cost and revenue analysis, etc., etc. and then the tool will take around two minutes. So now it's a bit accelerated, but in two minutes it will produce this monthly review. And now you have all of the narratives of what happened in the financial year of December 2024, plus the tables plus the graphs. And you have section by section. So first executive summary, then [00:30:00] revenue analysis, then cost of sales analysis. And you can tailor that. You can say oh I want a section more or less. I want to reduce the amount of information. I want to add more information. And here you can see that you can delete also the information inside you can add your own text inside.
Blake Oliver: So what is the name of this? What is the name of this course?
Nicolas Boucher: Dot io.
Blake Oliver: Concourse.io. And? [00:30:30] And so it plugs into QuickBooks online. Uses the data from QuickBooks through the API to generate these. Uh, what would you call this type of report?
Nicolas Boucher: It's a monthly summary review. You know, like with when you give your comments with some graph of what happened, uh, on last month's financials. Uh, what a CFO will show to the rest of the board, or what a fractional CFO will have to do for [00:31:00] their clients. You can do that, and you can tailor it here. We can see that we just added a new graph with total revenue. And I was really amazed that it's possible to already do that today, because the part that is hard is to combine commentaries with data and with visuals. All of one thing that is Not only just pretty, but also efficient.
Blake Oliver: Right? We can we know we can generate [00:31:30] financial statement commentary like executive summaries. You can do that if you upload a set of financial statements. But then integrating that with charts can be a really manual process.
Nicolas Boucher: Yes.
Blake Oliver: Just building it. So so this tool basically leverages AI to generate the narrative. The summary. What do we have here. We have we have an executive summary. If we go down further what else do we have.
Nicolas Boucher: You have like uh revenue. So executive summary you have after revenue analysis [00:32:00] you have after operating expenses here. It was cost of sales. I think you have something around like the overhead expense or here top ten vendors by spend. So this is also something you can customize or delete the sections you want.
Blake Oliver: So I can customize the sections. And then the next time The next period I can run the reports again and I get it in the same format.
Nicolas Boucher: Exactly. That's what Matthew told me. So the founder [00:32:30] is that it learns what you did last month. It will repeat so you don't have to again. Re-explain what is your type of report you want? And it has behind some different type of agents. You have an agent that is there for commenting. You have an agent that is there for graph. You have an agent that is there more to calculate the KPIs. Um, and you can always go back and check also uh, the where the data come from because it's linked with either QuickBooks. They also [00:33:00] have NetSuite and I think they are working on other integrations.
Blake Oliver: So what is what does this mean for finance jobs? Because ever since the invention of the electronic spreadsheet jobs in Fpna financial planning and analysis have exploded. Um, that's where that's where all the job growth has been. But now we have a tool here, a set of tools that basically takes all those really [00:33:30] complicated Excel skills. And the time that we spent building spreadsheets and doing manual analysis, building charts, like that's all going away. It looks like. Like, what does this mean for finance?
Nicolas Boucher: So I don't think all of this is going away. It's just like changing, you know, the the amount of time you spend on it because you will still need one person who is behind that to pilot the creation of this report by prompting [00:34:00] the machine and asking exactly what type of report we want this month, because we know that this business has a problem with vendors. So we want to have a section on vendors, but instead of spending a week with five people building, that is just like going to be a 30 minute work, and then this time you can reinvest in now analyzing which vendors are good, which ones are bad, and then starting to work with procurement to challenge them on to make some [00:34:30] some savings, to take some some commitments for the next board meeting. And I think for us in finance, it will help us really be more in, um, like navigating the business or help the management navigate the business rather than just being people reporting figures, which I think like if you read all of the finance thought leadership texts since 10 to 15 years, we talk a lot about business partnering. We [00:35:00] talk a lot. We talk a lot about adding value. But when people are behind their Excel file, they cannot do a lot of this, right? And it will free us from this time to put us more in front of our our management, more in front of the other part of the business. And because we are so lucky to be at the intersection of all of the other departments, all of the data being really close to management and understanding the strategy, we can help more. We can [00:35:30] do much more in finance. And that's where I think our resources are going to go.
Blake Oliver: So it sounds like we're going to be spending less time making the Excel workbooks and the reports and more time helping to solve problems.
Nicolas Boucher: And also right now, if you talk to a lot of CFOs, they have anyway a hard time to find the stuff. Um, you have less people coming into the accounting job. [00:36:00] You have a lot of people going to retirement. The turnover is really high. So as a company, you need to be attractive to give this tool to people, to tell them, look, when you come to me, you are going to try to work with the right tools where you are going to spend time working with humans rather than just being an Excel file. And I think it's also really attractive for a company to to think like this in the future and make plays to what we are the best at, meaning [00:36:30] being human, rather than letting us doing half of our time mundane and robotic tasks.
Blake Oliver: Yeah, it's it's it's interesting because, you know, there's a lot of fear around job loss. If your job is sitting behind Microsoft Excel and and copying and pasting in data and, and making a bunch of workbooks day in, day out, that's going to be [00:37:00] reduced dramatically. I mean, we saw in your example.
Nicolas Boucher: If you are like this, I think you are also the best place to be the person who is automating that. Right? I had the chance to meet the and now is a good friend, the CFO of Oracle. And he said one thing one day that I will remember is that the responsibility of the people is not only to do their job, it's also to improve the way they do their job. So [00:37:30] if the technology is there to make their job more efficient and to reduce the amount of time they spend on it, it should be one of their responsibility to work on that. And if you infuse that in the people, in the responsibility of their job description, then the person who is behind all of these Excel files, if they get the skills to automate, to use more AI, then they will be really valuable for the company compared to somebody who is not investing in those skills.
Blake Oliver: Where [00:38:00] where can I, you know, other than like talking with you, Nicolas? Where can I learn these skills? Like, how did how did you learn how to do this?
Nicolas Boucher: So I think, Um, I'm lucky that I'm really curious. And I always want to get the best out of the technology. Uh, two weeks ago, for example, I built an agent that is reading my email and categorizing the emails for me and then even drafting [00:38:30] emails. So. But I did that because I spent time on YouTube watching people doing that. I called the person who built the template for that and asked him, okay, like, how can I implement it? And those are extra steps that I think not all employees can do it, because maybe people at home, they want to take care of their kids. They have they want to go and play sports. Uh, but what I think is first you need to follow the right people on LinkedIn [00:39:00] or YouTube. So I'm showing all of this on YouTube if people want to follow. But then if you are already a manager and you want your team also to grow on this, then you need to start also implementing some start of some type of educations. I run trainings with finance teams and just in two hours is crazy to see how much people can learn and can implement everything that I showed you today.
Blake Oliver: So your YouTube channel is [00:39:30] at Nicolas Boucher Finance. That is Boucher b o u c h e r. Nicolas Boucher Finance I've got the URL there at the bottom of the screen for those watching on YouTube. Go, go check it out. That is that's what made me decide I had to talk to you. Nicolas was watching some of your demonstrations. Because you just make it so clear how much time there is to be saved doing this. And [00:40:00] it's it's incredible, honestly, like the outputs that you're getting. Um, and I noticed that when you were, when you were doing these demos like these were using not even the latest models. Uh, yeah.
Nicolas Boucher: The the question about, uh, which model is the best? Is it better to use ChatGPT or Claude? I'm always replying. Imagine if you have tomorrow to, uh, to go in a run against Usain Bolt. Does it [00:40:30] really matter which shoes you have against Usain Bolt? It doesn't matter. Like Usain Bolt will beat you. Like, will just go faster than you because Usain Bolt knows how to run. So before you choose which model, which tool, just know how to prompt and know which use cases you can do in these tools. And then you can see which tools you like the most. But don't try to buy the best shoes before you know how to run.
Blake Oliver: Mhm. And you know it doesn't hurt to subscribe [00:41:00] to multiple tools. They're not that expensive if you think about how much time you're saving I mean I would hope that most companies would authorize $2,030 a month for a finance employee, who is probably costing them quite a lot in salary. It's just.
Nicolas Boucher: I push, I push that when people ask me, okay, where should I start with my finance team? And I'm telling them, like, make sure everybody has a license [00:41:30] of LLM tool. If you are already embedded in Microsoft. So you use outlook, you use SharePoint, you use power BI, you use Azure. It just makes sense to go with Copilot because it's the same security standards you have. You don't need to add logins, details. Everything is integrated and is in. The contract is just a line more in the contract for your IT team and your procurement team. If, on the other hand, you are already really into Google, then it makes [00:42:00] sense to take Gemini because it's going to be more and more important. How much integrated is your AI? Because now if I am in Gemini and I ask to draft an email with just one click of the button. It can cover that in a real email in Gmail and um, and or if you are in copilot and you ask a question, it can retrieve the information from your email. If you have copilot business. And that's like I saw one client two weeks ago that was doing that in front of me, where they were asking [00:42:30] how many stores they are going to open next year. And it was just like a simulation, because we were doing analysis on the best amount of stores to to open next year. And copilot retrieved one email where this discussion already happened and told exactly the number of stores that was decided and everybody was blown away that copilot could do that.
Blake Oliver: It went into the email history and found the particular discussion.
Nicolas Boucher: Exactly. [00:43:00]
Blake Oliver: Yeah.
Nicolas Boucher: That's valuable though it makes more sense. Like how much integrated is your AI tool rather than how powerful? Because every 3 to 6 months those models will update and will be better than the best ones six months ago. And if, like me myself, I am with my team like we are a small entity, we are not really integrated with Google or Microsoft. So I went to ChatGPT, but I could move tomorrow to Gemini without problem as well. So [00:43:30] really it doesn't matter, just get licensed, make sure your team is using it without having fear of data security. And those tools, they use the best standards in terms of data security. So if you sign a contract with them, you can read the data security protocol and just make sure you opt out for data training, which is normally by standard. But make sure for that. And then normally you are good to go. And you can use like a lot of my clients, you can use those AI tools for financial data.
Blake Oliver: Copilot [00:44:00] Gemini backed by Microsoft and Google. Same data protections if your data is already in the cloud with either of those. Really. There's not you know, you can't say there's yeah it's a no brainer. Right. And then and then for ChatGPT, like you said before, it's important to make sure that you sign up for the the team plan and opt out of data being used. You have.
Speaker3: To prove or prove if you are alone teams if you have a team and if [00:44:30] you are really a big company enterprise.
Blake Oliver: Yeah, yeah.
Blake Oliver: And with pro and team, I think you have to explicitly opt out which is.
Nicolas Boucher: With Pro.Yes. With teams is by default.
Blake Oliver: Okay. And then.
Nicolas Boucher: You do need to to to check to make sure. Because as they change their uh their settings. Yeah.
Blake Oliver: Be sure to check the settings for the opt out. You don't want the any client data being used to or company data being used to train the models. Um, well, this has been this has been amazing. Thank you, Nicolas, for sharing these [00:45:00] examples with me and our audience here. Um, where can people follow you, I guess LinkedIn, right. That's the spot.
Nicolas Boucher: Yeah if, um, if they don't follow me yet on LinkedIn, I think where I provide a lot of value. If you like the YouTube format, uh, check my videos on YouTube. I have videos about how to create a, a presentation with AI. So PowerPoint presentation, how to make financial analysis, how to use ChatGPT for accounting. And I [00:45:30] also invite experts on my YouTube channel where those experts will explain, for example, how to build a forecast tool for a big company using AI or how to use AI in investing. So this type of expert I have. So YouTube is the a good place to spend a lot of time together.
Blake Oliver: And one of your examples I really liked was um, researching KPIs, key performance indicators. What what what KPIs should I be considering [00:46:00] for my company? And and that's one of those examples where you may have quite a few KPIs you are familiar with and you know, but maybe there's some you have never explored. And AI is a really great tool for finding new ones, exploring new options.
Nicolas Boucher: And also where they can find me maybe to finish, if there are people in the audience that feel that they need to stay on the top of their game, especially on AI because [00:46:30] they have clients or they are leaders in their company. But we all know it's like so much time, so much information out there. I built a community called the AI Finance Club. It's exactly for that, where every week we give the most important content in forms of guides or masterclass or video course where you have, um, experts that I bring in and they will teach us in the community. What are the best way to use AI for finance. And this month, [00:47:00] for example, we have a big focus on how to create agents for finance. So that's the I finance club.
Blake Oliver: And that's really powerful because I. Agents can be used to do the same tasks over and over again solving that problem of scalability. Which, you know, we don't want to have to be constantly going into these chatbots and asking over and over again for the same thing. So and that's, that's like all within the last few months that AI agents have [00:47:30] become really possible. It's very exciting. Exactly, Nicolas, thanks for your time. Great chatting with you and I hope to catch up with you again soon.
Nicolas Boucher: Thank Blake. And could you like this? Because really cool to have this type of format in the, in the industry of finance. Uh, I think we have really the luck that so many people are willing to share. So you having the platform also to allow that. Thank you for having that.
Blake Oliver: My pleasure.
