Prompt Engineering for Accounting Closing: Netspeed Masterclass

We are the team that challenges what seems complex. and finds solutions to simplify paths. Since the year 2000, Netsped has been a partner brand for thousands of Brazilian professionals. We are present in the daily lives of companies, freelancers, and accounting firms that need to streamline accounting and management processes. More than just systems, we offer specialized knowledge in the segment and guarantee support for the best use of our technology. Come challenge what seems complex with us as well. Net Speed, accounting without the headache. Hello, everyone. Good afternoon. It's great to have you here with us. Welcome , everyone. It is truly a pleasure to wrap up Accountants 'Week here today. You who participated with us this week had the opportunity to be with us during these afternoons and also see which subjects were discussed that are of utmost importance to the accounting and management fields. Now, to close with a flourish, a subject that really, as we joke here, is sometimes controversial, right? But in truth, it is necessary; it is a necessary discussion and also a necessary moment for learning. So, if you are busy with so many other things, I ask that you stop for just a moment, if possible, to follow this super class we are about to have. We have set aside this last day for a masterclass on AI, AI and accounting. I want to thank you right now for following along with us here. I ask that you start interacting in the chat, okay? Let us know where you are watching this last day of Accountants' Week from, right? And also, of course, during our masterclass, you will be able to ask your questions. However, I ask that you hold your questions for a bit, because Artur Santini, our guest who is already here with us, will be able to address the issues that arise at the end, okay? So let's go, thanking Loreto for the technical support here in the studio, I will introduce our guest for today very quickly, so as not to take up too much time, because we have a lot of great content ahead, okay ? Our guest today is Artur Santini. Artur is a consultant, author, speaker, and also the founder of IAS Consultoria . He works as a professor, is certified by the Getúlio Vargas Foundation, and has vast executive experience in large companies in the technology and finance sectors. See what I mean? Yeah, it’s, it’s, it’s a quick overview of our guest's professional journey today. Well, Artur, I won't go on too long so I don't take up much time. I know we are all looking forward to today's masterclass. So, a very warm welcome to our Accountant's Week. And I'll say it again, it is a pleasure to have you here to close out this year's program.

Good afternoon. Thank you for the invitation from Netped to be here this afternoon with you all, talking about a topic that, as people have already mentioned, is controversial. I consider it controversial as well. I agree with that. And the idea is to talk about the practical use of AI, to talk about its use in accounting and finance, but in a responsible way, in the right way, and in a way that respects what we need to respect. So, also respecting your time, I want to take the opportunity to say thank you. So, you have there—we already have a lot of people connected —go ahead and grab the link, send it to that WhatsApp group I know you have. Because the chat today is going to be good. We will break this down into a few stages here; at the end, we will open it up for questions from everyone, and let's get to what we came here to do today, which is to talk about prompt engineering, talk about artificial intelligence, and what this "prompt engineering" thing is all about. And I'll start here with an image that is a bit of a provocation, right? How many people sometimes see artificial intelligence, right? For those who are maybe my age, they might catch the reference a bit; you'll remember a Tom Hanks movie from the 80s or 90s called* Big*, where he makes a wish to a machine, right? He goes up to the machine and says he wants to be big—a totally magical machine. And the next day he wakes up and he's an adult. And sometimes we find ourselves thinking that artificial intelligence is just the same way, right? That we’re going to ask it for things, something magical will happen inside that box, and it will solve all our problems, right? As if it were something very easy, very smooth, as if we didn't have so much responsibility in using the technology. So, I start with this provocation, with this image, because in my day-to-day work, we end up encountering a lot of this within the business environment. Uh, both within accounting firms—I have great friends in the accounting field—and in the general market, we see a lot of this scenario, right? Sometimes we’re talking to a business owner, talking to someone, and they say , "Gosh, Arturo, I’ve had this problem in my company for years, I don't have an AI.""To solve this problem, right?" As if it were some kind of box we're going to open, that’s just going to happen, right? That it’s going to grant our wishes. But in practice, for those who use it daily, for those who have had experience using AI—and I’ll take the opportunity to ask in the chat here —what artificial intelligence tools are you using? Do you use ChatGPT? Do you use Gemini, or do you use Claude? What are you using there? Write for us in the chat what artificial intelligence you are using, so I can use that as an example here as well, right? And in practice, we see very different AI responses, right? So, we came here to talk about artificial intelligence, but with a focus on prompt engineering. So, why is it that sometimes, for the same request to the AI—I'm using ChatGPT and someone else is too, I'm using Claude and someone else is too. Why do we sometimes get such different responses, right? I can get responses that are extremely simple , right? Uh, coming from an AI, or I can get responses that are more accurate. Who here has never fought with an AI, right, trying to get an answer and couldn't get it, or it answered something where I looked at it and I knew that answer was wrong? So, why does this happen? Our goal—all of this was just to contextualize what our objective is today, which is to understand why this happens, okay? Vanessa has already introduced me, but I'll introduce myself quickly here, for the sake of, uh, manners, right? My mother always taught us not to talk to strangers. So, I'll introduce myself briefly here. Vanessa already mentioned it. Uh, I think the most important thing here, beyond my resume and everything else, is my role within companies, since I have always been in the technology field, right? My background is in technology, and at some point in my career, I ended up moving into finance, working with accounting, auditing, and controllership. It is very difficult for someone from a tech background to commit to and understand technology as a purpose. What I mean is, we are very passionate about technology. Anyone who likes tech and relates to this knows that when we see something new, we say, "Wow, I want to use this no matter what ." But when we are in charge of finance , when we are in control of the company , we cannot simply implement a technology just because we like it. We need to implement a technology that makes sense for the company, that makes sense for the strategy, and that provides a return. So, my role within companies has always been this, right? One foot in controllership and accounting, one foot in finance. And this gave me, within consulting, a background—let’s put it that way— a practical experience of where to apply artificial intelligence, technology, and innovation, all with purpose and delivering a return to the company. But I don't really like this slide, actually. It's just so you know who is speaking with you here this afternoon. And when we start talking about artificial intelligence, I want to bring up a first realization for you here, okay? I often say that AI is a mirror of the one who asks. What is very common? "Gee, Artur, I made a request to the AI, I sent a prompt, I asked it for something, but the answer wasn't very good, the response from this AI wasn't very nice." I usually turn this question back as if it were a mirror. Did I ask the AI in a way it will understand, in a way that it can give me that answer? Because let's be practical here, right? When I'm using an artificial intelligence, let's imagine you are using ChatGPT, Claude, Gemini, any of those out there. You, as an accountant, are using it, but you will also have a doctor using it, a chef using it, a student using it. So, these tools we call LLMs, or these large language models which are the artificial intelligences, they have concentrated within them practically all human knowledge generated to date. So, well, if I asked that artificial intelligence a question and it answered in a way that I judge to be wrong, what is the highest probability? I'm not saying it's always this, but what is the most likely case? Not that it doesn't have the answer within it, but rather that I didn't manage to get that answer out of it. That’s why we say that AI is always a mirror. It is neither good nor bad. It will respond to what I am saying. This happens for several reasons, some of which I have listed there. The first one is that it doesn't have all the data. So, even though the AI has all the knowledge generated by humans to this day, it won't have that private data from your company. So, it won't have the P&L statement, it won't have the balance sheet, it won't have the chart of accounts for your company or your client's company, because that is not public or accessible data. So, it's no use for me to ask the AI about that. There are secure ways for me to train the AI so that it has this information. We will talk about that, but this can be one of the reasons. The second fact is that the AI can make things up, and the problem with AI is that it makes things up with great co