Earmark Podcast | Earn Free Accounting CPE | Excel and AI: Code Your Way to Hours Back Every Week

You open Excel daily, but are you using AI to make it work for you? While 58% of professionals have tried AI, only 17% use it regularly—a missed opportunity. Join CPA Kyle Ashcraft in this hands-on webinar to learn vibe coding—a no-programming approach using Cursor AI to automate repetitive Excel tasks. Watch Kyle transform messy spreadsheets, organize GL data, and reconcile transactions with simple AI prompts while keeping data secure. You'll get three ready-to-use scripts plus a framework to automate countless tasks and reclaim hours weekly.

(Originally recorded on October 20, 2025, on Earmark Webinars+)

Chapters

(02:18) - Meet Kyle Ashcraft and His AI Journey
(02:31) - The Importance of AI in Accounting
(03:30) - Kyle's Background and CPA Review
(04:38) - Live Webinar and Audience Interaction
(05:12) - Kyle's AI Projects and Cursor Introduction
(07:39) - Data Privacy Concerns with AI
(17:18) - Practical AI in Excel: Examples and Demonstration
(20:03) - Getting Started with Cursor
(23:55) - First Cursor Project: Cleaning Up Excel Data
(31:38) - Jumping into Financial Document Verification
(32:50) - Exploring Cursor's Privacy Settings
(33:48) - Understanding Data Retention Policies
(36:23) - Comparing Excel Files with Cursor
(36:48) - Analyzing Complex GL Details
(42:22) - Using Cursor for Recurring Accounting Tasks
(49:20) - Leveraging AI for Audit and Analysis
(50:53) - Practical Tips for Implementing Cursor
(53:56) - Q&A: Advanced Cursor Features
(59:04) - Conclusion and Next Steps

Earn CPE for this episode: https://www.earmark.app/c/2854

Sign up to get free CPE for listening to this podcast

https://earmarkcpe.com
https://earmark.app/

Download the Earmark CPE App

Apple: https://apps.apple.com/us/app/earmark-cpe/id1562599728

Android: https://play.google.com/store/apps/details?id=com.earmarkcpe.app

Resources

Intro to Cursor PDF Guide - https://mcusercontent.com/02dbcae4a3e3f15021db25a0c/files/deff5647-e0a3-51d7-4225-cf8b3a48532d/Cursor_AI_Quick_Guide.pdf

Webinar presentation - https://ai.maxwellstudy.com/

Connect with Our Guest, Kyle Ashcraft, CPA

LinkedIn: https://www.linkedin.com/in/kyle-ashcraft-cpa-7638a42a

Learn more about Maxwell CPA Review

https://maxwellcpareview.com/

Connect with Blake Oliver, CPA

LinkedIn: https://www.linkedin.com/in/blaketoliver
Twitter: https://twitter.com/blaketoliver/

Creators and Guests

Host

Blake Oliver, CPA

Founder and CEO of Earmark CPE

Guest

Kyle Ashcraft, CPA

Kyle Ashcraft, CPA is the founder of Maxwell CPA Review and an accounting educator whose YouTube channel has attracted over 500,000 views and 15,000+ subscribers. Known for making complex accounting concepts accessible and actionable, Kyle has been helping aspiring CPAs since 2019 and has personally guided over 1,500 students toward exam success.

What is Earmark Podcast | Earn Free Accounting CPE?

This show is brought to you by Earmark, your source for podcast-based continuing education for accounting and tax professionals. You can earn CPE by listening to the episodes on this podcast and more! Sign up for our mobile app to earn free CPE whenever you want, wherever you go. Learn more at https://earmarkcpe.com.

Attention: This is a machine-generated transcript. As such, there may be spelling, grammar, and accuracy errors throughout. Thank you for your understanding!

Blake Oliver: [00:00:04] Welcome. We'll get started shortly. Before we begin, here's how to earn continuing professional education credit for attending this session. There are no polling questions or attendance checks. Instead, you will demonstrate your knowledge by taking a brief multiple choice quiz using the earmark app. Visit Earmark app in your web browser. You can also download the mobile app by scanning the QR code on the screen. Log in or if you're new, create your free account. Be sure to use a valid email address, because that's where we'll send your course completion certificates from the home screen. Scroll down to find the course. You can also tap the magnifying glass icon to search for the course. Tap on the course artwork, then the enroll button. On the course tab, you'll find the recording and quiz questions. Complete the quiz to earn your CPE certificate. It takes us a few days to create the course on the app, so if you are attending this session live, please wait a few days. We'll email you a link to the course when it is ready. If you have questions or need help, email us at support. And now on to the event. Hey everyone, you open Excel every day, but are you using AI to make it work for you? Well, 58% of professionals have tried AI. Only 17% use it regularly and that's a missed opportunity. Hey everyone, I'm Blake Oliver and welcome to earmark.

Blake Oliver: [00:01:31] Today I'm joined by Kyle Ashcraft of Maxwell CPA review for a demonstration of a really fascinating use of AI. Kyle and I met at the American Accounting Association this past summer at their conference, and we started talking about AI and how accountants can use it. And Kyle brought up this idea of using this app called Cursor and Excel to automate work. And he gave me a little preview of this. I guess it was like a month or two. Right, Kyle? And and, uh, I was just blown away, and I wanted to share it with our community. So Kyle has joined us today, and we are going to see, uh, some practical AI in Excel that you could use and get started, like in under an hour. You could be doing this yourself. And Kyle is going to show us how, uh, today's webinar is also a live podcast recording. So this, uh, this episode will appear on the podcast. And if you haven't heard that, go to podcast.com CPE. Check out our past episodes. Those are all available for CPE on the earmark app. And if you're listening to the podcast and you haven't checked out our webinars, go to earmark CPE webinars or look for the events section in the earmark app and you can check those out as well. So, uh, Kyle, welcome to the show. Great to have you here.

Kyle Ashcraft, CPA: [00:02:56] Thanks so much, Blake. Uh, yeah. It was great to meet you in August in Chicago. And what I'm hoping we accomplish through, uh, today's talk is making you a little less scared of implementing AI into your everyday workflow. I think the more advanced AI becomes, we can either take two directions, you can continually veer away from it, and the more that comes out, you step farther and farther away from it. Or you can make it a goal to learn, let's say, one new tool a week and you start staying more connected to this AI movement because it's not going to go away. It's just going to become more integrated into everyday work culture. So the the better we can learn these dependable strategies, the more you can, uh, protect your career, the more you can innovate and ultimately save time. So, I'm very excited for us to talk through, uh, these, these topics here. Now, for those of you who don't know me, um, I became a CPA six years ago, and when I took my exams, I noticed a particular strength with them. Uh, I was able to pass all of them with 90 and above, one of them with 95. Another 98. And I wanted to combine my passion for teaching with, uh, my abilities with CPA exam. So five years ago I started Maxwell CPA review, and it is an online review course that students take to prepare for their exams. And then I also offer private coaching. And since then, I've been able to help uh, 1500 students pass their exams. And the main method that I teach accounting is simply through free YouTube content, intros to the auditing process, how bond accounting works, how lease accounting works. So I really enjoy posting this. This kind of content.

Blake Oliver: [00:04:58] You're auditing 101 video has over 100,000 views. I think that might be a record for an audit topic on YouTube.

Kyle Ashcraft, CPA: [00:05:03] Kyle. Well, anytime you can make audit slightly less boring is a it's a miracle.

Blake Oliver: [00:05:10] Oh, and before you continue, I just want to welcome our live viewers. Uh, if you have questions for Kyle, please enter them into the chat. I will be watching that throughout the webinar, and we'll we'll pick a few to answer at the end. Um, and also, you know, if you just want to, like, let us know what you're thinking. Um, we got quite a few of you on the, on the, on the webinar today, so, you know, let us know your thoughts. I'm curious to know what you think about what Kyle's going to show us. And also, um, what you're using AI for that may be beyond what's typical, like writing emails. All right, Kyle, back to you.

Kyle Ashcraft, CPA: [00:05:43] All right, so where I started transitioning a bit into experimenting more with AI. I was preparing for this accounting conference in Chicago, and I wanted to test the limits of what one person who is a non-coder can create. So I created this assessment platform where students take a baseline test. It identifies their weakest areas. It emails them follow up practice apps and videos. And then if a professor is implementing this, they can see all of the analytics for their entire class. And this was a four month project just again with no coding experience using cursor, which I'm going to show you all. Uh ChatGPT. Model oh three and then Claude Sonnet for the visual parts of it. And this really shows you that it's possible to not have any idea what the code itself is saying, but through clear communication and with patients, you can really accomplish things that would have been impossible just two years ago. So I think the possibilities are limitless when it comes to this, and I just want to get you going with it. I want to get you using cursor AI and vibe coding on a daily and weekly basis. So what really intrigued Blake and I about this topic is you have nearly 60% of employees using AI to some extent, but 17% on a daily basis. So there's this there's this gap. And what that tells me is they don't have dependable strategies for implementing AI. The way I picture accountants currently using AI is like opening up ChatGPT, throwing in a spreadsheet, and then giving it a prompt and seeing what it comes up with.

Kyle Ashcraft, CPA: [00:07:43] Sometimes, like a Hail Mary where you just want to see if it gives you an acceptable output. If not, then you put it to the side and don't use it again. In. There are a few issues with this. One is that it often takes multiple attempts to get the output that you want, because ChatGPT can't read our brains. We need to give it as much context as possible and show it as clear steps as possible. The other issue from a privacy standpoint is when you do this, your Excel document is going directly to OpenAI. Your prompt is going to them, and the prompt that they output to you is going to them as well. And OpenAI, the maker of ChatGPT, is particularly questionable with AI data retention. They're currently in a lawsuit with The New York Times, and they're required to permanently retain these logs. So we want to be as careful as possible. As accounting professionals. And that's why this carbon survey says 70% of accounting professionals are concerned about data security, which is, um, the most valid concern that we can have dealing with clients tax information with their Social Security numbers. So any kind of dependable AI strategy we have has to put data privacy at the forefront of any tool that we're going to use.

Blake Oliver: [00:09:13] And Kyle, what you highlighted is one of the issues I've had with using my spreadsheets in ChatGPT, which is you upload the spreadsheet, you give it a prompt, and then you get back a spreadsheet, but you don't necessarily know exactly what was changed in that workbook. Sure. And and that's been one of my hesitations is like, well, okay, am I going to rely on this? I don't know if other parts were changed. I'd have to go through and check everything to make sure it's right. Um, seems like like using AI directly in Excel solves both that and this privacy privacy question.

Kyle Ashcraft, CPA: [00:09:48] Right. We need an intermediary between the AI and the Excel file to better track the changes and to protect the flow of data going to these powerful AI companies.

Blake Oliver: [00:10:02] We don't have to send the whole workbook along with our prompt. Basically.

Kyle Ashcraft, CPA: [00:10:06] Right. We can tell it to only analyze a specific column of text. You can you can get very focused and uh, minute in terms of what you're focusing on.

Blake Oliver: [00:10:19] All right. So we're going to be using cursor. That's that's the tool we're going to be using today for for working with our Excel file.

Kyle Ashcraft, CPA: [00:10:26] Cursor is what we are going to download. It's going to be our main interaction. And what makes cursor so safe is that it's using something that are called scripts. A script is just like an automation. It's like a list of steps, like a a recipe for what the code should do in your Excel, and I'll show you an example of that in a moment. But picture it like the script is in the middle between the AI tool and your Excel, the AI tool cursor. It is writing this script based on your direction, but then the AI tool itself is not directly interacting with your Excel, so it doesn't have a direct trace of the Excel input or output data. And I find this particularly powerful, and I hope it can alleviate some fears with using AI. So a script is just like instructions to say, uh, let's see. Delete these first couple of header rows, find these columns and change them to a more consistent format. And the way it actually looks is like this. We don't need to know what this means at all. We just need to know what it is going to output for us and see if we like the output of it.

Blake Oliver: [00:11:50] It looks like computer code, like it's, uh. Yeah. Lines of code here. Um. So is this what is this? What cursor is going to generate for us?

Kyle Ashcraft, CPA: [00:12:02] Yes, it is going to create a script just like this for any task that we give it. And then once it creates this, then that set of code is going to interact with the Excel file.

Blake Oliver: [00:12:14] And what's amazing here that I'm seeing on the screen is that we've got like hundreds of lines of code in the script. And that would take a really long time. And a lot of expertise that we don't have. And cursor is basically going to write that script to accomplish that task for us.

Kyle Ashcraft, CPA: [00:12:29] Right. Even if we just had a basic understanding of coding, we couldn't get to this level without having years and years of experience only working in coding. It really puts us a leap forward for what we can accomplish. So we hear this term vibe coding, which is just coding without being a coder. This is what all of us accountants should strive to be. Um, you give the instructions in everyday language, and then cursor is going to create the code that you're going to use. It's going to create that, that script, uh, the coding language that it uses is Python. That's just the language that's used for these scripts. And anybody can be a coder. Um, you just have to be open to it, try it out, have a little bit of an experimental spirit with it, and you can get really, uh, incredible results. Now, I like this one.

Blake Oliver: [00:13:33] I'm a musician, Kyle, so I like this. And I like this, uh, metaphor here.

Kyle Ashcraft, CPA: [00:13:37] Uh. Excellent. Uh, we, you know, the way we think about AI and how we interact with it is so important. What do you expect out of AI? Because I think, um, an introductory AI user is going to picture AI like this black box where they throw something into it, they don't know what it's doing, they don't know what they expect, and then they just see whether they like what they get out of it. And then 1 or 2 attempts is going to determine their entire perception of AI. But instead, let's picture it like an orchestra and a conductor. You're the conductor. You are in control. You set the tempo. You set the vision of what you want to achieve. And it's the orchestra that's doing all of the hard work. It's one who's actually executing this piece of music. And then you as the conductor, you can stop it at any point and say, I think we're we're missing something here. I think we need to You take a step back and fix this. So the way we think about this is three steps. You're the one who's directing giving a prompt as clearly as possible. The AI is the one who is executing, doing the boring work for your recurring tasks that we don't want to do. And we shouldn't be wasting our our mental energy doing them. And then this key piece of reviewing the work. It can take multiple attempts. And I think of it just like if you're working with a new employee at your company and you give them one chance to balance a trial balance, they don't get it right the first time. You don't just throw them out of the company and, uh, never talk to them again. Instead, you go step by step. You see what they're missing? What context they're not understanding, and make these small improvements to help them to to advance along.

Blake Oliver: [00:15:51] Yeah. And that's the key step. That's the key step that you have to remember to do. You still have to review. And I think that was what, uh, what happened with uh, uh what was it one of the big four in Australia recently. Just had to refund a big government contract because they didn't review the work of their AI that invented citations.

Kyle Ashcraft, CPA: [00:16:12] Yeah. So they send an entire report to the government. The government clicks on the links for it and they don't exist. Yeah. Um, disastrous if you don't actually, uh, review. So you have.

Blake Oliver: [00:16:26] To I sorry to interrupt you on this, but, like, this is like a this is something that's fascinating to me about AI. Is that, like, there's this assumption that AI is going to, like, eliminate a lot of work. But what we find in reality is that it shifts the work from doing it, the it, the execution of it, like you said, to the reviewing of it. So that job is not going away. But now we are reviewing the output and providing feedback, like you said, like the conductor to the orchestra.

Kyle Ashcraft, CPA: [00:16:55] Sure. And so contrary to the idea of AI making you more lazy, you have to be more, um, more communicative and you have to pay more attention to what it's actually outputting.

Blake Oliver: [00:17:10] Mhm.

Kyle Ashcraft, CPA: [00:17:10] Um, and so you have to set your expectations. If I'm initially spending ten hours a week on something with Excel, uh, I could think, okay, well with AI, I only spend ten minutes on it. Don't try to get such a massive exponential cut in time. Instead, try for maybe two hours instead of ten hours, and you're spending that extra time reviewing to get better results. So in other words, just think if I can save 80% of the time instead of getting greedy with time spent and try to save 95%. I'm going to get really good results, but just don't expect it to completely automate the entire process. So we're going to start looking at three Excel examples that we're going to clean up. We're going to fix and analyze with cursor. So this is the first Excel which has columns of data that are in different formats. It has some blank rows. It has some that are missing reference documents. It's kind of hard to look at. And as we run it through cursor it's going to output something similar to this. We have some basic styling which makes Excel's that much easier to look at. Um, it is clean, it is consistent. And we're just going to do that with one simple prompt.

Blake Oliver: [00:18:36] Now this is a classic example. For some reason Exports of ERP system data or payroll system data, or point of sale data, like they're just not consistent with the formatting or like bank statements. So we're going to put all that into the correct columns so that we can work with it.

Kyle Ashcraft, CPA: [00:18:52] Sure. And in the second example, what if you have an export of your entire GL detail across your entire accounting system manually? Maybe you'd do a filter in Excel by account and then copy it into each tab. Very time consuming. But we are going to summarize each account's activity to then say for product sales code 4010 we have nine transactions with this amount of credits. So we'll have a summary. But then it's going to create a unique tab to show all of the detail but just by account number. So in terms of segmenting, summarizing Rising and separating data. Incredibly beneficial and time efficient.

Blake Oliver: [00:19:44] Mhm.

Kyle Ashcraft, CPA: [00:19:46] And then a third example is what if you're working with two different Excel documents. One is an export of just your bank statement and the other is an export of your cash GL detail. And you want to find what's missing between the two instead of just scrolling row by row. You want to see, uh, which items are in one but not the other, or vice versa, which items are matching. And so we're going to compare them. And it's going to give us a summary of how many are missing. And then a specific listing of the items that are missing. So this could also be very helpful in an auditing context when you're trying to find, um, uh, deficiencies, discrepancies between two different, uh, client files. So what do we need to do to get started with using cursor. The first thing is simply downloading the Python coding language. It's simply going to python.org slash downloads and clicking on download for Mac or PC. I'm just using a mac here. You need this because that's the language that it's using for all of these scripts and automations. So before you even download cursor, just go there, download this in five minutes.

Blake Oliver: [00:21:07] And that is python.org downloads for everyone following along at home. I've put the link into the chat there for you.

Kyle Ashcraft, CPA: [00:21:17] And then you download cursor cursor.com. It's a free download. You don't need to create a login until you download it and open it up for the first time. They have different tiers of pricing, just like all the AI systems. Start with the free platform. I'm using the $20 per month platform because I'm using it on a daily basis, but it's very, um, powerful even at the free, the free tier. So start out, start out with that. Now we get step more technical. But I want to show you some screenshots here to prevent you from feeling overwhelmed the first time that you're opening up cursor because you open it up and you give it your first prompt and it needs certain libraries of information. So just for example, for it to interact with Excel documents, it needs the pandas required package. So it's going to tell you I need to download this I need to download this. And at first you're going to have to manually approve for it to run those downloads. What I would suggest, as we see down here is you click this drop down button and it's going to simply allow you to choose run everything. So you just trust the platform. It's very reliable. And then anytime it needs a new required package or library, it's going to automatically download that.

Blake Oliver: [00:22:57] Okay. So we don't have to we don't have to initiate this when we start using it to work with Excel. It's going to ask us to install these packages.

Kyle Ashcraft, CPA: [00:23:05] Yes. It's just going to give you dialogs as you go throughout, uh, to run each installation. So you don't have to start out doing this. You can manually do it. And then simply at some point when you get tired of clicking run to often click on run Everything. So you'll just click in that bottom left box of the dialog, uh, for that. And this style of prompting, we're going to use its three tiered. We start out with the goal of what we want to achieve, kind of the vision as the conductor. Step two is what steps do you want it to take to achieve that goal? Like analyze each column, fix any inconsistent formatting, and add some basic styling and color. Three is the output. Do I want a new Excel from this? Do I want it to update my existing Excel and just put it into a new tab? Um, I suggest you always do a new Excel so you can track the changes that are occurring. Um, but you can ask for any, any kind of format that, that you want. But the more you break it up into goals, steps and output, it's going to understand your context and what you're trying to achieve much more clearly. Clearly. So now we're going to open up cursor and start working in it.

Blake Oliver: [00:24:34] Okay. So we've we've installed Python by going to Python.org downloads. We've installed cursor going to cursor.com. And uh and here we are in cursor.

Kyle Ashcraft, CPA: [00:24:48] We open it up. It wants us to open a project. A project is synonymous with folder because cursor works all within one folder on your computer. So if you want to introduce any Excel into cursor, it needs to be in whatever folder you are working with. So let me just show you, um, for example, here, just on my Mac, but I'm choosing this vibe coding folder. Um, so you choose the folder and after you choose it, don't change the folder name or any document names because it will become confused. So now we see this, uh, main user interface. And there are three different sections that we are looking at. The section on the right is what you're going to be using 99% of the time, because this is simply the chat command box, where we're going to be given all of the instructions for working with these files. So down here is where we're going to enter the goal of what we want to achieve the steps and then the output.

Blake Oliver: [00:26:07] And you're structuring this this prompt with goal. And you wrote goal colon. And you're going to write the goal. And then you've got steps and you've got output. So okay what are we what are we going to what are we going to do first.

Kyle Ashcraft, CPA: [00:26:20] So this middle section you don't use it all because it just gives you previews of the text. So it's just going to be a blank area here that we don't use. Okay. And then up in the corner left it mentions the files that you have within that specific folder.

Blake Oliver: [00:26:36] And so see we've got four files here. We've got messy data, income statement, GL detail bank statement and cash GL detail.

Kyle Ashcraft, CPA: [00:26:46] Now we're going to do the first example. Now cursor needs to know which file do you want to work with. And so you go over to that file on the left and you right click it and a dropdown appears. And you click on add file to cursor chat. So now you can see that it's popped up right here in the chat I see. So it understands that when I give it this prompt it's working with that Excel file. That's important because context is everything with these platforms. So let's give the first the first prompt. The goal will be simple clean up this Excel file, but it's probably wondering what does that mean to clean up the Excel file? First, identify any consistent formatting. Add basic color and style. And um, we could also say analyze each column to better understand it's, uh, it's it's format.

Blake Oliver: [00:28:01] And you've just put that under the steps section.

Kyle Ashcraft, CPA: [00:28:04] And then in the output I'm saying I want a new Excel document. Okay. So go clean up this Excel file. Steps. Identify any inconsistent formatting. Add basic color and style. Analyze each column to better understand its format and output. Is a new Excel simple. Now we are going to run that. It's going to take a little bit of time to process this.

Blake Oliver: [00:28:33] And we can see cursor is thinking it's giving us its its thought process. This is always fascinating to me now that we have these like thinking models that can think through problems and solve them. Like it's like watching somebody think out loud.

Kyle Ashcraft, CPA: [00:28:51] My favorite is when I give it instructions and it says, like trying to figure out what fix this column means. Yeah. And then you see it guessing the possibilities of what that might mean. Um, and so you get to see pretty quickly, quickly how clearly you're communicating.

Blake Oliver: [00:29:08] Yeah. And that's the biggest challenge I think is with, with prompting is, is you have to be you have to be a good communicator. You have to be able to write well to do this kind of work. And that can be a challenge. It's just like meaning what you say and saying what you mean. Mhm. Um it's critical. Okay. Now I see we've got, we've got um, we're getting a response. What's happening here.

Kyle Ashcraft, CPA: [00:29:35] So it has finished the task and it's telling us everything that it did. It's telling us we found some extra spaces and inconsistent formatting. There were 14 different date formats and it applied some professional styling to it. And ultimately it outputs this new Excel, a cleaned new Excel. So let's just take a look at that new Excel. And uh, let's just see, let's see uh, let's see what we think about it.

Blake Oliver: [00:30:16] And we've the the new file is called, uh, a messy data hyphen cleaned. And actually, can you show us what the original one looked like as well? Yeah.

Kyle Ashcraft, CPA: [00:30:29] So so we started out with a pretty messy Excel document here.

Blake Oliver: [00:30:37] Okay. So here we've got okay. This is I see this is the one we saw before where we have transactions. We've got empty rows. The the column headers are not in the first row. We've got like inconsistent formatting. Got it okay. Oh yeah. And then there's the the total at the bottom there. Okay. And now look we've got everything in a nice clean table.

Kyle Ashcraft, CPA: [00:31:03] It looks clean smooth some nice shading. Uh it's just easier to, to look at uh overall. And that was just with the simple prompt of cleaning it up and fixing any inconsistent formatting.

Blake Oliver: [00:31:18] Yeah.

Kyle Ashcraft, CPA: [00:31:19] And I think accountants in general, we don't want to tire ourselves with always looking at black and white excels with no formatting. Um, you're going to, um, work much more efficiently if you, you actually like what you're looking at. So just always add some basic styling and color. Um, it just helps, uh, drain on you a little bit, uh, a little bit less by having just some basic styling to these Excel documents.

Blake Oliver: [00:31:48] Mhm.

Kyle Ashcraft, CPA: [00:31:51] So there we have it, basic formatting. Um now what I would do is a side by side comparison to see how it is. Maybe it missed something. Um, I would add up this entire um column here and compare it to the other one. Um, we're going to jump onto the next example so we can get through more more examples. But you want to do a spot check. And the best thing with financial documents is add up the column amounts and then um, go from there. Or you may notice.

Blake Oliver: [00:32:26] I'm guessing we could also like prompt. In our prompt we could say like sum up the columns. Right. Could I mean could it, could it do its own like checksum.

Kyle Ashcraft, CPA: [00:32:35] That's great. So I already have the initial Excel added. And then I'm going to add in the new version. So I'm going to say can you verify that uh, the new Excel you created. Uh like accurately.

Blake Oliver: [00:32:55] Contains. Yeah. Like it's not missing any. Yeah. It doesn't missing any data from the. Yeah.

Kyle Ashcraft, CPA: [00:33:02] Perform a sum of the numbers column.

Blake Oliver: [00:33:06] So we're basically we've got both we've got both the, the original file and the new file. And we're asking cursor to compare the two. And while that's.

Kyle Ashcraft, CPA: [00:33:20] Loading.

Blake Oliver: [00:33:21] Oh go ahead Kyle.

Kyle Ashcraft, CPA: [00:33:21] Yeah I just want to show you around the platform a little bit. So up in the top right corner we have our settings. And if you scroll down a little bit then you have your privacy settings where you can choose that. You don't want any of this data to be used for their training, which I think is best for us to do. So you choose the privacy mode and then no training data is used, um, for their systems instead of doing shared data. So when you first open it up, I would I would set that and then you're, you're setting yourself up for, for a better privacy, uh, privacy policies.

Blake Oliver: [00:34:00] And Kyle, since we're on that topic, we have a question from Brett who asked, uh, when you load the project into cursor and it shows you the Excel files, does this AI platform not retain that client data? How is this different than uploading the Excel into ChatGPT?

Kyle Ashcraft, CPA: [00:34:18] Great question. And it does not retain this data because with this process that it just did what cursor was doing was it was creating this Python script, which is just Python code. It's offline. There's no record of this script. And cursor AI doesn't retain these scripts either, and it's just doing a full example of different kinds of formatting that it may have and the way it wants it to be cleaned up. So then if we noticed what it did. It said that it created this script. And then after that it runs that script and applies it to the Excel. So it's just this offline script that's getting applied to the Excel. So your Excel info is not going anywhere outside the platform. It's just being used for the coding. But there's no log of your actual Excel information.

Blake Oliver: [00:35:20] And. The data that is sent to. I think cursor uses Claude, right. Is that correct? The Claude you can.

Kyle Ashcraft, CPA: [00:35:31] You could just do auto and it chooses whichever one you want. I personally choose, uh, Claude for it. Claude 4.5 sonnet.

Blake Oliver: [00:35:39] So my understanding too, is that the data that is sent to Claude to be like. That's the that's the AI that powers cursor that goes through their API, which has a, like zero data retention policy.

Kyle Ashcraft, CPA: [00:35:56] Right.

Blake Oliver: [00:35:57] So it's different when you use the the chat interface for ChatGPT or for Claude. Um, logs are retained, files are retained for, for uh, many different purposes. But when it goes through the API, these companies have like terms of service and policies that they don't retain it or they only retain it in very specific situations. And so, um, you know, that's because they want large enterprises to be comfortable.

Kyle Ashcraft, CPA: [00:36:27] Right? So the information is better encrypted when they're using APIs. No data retention. The max data retention I've seen with cloud, for example, is seven days. But in general, uh, APIs are much more secure, much more encrypted. And, um, just just so far above the security of just logging into ChatGPT and interacting with the platform there. So, um, it looked through all of our transactions and it didn't create a new Excel, but it's saying all 20 transactions are present. Uh, all the amounts were correctly processed. It kept this sum of 19,000. And then I would just do a visual spot check. So I'm not completely depending on it, but just go into it and see see if I agree with that. So now let's go to the second example where we have an Excel that is a full GL detail. And it has so much. Going on that navigating through it would take too much time because we have service revenue in one row. We have employee benefits in another. It's all of the company's income statement GL accounts all within one Excel.

Blake Oliver: [00:37:48] And how many rows do we have here?

Kyle Ashcraft, CPA: [00:37:53] Looks like we have about 400 rows.

Blake Oliver: [00:37:56] Yeah okay. Got it. So yeah. So to basically you know we want to we want to analyze this GL detail and to do it in the past I mean I would do a pivot table. That's how I would I would do it right. But we're not going to have to go to that effort.

Kyle Ashcraft, CPA: [00:38:11] Yeah. It's going to be just a prompt. And then voila. So let's jump back into cursor. Let's give it this, uh, this new context, because right now, you see, it still thinks we're working with those example one items. So we need to exit out of those clean.

Blake Oliver: [00:38:30] Slate, those files. Okay.

Kyle Ashcraft, CPA: [00:38:33] Now I'm clicking on example two Excel right clicking and then add file to cursor chat. So now it has that context. So the goal is to um. Identify the transactions by their account number. Now here I'm going to tell it what column find that account number in. That'll give it just a little bit better description. So I'm going to say in column B. Your goal is to create a tab for each account number with all of those transactions. The steps a little bit more straightforward. Simply create one tab summarizing the totals for each account, and then create a tab for each specific account. Then the output is a new Excel file. To identify the transactions by their account number. That's the goal. The steps are to create one tab summarized and then create a tab for each specific account. And we want a new Excel file. So if you know more about models that you like to work with down here in the chat box. Uh, you can either click on this auto that I want it to choose the best model based on the task. Or you could go with GPT five, for example, OpenAI cloud.

Blake Oliver: [00:40:31] And.

Kyle Ashcraft, CPA: [00:40:31] A few other other options.

Blake Oliver: [00:40:34] This is a this is a great reason to use cursor to do this kind of work, instead of just Claude or ChatGPT, because with cursor you can choose a different model. So when a new model comes out, you don't have to like switch platforms, like, like like when when it came out, when GPT five came out, it was so good that I switched a lot of my work from Claude to ChatGPT. But if I was using cursor, I could just switch the model. I don't have to change my workflow.

Kyle Ashcraft, CPA: [00:41:07] Exactly. You're still working through cursor, and you just simply click that new model instead. And from the most recent releases. I've noticed them installing it into Claude. The next day, so there's really not a delay in terms of how quickly you can start using that model. Okay, so now it has told us everything. It's done. Um, I don't care so much about what it's told me. I want to see the actual results of what it created. So it created this new Excel, and, um, let's see what it looks like. So now we have a summary by account. We'll do a spot check of this one because it says product sales has nine transactions. And then in each tab we have those specific transactions.

Blake Oliver: [00:42:01] You've got a detail. You've got a summary summary of all the let's let's go back to the summary. So we've got yeah okay. Summary of account code account name. Number of transactions, total debits, total credits and the net amount. And we've got a grand total for transactions debits credits and the amount. And uh and then over on each tab we've got the detail for each account. So we just created a like a GL uh what would you call this report GL summary and detail report basically.

Kyle Ashcraft, CPA: [00:42:36] Yeah.

Blake Oliver: [00:42:36] From a GL export.

Kyle Ashcraft, CPA: [00:42:39] Instead of maybe going to each account in your accounting system and exporting the GL individually, just export all the accounts together and then run this through to separate them all by by by account. It's pretty powerful and I hope, I hope um, I hope your wheels are turning on how you can implement this into your, your workflow. Because in different accounting roles, whether it's private or public, um, private roles always have month end Then closing, um, on the, uh, public end. Um, clients always need, uh, amortization schedules for their notes. They always need depreciation schedules, even, for example, taking the client's information from last year and then creating a checklist for what forms you expect them to have this year. The goal is identifying your most reoccurring tasks and start there. You know, don't start with the most obscure task, but choose the lowest hanging fruit to see how you can, um, how you can implement this starting tomorrow and how you can already see the the impact from it.

Blake Oliver: [00:43:58] And I'm just picturing, um, I'm picturing our, our listeners, our viewers who work with some older ERP systems that don't have very customizable reporting Wording. And who are doing a lot of manual formatting of of data manual reporting in Excel. Like now you can automate that recurring task every month or every week. And you can save. Can you save your prompts into cursor and reuse them? Like how do you what do you where do you keep them? If you're doing a recurring task.

Kyle Ashcraft, CPA: [00:44:31] So, um, the the script will naturally. Um, okay. So when we're thinking through the prompts that we've given it so far, um. You can go back to that box and click on it, and then it you can copy it. I don't think they have a great memory feature to say like save this prompt as ABC. Um, so I think just when you find an effective prompt, you should categorize it and maybe just put it into a word document. But what is worth reusing is when you get a really good script that you've created, and then instead of having it create another new script, say, well, I used this one script to clean up my Excel. Now I'm working with a different Excel, so the only thing that's changed is just the document that I'm working with.

Blake Oliver: [00:45:26] Yeah, the underlying data.

Kyle Ashcraft, CPA: [00:45:29] Instead of creating something new every time I'm noticing my results the past few weeks, getting better by taking five seconds to just look back to what the most recent script I used was and then saying, take this and apply it to this new scenario because I know that it already works.

Blake Oliver: [00:45:46] Okay, so on the left hand side in that file explorer, we've got these files that end in.py. So that's the Python script that was created. And I think this is something worth highlighting. This is what's powerful about this is we can get consistency here because we can reuse the script that was already created, apply it to a new Excel file and get the same expected result without having to check everything over again, because we've already validated that the script produces a reliable output.

Kyle Ashcraft, CPA: [00:46:19] And it's very step by step. So it is a very repeatable process.

Blake Oliver: [00:46:24] Uh, the script.

Kyle Ashcraft, CPA: [00:46:25] Is.

Blake Oliver: [00:46:26] Deterministic. The script is the script is code. It's not like it's not like we're doing a statistical, um, thing every time. And so like that answers this question about like the risk of hallucinations that we had from Andrew. Andrew said, is there a risk of hallucinations in data using cursor? And I mean, the answer.

Kyle Ashcraft, CPA: [00:46:47] Is because you're not having you're not having a AI model interact with the Excel information. You're having this step by step, uh, script that's going to say do an auto sum of column B. And so it's taking Excel's functions, but it's just telling it what to do. So it's different than throwing yourself into ChatGPT and then having it do autosum, because this is a specific set of instructions.

Blake Oliver: [00:47:18] Yes. And you can reuse them.

Kyle Ashcraft, CPA: [00:47:21] And when you want to reuse them, just like with the Excel's, we were right clicking and then add file to cursor chat. You can click on that same. Script. And then now it's in your context and you can say apply this same script, but instead to um.

Blake Oliver: [00:47:41] And then point it to the new Excel file.

Kyle Ashcraft, CPA: [00:47:44] Yeah, I'm just going to say to this Excel. And then I could attach this.

Blake Oliver: [00:47:49] This is what I love. This is using AI for its highest and greatest purpose, which is to it's not to replace what we've been doing, which is create rules and and code and deterministic, um, automations. It's it's automating the creation of those rules and the script. And that's how we get the productivity bump with the reliability.

Kyle Ashcraft, CPA: [00:48:17] And the quicker we can get to analyzing the information, the more useful we are. Um, I just know in my first couple of years in an auditing firm, our first main task was importing the trial balance into the Caseware auditing software, and it didn't add any value. Um, maybe it taught us some basics about just balancing debits and credits, but it's something that had to be done every time. But the sooner it was into the system, then you could start actually running audit on your HR and your cash accounts. So the sooner we can get rid of those non-value adding activities and get into analyzing the the more valuable your work is going to be. And I think if you learn this in your accounting firm, whether you are staff level or manager level, uh, you can use this as staff and then show it to someone of how effective it is and start kind of sharing it around the people around you. Because I think these firms, they, um, maybe the manager and higher levels are saying how much they want to implement AI, but maybe they don't know how they want to actually implement it. But if you show them something you just did yesterday, then they might say, hey, let's do a class and have you share this with the rest of the the accountants. And so it can really be contagious for sharing these, uh, these technologies that we thankfully have available.

Blake Oliver: [00:49:50] I mean, the possibilities are really endless. Like, yeah, it's specifically an audit. Imagine, uh, using cursor to create a script to do sampling. Right. Here's the GL detail. And now let's, let's prompt it to do the sampling based on our. I don't know, materiality threshold and, uh, you know, sample size and like create a script to do that. How much time would that save versus doing it manually. I mean, yeah, really add up.

Kyle Ashcraft, CPA: [00:50:20] Or if you're analyzing a company's income statement and you want to identify any balance that's changed by more than 10% and X amount, uh, highlight that dollar amount. And then I'm going to go ask the comptroller about this. It's trying to take away that that selection work because, um, if we can randomize it we get more accurate formatting. Um, and you're just saving time for the more important tasks that will that's what we're ultimately trying to to reach. And that's why saving five hours with this, you can apply to five hours to client consultation and bringing real, real value. That's how you're going to stand out in your firm. That's how you're going to really, um, kind of insulate your career, uh, path and create a good moat around your. Your job position is by adopting these tools, trying them out, running some roadblocks. But then, um, like, reach out to me as you're installing this with roadblocks you run into or send screenshots of what, uh, what you may be struggling with. But I want to get you to to actually be be using this sometime over the next week. Try out your first task with it, see how it goes, and see. Just see how you're liking the the experience.

Blake Oliver: [00:51:36] And use that standard prompt format, which I love because it's very simple goal steps, output and try to anything you want to do, try to do that. Explain the goal clearly and then write down the steps or the process you want the you want cursor to follow, and then define your output. Do you want a new tab? Do you want a new workbook? What do you want it to do? And, uh. Yeah. And and try it with, uh, try it with like, a lot of different stuff. Like, that's, that's how I figure out what this, what it's capable of is, is not just following these examples, but also trying to apply it, like you said, Kyle, to the work I'm doing every day.

Kyle Ashcraft, CPA: [00:52:23] Sure. Apply it to your specific scenario, and you can start out with some information that is not private at all, maybe one of your your documents, and then start working with it. And patience is going to determine how effective of results you get from this. So there's just a direct correlation between my level of patience and the results that I have. Because the more patience I have, the more I'm going to follow up on that review step. Give it tiny pieces of feedback. It improves over time. I have this script that I like, I save that I reuse that every single week for my tasks and um, so just really, uh, practice your patience when it comes to, uh, to working on these, uh, to these, uh, these, these categories.

Blake Oliver: [00:53:11] And I like to say that I like to say that, uh, when I'm trying new tech like this, you know, 80% of what I do doesn't have a payoff, but then the 20% has a huge payoff. So don't get discouraged if your first few attempts fail.

Kyle Ashcraft, CPA: [00:53:27] Absolutely. And as you're attempting it, you're learning it. You're absorbing the technology more. Uh, yeah. The more you accept that that failure rate, the more the more you're not going to just want to run away from it or close your laptop. And so I'm going to be sending this PDF to everyone who's attending here, which is all of the steps that we've talked about where to go to download Python, how to install cursor and what to do your first time using it. So I'll email that PDF to everybody. And then if anybody has any questions about installing it, or maybe wants to brainstorm some ideas of how it could apply to their specific work situation, I love talking about these topics. It really, um, it really, since I'm so, so new to it and just thinking about it so often, it really encourages me and gives me ideas when I hear how other people want to use it in their in their day to day. So feel free to to email me to ask about, uh, anything you run into with it.

Blake Oliver: [00:54:27] We've got a few minutes for questions, and I've got a couple that, uh, have just come in. Citlali asks, how well does this work with Vlookups? So having a database excel sheet that it needs to retrieve to fill in data.

Kyle Ashcraft, CPA: [00:54:45] I'm thinking if you are basically the idea being using the script and cursor to actually perform the Vlookup, like I think you can't use the Vlookup formula and then have cursor interact with the Excel file. You would have to ask cursor to do the same thing as the Vlookup. I think it would be pretty similar to when we were going through the full GL detail and having it organized by specific column information, because after all the yeah.

Blake Oliver: [00:55:21] I assume you could have cursor. We could ask cursor to like add a column and create the Vlookup formula. And like we just describe what we want it to look up. Right.

Kyle Ashcraft, CPA: [00:55:31] Like that's that'd be great.

Blake Oliver: [00:55:33] Yeah. That's I mean so this is how I started with Excel with like ChatGPT is I would just ask it, you give it the file and you say, okay, what's the formula to accomplish this? And then it gives you the formula and you go and you plug it back into your original workbook and you see if it works. Yeah. Um, but but with cursor we could actually tell it to create the formula. And we describe what we want in plain English. Us.

Kyle Ashcraft, CPA: [00:55:58] And maybe that makes it more easily communicated to your team to show them not necessarily a script that you're using, but how it helps you to actually put in the formula.

Blake Oliver: [00:56:08] Well, and that's interesting. We don't have time to do it here, but that would be an interesting experiment. Would be like you had that source GL detail example. Right. Could we I bet we could do this. We could prompt cursor to create additional tabs that use Vlookups instead of just, you know, a script like we could actually have it do the thing where, you know, I have my source tab, and then I have my, uh, analysis tabs that are using formulas on the source tab. And that way my team could just I'm basically building something my team doesn't have to use cursor for. They can just go and copy paste the new data into the source tab, and everything flows into those other tabs. Like we could do that with cursor for sure, right?

Kyle Ashcraft, CPA: [00:56:55] Yeah. Because as soon as you have that new information each month, instead of rerunning the entire thing through the cursor script, you already have the formulas in Excel. So it'd be very, very repeatable, almost like a template that you're creating. Yeah, I haven't experimented with that, but I, I 100% think that that's doable with with cursor.

Blake Oliver: [00:57:13] The trick is just being able to explain it in plain English, to actually write it down. What you want it to do sometimes can be a challenge.

Kyle Ashcraft, CPA: [00:57:20] Right. Yeah, but try it and see what result you get. And then, um, sometimes even like, maybe I'm not communicating something to cursor. And so then I'll go to ChatGPT and I'll say this is what I'm trying to accomplish. Help me elaborate on this idea, and then I'll feed that into cursor. And it can just give much more detailed explanation compared to what I'm capable of communicating.

Blake Oliver: [00:57:45] Donna asked. Uh, just curious, what role is Python doing in this scenario we are working on? How do you link Python with cursor?

Kyle Ashcraft, CPA: [00:57:56] So when you download Python before you even open cursor, when cursor first sees your prompt and wants to create a script, it's going to look on your computer. Do you have Python installed? So I tried this on a computer this weekend that didn't have anything installed. And my first prompt it identified that I had installed. And so now it knows where it is and where to find this this Python language. So install it before so it knows to find it. And then Python is used in all of those automation scripts. It is solely use. So Python is the is the sole language that's being used for for all of those scripts.

Blake Oliver: [00:58:38] Yeah. Think of it as like the, the dictionary. Uh, it's the, it's the it it tells cursor. It gives cursor the ability to translate your prompts in English or any language. By the way, you don't have to use English. That's what's amazing about this too. You could write these in Spanish. You could write these in, like, Chinese. Any any language in the world. It can. It knows and it will translate that into the code. And that's the vibe coding thing we're talking about here, is everyone can be a developer now. Everyone can write code. Um.

Kyle Ashcraft, CPA: [00:59:14] You just have to experiment and have as much patience as possible. And even though we were working in Excel, you could easily throw in PDFs into this. And then it still uses Python to, uh, to scrape the information from them and to analyze them. Word documents, any document, uh, that you have, you can you can throw it in the cursor and start analyzing it.

Blake Oliver: [00:59:35] Thanks, everyone for joining us today. We are at the top of the hour. Kyle, it's been so great to see this, uh, content Kyle at Kyle at Maxwell cpa.com. And if you joined us live like Kyle said he's going to email you out a link to the instructions for how to set this up yourself. Um, if you are listening to the podcast, uh, we're going to we'll we should we'll put a link to those instructions in the show notes. Uh, and stay tuned for a brief message about how to earn continuing professional education for attending today. Kyle, great to talk to you. Thanks for, uh, thanks for showing me this.

Kyle Ashcraft, CPA: [01:00:12] Thanks, everybody. Try it out this week and see how it's working for you.

Blake Oliver: [01:00:16] Today's event has ended. To earn CPE, go to earmarked app or scan the QR code on the screen, then log in or create your free account. Search for the course by tapping on the magnifying glass icon on the home screen, and then entering the name of this event. It takes a few days to make the course, so if you're attending this event live, keep an eye out for an email letting you know when the course is available. Pass the quiz to earn your CPE, then tap the button to email yourself a certificate. Stay tuned for more events from earmark. Thanks for joining us and we hope to see you again soon.