Top Accounting Firms Are Obsessing Over Custom GPTs - Discover Their Secret

So today we're going to be talking about my favorite feature of ChatGPT. I mean, there are so many things I love about ChatGPT. Those of you that are regulars of.

My sessions, particularly if you're in the Value Pricing Academy, you'll know that I now teach a lot of stuff around ChatGPT because it's extraordinary what it can do to save time. In fact, I was this week doing a training session for my VPA level 3 mastery members next week. And I've been doing a lot of research. And what I'll be teaching next week is how to train ChatGPT in your writing style so you get much better written answers. But then I've also just built a GPT which then makes it so crazy fast to answer emails. In fact, I just did one today as a test. I sent an email to somebody and I did put in the PS this email is 100% written by AI. I can't claim any credit. It was a really well written email and it saved me loads of time having to write it myself. I just had to feed it a couple of bits of information and it did it. So today we'll talk about my favourite features of ChatGPT, which is called GPTs. Now, before we dive into that, as usual, I've got a few quick things to. To go through and firstly, just to let you. If you're new, if you're new to my live streams, I've got a lot of content to share with you, but I'll also be answering questions. So please put in the. Put in the comments wherever you are. I've got people here from LinkedIn, Facebook, YouTube, wherever you happen to be, put in the comments any questions you have. But please, if you have a question for me, remember, put two capital Q's in front because as you've perhaps seen already, there's a lot of stuff goes through the chat, particularly on YouTube. I do sometimes miss some of the comments. I'm happy for you to comment, I try to read them all. But if you want me to answer your questions, put two capital Q's in front. And at this point in time, somebody normally actually does that, just so I can demonstrate and show you what I mean by that. I will show that as soon as it comes through. And also, if you've got your smartphones to hand, grab them now, because I'm going to show you a few things, some QR codes of some resources you might want to grab. And firstly, one of the things that those of you who are regulars will know is that we are part of the amazing B1G1, where what we do is we make a contribution to a project somewhere around the world for people less fortunate than us and we make a contribution for every single person who is here live today. So the way that you do it is simply hold up your smartphone to that QR code or wait till Sarah puts the link into the chat and you'll see a screen that gives you four choices. Pick whichever one of those projects resonates with you the most. Click on the yellow button underneath and that will send a payment from us, not you from us. We will make a contribution on your behalf to that project. So please do that. Help to make the world a better place. Okay, next thing, next thing. If you are again new to this and you want to learn more about ChatGPT, we have an ebook you can grab. It was recently updated and if you want our ebook, then either use your smartphone and grab that QR code in the top left or if you don't want to do that, Sarah's going to put the link into the chat. I can see she's done that already for the B1 G1, which she will put the link into for, for the ebook as well. Soon. Just have to be patient because she has to do it across all the different social platforms. Next. And then the next thing to tell you about is if you want to learn more about, even more about chat, GPT and AI, we are running another event next week, another virtual summit. It's three hours of training and at the moment we're going to be giving away some complimentary tickets. So that's next week, next next Wednesday, same time as today, May the first, and actually starts a little bit earlier than today's session. And so if you. I'm just thinking, have I got the time? I think I might have mistyped the time. I think it's 10:45 Eastern Standard Time. It's 3:45 in the UK. I think I got that wrong. I think that's 10, 10:45. Apologies. Anyway, go to that QR code if you've got your smartphone or wait for Sarah to put the link and you can grab yourself a ticket for our three hours of training on chat GPT. Okay, great. And I can see the, the links are going in the chat and also. Thank you, Lynn. So Lynn says if Lynn shows how to ask a question, please put two capital Q's in front because that just jumps out from all the rest of the chat. But yeah, please keep the chat coming. We've got Let me just see where we got to. Is there anybody who's.

No, I think I've given all the shout outs. Right.

Keith says he uses ChatGPT to rewrite meeting notes to clarify things. Better. Better for non accountants. That's a great use of ChatGPT. Rewriting meeting notes. I like that one. Okay. I use it every single day, by the way. Every day because I've got into the happy audit and it saves me so much time. Hi there, Karen. Good to see you. Right, let's get into today's session. The secret behind custom GPTs for elite accountants. And what we are going to cover on today's session in the next 30, 40 minutes is I'm going to talk about actually first what some of the latest news in the world of AI because it's moving crazy fast. I'm also then going to talk about what are GPTs? Why are they so powerful? I'm going to do a quick demonstration for you. I'm going to share with you a number of other ideas of things that you can do with GPTs and then I'm also going to show you how you can and talk you through how to build GPTs. It's extraordinary.

So before we dive into GPT, just a quick bit of news as to what's happened in the last couple of weeks. Because this space is moving so crazy fast. In fact, what I've 2023 was a huge year. 2023 was a huge kind of race with all the major tech companies with large language models and AI. Some extraordinary developments over the course of 2023. But now we're in 2024 and I would say that as we're recording this in late April, there's been more changes and developments in the first few months of this year, 2024 than the whole of 2023. It's just accelerating the pace. It's extraordinary. I came across some research recently that showed that.

Three of the big four accounting firms, three of the big four accounting firms have subscribed to and use ChatGPT Enterprise if you're not sure what that is. It was brought out by OpenAI back end of 2023. It's their top level plan for really big companies. Probably not for you and I. You are better off with ChatGPT plus or ChatGPT Teams. We have a teams account enterprises for the really big fir. But interesting that three of the big four are obviously very serious about ChatGPT in our profession. Something else I came across that was quite cool recently, just a few days ago I heard that. Nothing, Nothing. That's a, that's a company that makes now phones. And what they've done is they've brought out some earbuds where they've integrated ChatGPT into their earbuds, which opens up all sorts of possibilities. Now, chances are you're probably not using nothing. However, what I can almost guarantee is that we will see that Apple will probably do something very similar. We know that Apple are going to do some amazing things with AI. They have their developer conferences in June I think, and I'm excited to see what they're going to be building into their products. Another announcement I don't know if you saw just recently. Mark Zuckerberg Facebook Meta they're really big in the AI space and they have recently launched their Llama 3 model which is, which they have now embedded into the search boxes of WhatsApp, of Instagram, of Facebook. So we're starting to see these large language models appearing all over the place. And I think the last time that we did a live stream it may have been Emily, I think, said that for the first time, another large language model knocked ChatGPT4 off the top of the pile as the best large language models. So ChatGPT4 has always been the best one across all the various tests. And a few weeks back, Anthropic launched Claude 3 and Claude 3 was getting better results than ChatGPT. And at the time the guess was, well, that's going to make ChatGPT respond. We know that they are going to at some point release chatgpt5 but what they have done in the last week is they've released, I think it's called ChatGPT 4 Turbo. But if you actually look into your ChatGPT, it labels it as dynamic. So if you're on a plus account, you may well now see that you've got a Choice between using 3.54 and now the brand new Dynamic, which is supposed to be much better still. So that's some of the new stuff happening. It's moving so crazy fast. There is always new stuff to tell you about, but let's get into today's main session which is GPTs. What and earth are they? Well, they were launched by OpenAI back in November of 2023 and they are my favorite feature. Now the downside is, is you do have to have a ChatGPT plus account at the moment they are not part of the free account, but if you're a regular to this, you'll know that. My view is I cannot see why you would not want to have a PLUS account because there's so many extraordinary things that you can do to save so much time with ChatGPT. And a Plus account is just $20 plus tax a month, which is tiny compared to the time it saves. So you will need a Plus account if you want to access GPTs. But I think when you've seen today's session, you'll realize that you need to get onto GPTs. What are they? Think of them as a. If you like a mini version of ChatGPT inside your ChatGPT, think of them as very specific, very specific tools that you feed with a prompt or a prompt or a set of instructions. And you can, and you can use them to do the same task over and over again, rather than having to recreate prompts and so on. GPTs allow you to do something extremely well and define exactly how it does it. And that'll make sense when I show you and walk you through a bit more.

So the big benefits of them is, firstly, you can give them a unique set of instructions. And there's an 8,000 character limit if you're aware of custom instructions. And custom instructions are awesome. Custom instructions are great, but you've only got two boxes each of 1500 characters. But you can think of a GPT as a set of custom instructions. And you can have different GPTs for different types of tasks. So you've got 8,000 characters, which is essentially where you can put quite an interesting prompt that can do standard tasks for you and get standard answers. That's the first thing that I love about them. The second thing about a GPT is they also have a knowledge base. In other words, you can upload files. So you can upload files like Excel spreadsheets, you can upload Word documents, you can upload pretty much any type of files that you want. And that GPT will always have access to that knowledge that you provide it. I'll give you some examples. In a bit, it'll become clear. And the third big thing I love about GPTs is they are shareable. Now, they're shareable in a number of different ways. One of the things you can do is you can share them publicly. In January 2024, January this year, OpenAI launched the GPT store. You can make your GPTs public. I am not excited by that in the slightest. And the reason for that is because they have made creating GPTs so easy and so easy to share that there are now tens of thousands of GPTs on the GPT store and 90% at least are a load of rubbish. Okay, so you could, you could spend forever browsing the GPT store for other people's GPTs to use, and you'll find there's only a small number that are any good. The real power, though, is in building your own GPTs to do things in exactly the way that you want them done. So, yes, there may be somebody else who's written a GPT on how to answer emails, but you're better off creating your own because then you can train it in your writing style, the format, the structure of the way that you do emails. So build your own GPTs. And they are shareable. Yes, you can put them on the store. I wouldn't bother. The key thing, though, is you can share them with other people, you can share them with your clients, you can share them with your team. And if you have a ChatGPT Teams account like we do, you can make your GPTs available to all team members. So, for example, I now have a chatgpt, sorry, GPT for answering emails, and I can share that across every member of my team. They can all access it and they can answer emails, saving huge amounts of time and answering in a consistent way, with a consistent voice. So they're the three things I love about GPTs. A unique set of instructions with up to 8,000 characters. You can feed it a knowledge base and they are shareable. So they are great for little mini systems within your business, for saving time and creating some consistency. They are best. They are best for specific tasks. You can make them very complicated, but they are best if they are really, really simple, specific tasks. And one of the reasons for that is if you create a GPT that's highly focused, it's going to be better than if it's too broad in what it does, it'll be better if it's more focused. But also because OpenAI released something, I think it was in January, that was really interesting, and that is that you can now, within a chat, you can call on different GPTs. So what you can do is, for example, I might show it later, I have a series of GPTs where the first GPT will analyze a website, let's say your website, and it'll identify the features of your product and the benefits, which is really powerful for building a marketing plan. But then based on that, you can then call in one that I've got called Ideal Client Profile, and it will then help you to, to create an ideal client profile for your business. And then when you've done that, you can pull in a third one called Client Persona, which will then take all that information of your features, your benefits, your ideal client, and it will create a Persona with their psychographics, demographics, their pain points, challenges, their wants, their fears, their aspirations, which is really powerful for marketing, which will help you then write greater marketing copy. And then it pulls in another GPT which then takes all that information and then will give you ideas on how you can improve your products, your services to better meet the needs of your ideal clients. It's really powerful stuff. So I love the fact that you can, you can create specific GPTs and then you can pull them in whenever you need them within a chat. So let's do a quick demonstration and I know some people are listening on the podcast. If you're on the podcast, you might want to go and catch up with the, with this on YouTube because some of this will be visual. So let me, let me. First of all, let's talk about, I'm going to show you a GPT that I did recently, really to help one of the VPA Academy members. So just a bit of background. One of the things I teach with value pricing is that if you're doing any form of cleanup work, let's take cleanup work. People don't know how to price that. So one of the things I teach is that you should absolutely value price cleanup work. You shouldn't time base bill, you should be value pricing it. But it's one of the harder things to value price because with a cleanup work, how do you know how much work is involved in cleaning up the books and records until you start doing the work? Which is why most people stick with time based billing. So the answer to that, how you solve that is you do a diagnostic first and you charge a fee for that diagnostic. In other words, you charge a fee to go in and do some work to analyze what's wrong.

In their system, their accounting records, and then you will use that information to then give them a proposal for the full cleanup work. And so essentially the way I teach it is that you are charging them a fee for the privilege of you doing the work to figure out how much to charge them and create the proposal. But what we want to do is make sure that that diagnostic itself is valuable. And so what you do is you carry out your diagnostic and at the end of it, as well as the proposal, you create a diagnostic report, a diagnostic report that creates value. And so one of our Value Pricing Academy members, Sarah Moss, asked recently a few months back in our Facebook group. She said, I've just started charging for diagnostics. It's going really well. But I feel a bit guilty that I'm not creating value. How can I make the diagnostic report more valuable? So I thought about that and thought, okay, let's think about how to make more value and then how can we use ChatGPT to make it so that you can create a powerful report in minutes? So let me show you very quickly. Let me just first of all show you a report I created in minutes as a result of a GPT and then I'll show you the GPT so you can see what it does. So if I go across to here, this is a diagnostic report where it sets out what was done, a summary of the findings, the problems that were found, the consequences, the solutions. It's a short report because I only fed it five problems. If you do cleanup work, you probably will find that you come up with dozens of problems in the, in the accounting records. I just pick five. It then goes on to give some a summary of recommendations, some technology and automation recommendations, suggested next steps and then in an appendix it gives much more detail on the consequences and the solutions for each of the problems found. It then creates another appendix with a risk assessment, a risk assessment for the business and then another action appendix with a more detailed action plan. The purpose of this is to add value to the client, to create something valuable, but at the same time make them realize that there's a lot of work needed to fix it and so they need you to do the full cleanup. So how do we create that in just a few minutes? Let me go across to chat GPT and if you have a plus account you will see that in the top you have all of these GPTs once you've started creating them with icons. So underneath or everybody will have their chat history. On the left hand side, this is my chat history, I use it a lot. But once you have a plus account you'll start adding above that specific GPTs. So let me pick the one I'm looking for. Diagnostic report. So I open up diagnostic report. Let me just close the menu, the sidebar. So what's different now is we can see it says in the top left diagnostic report as opposed to ChatGPT 3.5 or 4. And then it's got a few things to get started. I'm going to hit the let's get started here. And what it's going to do at this point it says great, it's going to tell me some shortcut commands that I can use and then it's going to ask for some information and I've created my GPT to give me this information because it's what I want to put in the report. So let me show you what I would then do is it's asking me for some basic information. I'm going to rather than you watch me type, I'm going to do control P so let's imagine I have a business name, a business called Lead Balloons. The client is called page and the review date is the 18th of April and it's for the year end 30th of December.

So if I just hit the send that message, what it will now do is it will say thanks for providing the information. Now could you tell me about the weaknesses or inconsistencies you found in the financial records for Lead Balloons? So obviously when you do a cleanup project you will find all these various problems. And let's imagine I've come across six problems and I'm going to feed the problems into here and this is the last piece of work that I need to do. So the problems are none of your accounts have ever been reconciled. This is by the way from a real cleanup project from one of our Value Price Academy members in Canada. So that's why it says this is a requirement of the CRA Amazon Holding Bank. There's a balance from journal entries that was recorded. This is not the best way to manage Amazon sales. Your bank account called product. This is not a valid bank account.

You probably see these when you do cleanup work, these sorts of things, mistakes that people make. Opening balance balance inquiry. There's a balance sitting in this account. We'll need to review and clear these entries. There shouldn't be a balance in this account. So all I've done is I've put in there the problems I found and I hit this button and what it's now going to do is it's going to go through that and it's going to create a list of those weaknesses and for each weaknesses it's going to explain the consequences and the solution. Why the consequences and solution? Because when you tell the client the consequences. So in this case, without proper reconciliation, the financial records may not accurately reflect the company's actual financial position. This can lead to discrepancies with the tax authorities, potential tax penalties and lack of confidence in financial statements. So by explaining the consequences to the client, they're going to realize the problems they've got or the potential problems and realize this needs to be fixed. Okay, so that's giving some value straight away and then we Say what you need to do to fix it. So reconcile all the accounts in QuickBooks Online. This involves matching transactions and it goes on and it's done that for each of the problems it's identified and come up with what the consequences are and what the solutions are. And so what you would then do at this point is grab that, copy and paste it into, let's say a Word document and you could then tidy up the formatting if you want, or you could put it into Canva and make it a beautiful PDF and that would make a great appendix because it's a detailed list of consequences. It then says type R if you're ready to create the diagnostic report, because that's just the appendix A. So I do.

And what it will now do is it will create the body of the report in the format that I want because I've specified this inside the GPT so you can see it's called an accounting diagnostic file review, a little bit of an introduction, a summary of the findings.

And then it's going to go and do some.

Recommendations, some and so on. So I'm going to stop at that point because I think you get the point. The GPT has been pre programmed to just go through a process of building up a diagnostic report and each step of the process I will then just copy and paste that into my Word document, review it of course, to make sure I'm happy with it and then that's it, job done. We have very quickly, probably an 8, 10, 15 page valuable diagnostic report.

So that's the next steps. And then it goes on to say, would you like me to export this to a Docx file? And so it guides me through the process. Right, let me leave that for now because what I want to talk about next is hope. That's giving you an idea of what a GPT can do. So what else could it do? What sort of GPTs could you build? And really the limit is your imagination. So some great GPTs that you may want to have is a GPT for answering emails in your writing style and in the way that you like to structure your emails where you just, in that GPT, you tell it, for example, how do you want to greet people? Is it hello, first name or hi, first name or Dear George, how do you sign off? How do you structure emails? What's your writing style like? We can, we can take all that information and put it inside GPT and then use it every time that we want to answer an email. And I did one of those. I've built One of those this morning for value pricing Academy members to make it really easy. Because one of the things I found is that that's the number one thing that people say they use ChatGPT for answering client emails. And so if we can automate that into a GPT to get better results so we don't have to do too much editing, we're going to save huge amounts of time. What else could we do with a GPT? Another great one would be, if you think about marketing, if you want to grow your accounting firm, then one of the things that we could do is content marketing. And as part of content marketing, that might involve you doing a blog post and so you could create a GPT. So we have one, we have a GPT for doing blog posts. The way that we do our blog post is we take a video from a YouTube channel, like for example, this one.

We take a video or podcast, we transcribe it, and then what we used to do is we'd edit that down manually into a typically a 500 word blog post. What we do now is we take the transcription of my video or my podcast and we put it into the GPT and it writes the blog post just like that. And then what it does is having written the blog post, it then based on that, write some promotional messages for Facebook posts, for exers, tweets for LinkedIn posts and so on. And my daughter does that. She built the GPT and what used to take her two or three hours to create a blog post and the social posts now takes a matter of minutes. It's extraordinary. What else could you use it for? You could use it for creating your ideal client profile. You could use it for creating a customer or client Persona, which is one of the most powerful things that you can do as part of your marketing strategy. Because the better you understand your clients in terms of what are their pain points or challenges, their wants, desires, their fears, what keeps them awake at night, what their aspirations are, their demographics, their psychographics and so on. The better you understand your clients, the better you can then create marketing messages, the better you can build the wording on your website and so on. Well, one of the great things is, is that ChatGPT can do that really, really well. It can give you a great starting point. So we have a GPT for doing that, for building a Persona. Those of you who are value pricing and you know, I talk about, we should give people choices, menu pricing, different packages, think of it as a bronze, silver, gold. If you've got a service and you're not sure.

How to create your packages, if you're not sure what to call your packages or what to put in your packages, then you could create a GPT that does exactly that. I've done that already. It's one I'll be sharing with Academy members fairly soon, where you can use that, you can feed it some basic information and it will do all of that for you. Another thing that you could do is you could use it for writing proposals. So if you want to structure your proposals in a particular format and layout, a bit like I showed you, the diagnostic report, then you could create a GPT that simplifies the process of creating proposals. The limit, there's no limit, it just goes on and on and on.

If I go back to my version, I've got dozens of these. I'm building them all the time. I've got one called Write Like Mark, which has been trained in my writing style. I've got one called Marketing Blueprint that builds a complete marketing strategy. I have Book Summarizer, which summarizes books for book reviews. So many. And these things just, they save so much time. Let me show you. Let me show you one more and then what we'll do is.

Yeah, I'll show you one more quickly and then I'll show you. How do you create these things? How do you create GPTs? So another one that's really powerful is I have a website here. It's the website of an accountant that in the uk, in London, if you're here, Hi there. They've got a great website, they're called makesworth Accountant. Great website, do some amazing stuff. They've got loads of awards and so on. So I took their website and one of the things they do is they have a particular service for business startups and it lists what they do as part of their business startup support. Now, one of the things I find is that one of the things that you might want to do is with all of your service offerings, you might want to ask yourself the question, how can I better communicate the value of my offering? How can I better communicate the value to the client of my business startup support? Because the better you communicate, the more people will buy and they will pay higher prices. And so we can use ChatGPT to help us with this process. So let me show you another GPT. It's called if I can Find it quickly. It's called Benefit Analyzer. And so what this does, it will analyze any service or products that you have and it will identify what are the features and then it'll help us turn those features into benefits that we can better use to build up the value. So I got a prompt here, it says analyze this webpage for a firm of accountants to help me identify the features and benefit of their service for business startups. And I've put the link to that webpage and so I hit this button here and what it's going to do now it's going to go and read that website and read about the business startup offering of the firm. It's summarizing it and it's.

Move on. And what it's going to do is it's now analyzing and it's going to create a list of the, the benefits.

It'S, it's using, it's analyzing because it's going off the website right now. Sometimes ChatGPT goes a little bit slower if you are using things like code interpreter or the website. But here you go, it's now building a table for me and, and it's of all the features and then it's creating benefits. So choosing the right business structure. I know that's on their website, I remember it's on here. That's a feature. People don't buy features, they buy benefits. So what's the benefit? Well, the benefit is it simplifies decision making for entrepreneurs, allowing them to focus on the other aspects of the business.

So I can add to that, I can change it, but I'm going to say I'm happy because then what it's going to do, if I remember right, I've not used it for a few weeks. What it's going to do now is start to create better benefit driven language that we can use in our marketing. So here we go, you can see that it's now.

It slowed down benefits. So choosing the right business structure. This feature simplifies the decision making process for entrepreneurs, allowing them to focus on other aspects of the business. That the ultimate benefit is reduced stress and clear path to growth. The problem solved is avoiding future complications related to business structure. And it's done that for all of the features it's picked. And so that's a great way of, we can use that firstly for creating some different words and language to put on our websites, in our proposals, in our pricing brochures. But also what we can do is because we've now done this exercise now ChatGPT understands my business better, my features, my benefits and we can now call on other GPTs and start to pull together some great marketing information. Okay, what I'm going to do now is I know there's going to be lots of questions coming in. I can see out the corner of my eye questions. Remember to put QQ in front if you have a question, otherwise I might miss it.

So let's talk a little bit about how do you build them? How do you create GPTs? So let's go back to ChatGPT. It's very easy. What you do is you'll see if you have the plus account or the Teams account. You'll see it says Explore GPTs. And when you click on Explore GPTs, what you will find when it opens up the page is you will find firstly all the GPTs on the GPT store like featured ones. I don't get excited by these for reasons I said there are tens of thousands of GPTs, most of them rubbish. But what it will show you is all of yours. And you can edit them very easily or you can just go to my GPTs in the top right and filter down to just your GPTs. But you also have in the top right a big green plus create button. So if I hit Create.

You then get into the the place where you can build your GPTs. And they've made this so easy. It's really, really nice. The functionality on the left hand side is where you build the GPT and on the right hand side you have a preview. And what that means is that once you've built it, you can then test it out without leaving that space. And if you notice it's not doing things right, you can go back and you can tweak the instructions. Now there are two ways to build your GPT. So on the left hand side we have two tabs called Create and Configure. Create is the easy way to build a GPT. You just simply have a natural language conversation with ChatGPT. So it says right at the start what would you like to make? And you just simply in the bottom left tell it what sort of GPT you want. I want to create a GPT to do a diagnostic report and then what it will do is it will ask a follow up question to get some more details and more information and you can have a conversation to tell it what you want to do. Meanwhile, what it does behind the scenes is it creates the prompt for you for doing that, which if you click on Configure you get to see in the instructions box it builds the prompt for you as you have that natural language conversation. Let's see how quick it's working. See if I can do it. I Want to create a diagnostic report for my clients to.

I'll just go with that. Okay. And so what it's going to do now, it'll say, yep, I'm working on that. It'll probably come up with some. It's updating the GPT and then it's going to come up perhaps with some more questions. And meanwhile behind the scenes, it's writing the prompt. Now that all sounds great. It's really easy to do it. The problem with that way of building your GPTs is that it tends to make a very, very oversimplified GPT.

If I go across the Configure tab, then we can see the prompt it's written is generate diagnostic reports for clients based on specific business metrics. This GPT should collect relevant information from users, ask clarifying questions and produce detailed reports in a user friendly format. That's not really very helpful. It's not very specific as to what we're going to do. So what we really want to do is my advice is never use the Create tab. It's too simplistic. Go into Configure and write your own set of instructions. So you need to understand a bit about prompt engineering and how to create a great prompt. But essentially what we're doing is creating a prompt here prompt that can be used over and over again. So what I'm going to do is I think it'll probably better if I go back to my. The one I've just shown. Let me just see if I can find it. Benefit analyzer.

And.

I don't want to hide from the sidebar. Just bear me a second. I always forget how to get to the edit bit.

Is it here, is it there?

Okay, I think I have to go to my GPTs. My GPTs. It's now opening it up. Yes, here we go. So these are all my GPTs that I've got. I just need to find where Benefit Analyzer is and then I can edit it and I can show you what's in it.

Actually let's go to Diagnostic Report. So if I hit the Edit button, I can now change my GPT. So diagnostic report, go into the figure tap, Configure. All you do to create your GPT is you give it a name. Diagnostic report. You can give it a description.

And then you create your instructions. Now if I expand this, you'll see that the instructions are quite detailed here. Here's the instructions that that process I went through to create the report, where it tells it what the purpose is, the mission, the instructions so I've got a series of steps to ask me for some information, ask me to supply the weaknesses, then to create the. The appendix A. Essentially that's the prompt to say I want a list of all of the consequences and the solutions. Then I wanted to compile the final report, then ask if I want a risk assessment, then ask me if I want to do an action plan. And then there's a few other things that I've asked it to do in terms of things like the writing style, the structure of that report and so on. So that's my prompt for doing it. You'll see it's a little bit more detailed, but that's because you can. Within the GPT you can be very specific about things like the steps it's going to take you through the format of the output, the format of your report, the writing style and so on. So that's where you create your instructions, you just build a prompt and then other things that you can do is you can have conversation starters. So these are little. When you first go into it, it gives you some suggested things, which is great if you're building a GPT that your team will use. So if somebody's using this for the first time and they don't know what a diagnostic report is, they can say what is a. The purpose of a diagnostic report? Or they could say, explain to me what you do and the steps we'll go through together. In this GPT, the one I always used is just let's get started and it will ask me that series of questions about what's the client name, what's the business name, etc. And then, and here's the other thing I said to you that's really powerful about GPTs. As well as you can have up to 8,000 characters, you can have a really detailed prompt and be very specific in. In your prompt. That's the first thing I said. The second thing I love about GPTs is you can upload documents. So what I've done with this particular one is when I was building this GPT, I created an example risk assessment, an example action plan, an example diagnostic report and example of the consequences and solutions. An example in the layout that I want, in the wording that I want, and I uploaded those four word documents and in the instructions it says somewhere, make sure you read your knowledge base and understand how I want to format the report. So that knowledge base is great for giving examples of things which you will then use to follow your structure. So that's really powerful. And then A couple of other things just to wrap up is you can choose whether you want to use web browsing Dali code interpreter if they're important in your GPT. You also have actions. I'm not going to talk about actions right now. Actions are where you can have APIs which then go and do other tasks. Think of Zapier. In fact, you can link it to Zapier, you can link your GPT to Zapier. That's more advanced. And also the consensus is that actions sound better than it really is. There's some more work required there and you have to understand about APIs. So I'm going to won't touch on, on that now. And then you hit the share button and you can choose whether to make your report your GPT available just to you, to your team, whether you want to share it through a URL and you can share your GPTs or whether you want to even put it on the GPT store.

So that's GPTs. Extremely, extremely powerful. I love them. I have fun creating them and they save so much, so much time. If you want to learn more about going deep on GPTs, then if you join the Value Pricing Academy. Mastery Level 3. One of the things I do every month is I go live for 90 minutes and I teach a curriculum called Leading Edge where I teach really detailed how you can master chat GPT and build prompts, build GPTs and I share those with VPA level three mastery firms. So check that out if you're interested. I might get Sarah to put a link into the link to those, but I'm gonna go to questions now. Let's see what questions we've got.

And.

I'm actually not sure I've got too many questions so at the moment. So I've got one, at least one, possibly two. Let me know what your questions are. Now's a great time to ask a question, so let me just go through a few people I think have asked questions without putting a QQ in front. I think.

I think I'm just checking.

So Justin says, does the ChatGPT has a software for diagnostic report? No, it doesn't. The diagnostic report is something that you, that I've, as I demonstrated earlier, you can build a GPT to do it. So your GPTs, you can build GPTs do anything, anything that you do over and over and over again. If you produce a particular type of report, then build a GPT for it. It's going to save you time. If you regularly answer client emails, that will be a great GPT.

We have a question from.

Clayton. Wick says, can chat GPT extract PDFs, such as invoices, and put them into a file on computer? Clayton, as long as you have the ChatGPT+ account, one of the things that you can do with the plus account is firstly, it can read and understand PDFs, so you can drag a PDF into either ChatGPT or into a GPT's knowledge base, and it will be able to read that and understand it. And also you can create PDFs. It can create PDFs now to do that. And not just PDFs, it can create PDFs, it can create PowerPoint files, it can create DOCX files, it can create CSV files, it can create Excel files. The way it does that is through what's called Code interpreter. So Code Interpreter is one of the things that you can choose whether to use in your GPTs. And that means that if, let's say, you've got an output, a response that you really like and you want that as a Word document, you can then say, please create a. Docx file of that response. And what it does, the way it does it with Code Interpreter, Code Interpreter is using Python code at the back end. So what it does is, is it will take your instruction. It'll go and go away and write some Python code to take that response and create a Word document or a PDF, and then it'll supply you with a link to that. I think I showed you. I think you may have seen one in one of the things I showed earlier.

So Code Interpreter is pretty good sometimes. One of the challenges is, well, firstly, you have to have a ChatGPT plus account to have Code interpreter. Secondly, because it's writing Python code, if it takes more than it used to be, 20 seconds, I'm not sure now, that gets to a timeout point where he'll come with an error message. Sometimes you have to be patient if you want it to create for you a PDF, a.docx file, an Excel spreadsheet. Sometimes it works really, really well. Sometimes, if lots of people are using ChatGPT, it can be a bit frustrating with the error messages. You just have to be patient sometimes.

Okay, I hope I answered that question right, Clayton. If not, just ask again. I've got loads and loads of comments and loads of people saying.

Saying, hello, I have someone from Botswana here. Hi there. Let's see if I pronounced that right. Moralagani. I hope that's right. Welcome to you. Right, we have a question here from accountants. Okay, if I create a GPT. Does OpenAI use my data to train its model? Okay, interesting question.

So if you are on a ChatGPT plus account, then any of your interactions, it will train the model.

Whether you are worried about that or not. I am not. I think it's a good thing that, that we all get a chance to, to influence how the model is being trained. They're not using your data, they're just simply using it to train the model. Now if, if you are creating a GPT, and this is important to understand.

If you're creating a GPT, then one of the things to be careful of is if you upload documents into the knowledge base and if they are proprietary information, just be careful. Now OpenAI has said that anything that you upload as knowledge into a GPT will not be used by them. They won't. They're not interested in the documents that you are uploading. So there's not an issue from that point of view. Where there does become an issue is if, if you turn on code interpreter and you decide to put your. If you decide to share your GPT outside the organization, either through a URL or by putting it on the store, it is very easy, very easy to hack a GPT and access the documents that have been uploaded. Very easy. Which is why they realized that early on when they brought out GPTs. And what they then did is when you look at the capabilities section, web browsing, Dall E for image generation and code interpreter, by default, code interpreter is turned off and there is a question mark icon next to it. So my advice is if you upload documents that contain proprietary information that you don't want anybody to access and you are going to share that GPT outside of your organization, then don't turn code interpreter off, which does mean people then can't do other things like create a Docx file from the output. So just be aware of that. It is possible to hack GPTs. I've done that several times. It's very easy. It's quite good fun. It's easy to go to the GPT store and look behind the hood of what people have put in their GPTs. Which is why one of the reasons I know most of them are actually rubbish, they're not very good at all. Okay, let's move on. Where do we get to. That was accountification. Carmen, where can one get stronger in prompt engineering? Carmen, great question. Prompt engineering is a really important skill. I often say that the most important skill we need as business owners is communication. Communication affects everything. Communication affects your leadership Skills, managing people, your sales process, your marketing. We have to be great at communicating, whether it's in writing, like at the word on our website or verbally. It's the number one skill that we can have in business. And prompt engineering is, it means it's how we communicate with AI. That's what it is. And given that we are now in an increasing AI world, prompt engineering is one of the most important skills that we can master. But it's not, it's, it's so new. There's not many places to learn. I have spent the last year learning prompt engineering. I'm fascinated by it because I love language, I love communication and I love the fact that we can then use language and to get consistent outcomes like diagnostic reports. So I teach prompt engineering at an advanced level inside Value pricing Academy Level 3 for mastery members. So if you want to go to the top level of prompt engineering, particularly stuff relevant to the profession, check out the VPA Mastery level. I do teach some basic prompt engineering in a ChatGPT course that we make available to our Value Pricing Academy Level one and two members. There's some great prompt engineering there as well. So if you want, you could go and check those out on some courses and training on prompt engineering.

Okay, next. Next question. We've got eight minutes left. Mark Johnson hi there. Can GPTs be linked to Gmail to allow for quick generation of frequently written short emails? I think the answer is possibly, but I don't know. You would do it by creating an action and they are advanced and you have to understand a little bit of basic programming and things like JSON language and you also have to understand APIs and I don't know whether or not you can create an API to link your GPT into Gmail. I just don't know and I don't know much about Google's stuff. So you can though what if you, if that's what you want, if you use Gmail and you wanted to create emails, then what you could use instead is you could use Google's Gemini. Gemini is a very powerful large language model. It's one of the better ones. And so you could create prompts for prompts for answering emails inside Gemini. And I get. And I think that links up. I think the thing about Gemini is it fits in with the whole Google space.

The one thing that makes ChatGPT different to all of the other large language models is the fact that it has these GPTs. And one of the things that OpenAI have said is that whilst everybody else, whether it's Google, whether it's meta, whether it's Amazon with their investment in anthropic, they're all going down the route of large language models, which are great. But OpenAI are increasingly going down the route of what's often termed as agents. So a GPT is sometimes referred to as an agent. An agent is a specific thing that you can call on which will then go and do something and the actions allow it to then do things beyond simply giving a a written response. And I think we're going to find over next 612 months this whole thing of GPTs, I think we're going to see them just get better and better and better. So we need to watch that this space. But if you specifically want to link to Gmail then you probably want to think about using Gemini. But I would stick personally I would stick with creating a GPT and then it doesn't take long to just cut and paste into an email. Okay, next.

Carmen says and we have six minutes left. Carmen says can code interpreter only be switched on during the creation of a GPT? I have tried creating a document from a chat with little success. Not using a custom GPT, just a normal chat. No, you can create, you can use code interpreter outside of a diagnostic, outside of a GPT. So let me just go back to. Let me just go back, let me start a brand new chat.

And when you start a brand new chat I'll just close the sidebar for a second.

If you've got the free account you can't use code interpreter because that's with 3.5. If you have got the ChatGPT plus account you have access to code interpreter as well as the browsing functionality as well as creating images. So you want to make sure you are using chat GPT4.

You will also see, you may see now that you've got dynamic. This is the new one that's better but interestingly the brand new one which is optimized for speed and intelligence. So giving you faster and better results doesn't seem to use or give you access to dalle browsing and analysis. So if you want to do at the moment, if you want to use code interpreter, you want to be in ChatGPT4 and so.

Let'S see if I can do something quickly.

Give me three ideas for a.

Blog post about.

Cash flow. This isn't a great prompt. I'm just hoping to demonstrate to Carmen the code interpreter. So I'm just getting some chat in there to start with. Hurry up. Hopefully you'll get there. I'm not bothered about the output. So with Code interpreter, one of the things that you one of the things you sometimes have to do is if you want to use Code interpreter or if you want to use this browsing functionality, you want to tell it that because it doesn't necessarily go away and use those tools. So in this case I'm going to say.

Create a Docx file for that response. And what it should do if it's behaving itself is it should now go into Code interpreter and start to write some Python code so we can see it says analyzing. If I hit the drop down box, it's writing Python code that will take that response and put it into a docx file. So that that's where it's using the code interpreter. Carmen, I hope that makes sense. If it's not doing it, then just put into your prompt.

Use your code interpreter functionality to do XYZ and if I click I can see download file. I've probably now got in my downloads. Yes, I have a document, a Word document for cash flow blog ideas. Okay, hope that helps.

Justin, does ChatGPT has software to conduct diagnostic report? I think I covered that earlier. No, no, that's the GPT I created to do that.

Okay. Jeff says can it be used for investing research and or technical analysis? Yes, it can. You can use it for investment vesting research. But of course you do have to be careful. Careful of all the usual disclaimers about past performance is no indication of future performance. And I think even ChatGPT gives you those disclaimers when you do it. Also though, if you're using it for any form of research, whether it's investment research, whether it's tax research or whatever, be aware that of the Knowledge Cutoff at the Knowledge Cutoff they keep updating it. So originally when it came out in ChatGPT3 came out in November of 2022 and the knowledge cutoff was I believe about September of 2021. A few months back they updated that so that with ChatGPT plus or Chat GPT4, the knowledge cutoff is up to April 2023 last April. So it's now got access to everything up to April 2023. But if you want to do any form of research and you want the latest information, you need to access the Internet. And So you need ChatGPT4 for that, the plus account to do that. And you can then use its browsing functionality. And what I always recommend that you do.

Start off by saying using your browsing capability.

Tell me.

Last night's football results. And it should be able to do that, I think, because it'll go off now and start analyzing the going through the Internet to find that. But sometimes you have to. There you go. It's doing it if you don't say using the browser. So if you want to, for example do some tax research and you want some information about tax rules around capital gains tax. For example, if you don't specify using browsing capability, it might just take what's its knowledge base up until April 2023, which may be okay, but with tax that may be out of date. So you really want to force it to use its browsing capability if you are doing research. Okay, so that's a really important tip there. I think we are about out of time. I know there's probably loads more questions. If you've enjoyed this, if you've got something of value it, please just hit the like button, the thumbs up, whatever, Also, I'd love it if you could just put in the chat box in your chat. Tell me what's the main thing you've taken away from this session? What's the main thing that you've learned? Because as we play out with some music soon, I will read all your comments. I'd love to know what's been most valuable. What have you learned?

There was a question from somebody about prompt engineering. That reminds me to say I did do a session on prompt engineering in my last livestream about three or four weeks ago. So it will look like this. If you go to my YouTube channel and go to the live stream section, you should see prompt engineering mastery, the six crucial levels. That's a great place to start with prompt engineering. And I talked through this particular diagram and gave you some of the key things to build into your prompts. So I hope you find that useful as well. Okay, I will hopefully see you. Remember today, this time next week, Wednesday, Emily and I are running the three hour ChatGPT and AI virtual summit. You can currently get a free ticket. We put the links in the chat earlier Sarah did and we'll also put a link in the description. Perhaps Sarah might post it one more time. So hopefully if you've enjoyed this, I will also see you next week next Wednesday where we'll go much deeper on some of these ChatGPT topics. I hope to see you then and that's also another time you've got an opportunity to ask me questions. If I've not asked your question to ask it today, I will answer it next Wednesday when you come along. Thank you so much for coming along and giving me some of your time. I hope it's been valuable. Let me know what's the most valuable thing. I'm going to read your comments as we play out with some music. Bye for now.

I wasted all my time I wasted all my nights. I wasted on someone who's indifferent. He didn't love me right he told me I was blind but I never really want wanted to listen.

You have been here since day one. See my ups and downs.

Never left my side once you st around.

I.

Knew it was you you have been here all along. You have always been the one. You, you knew it was you you have always been the one. That true.

You have been here all along.

I never saw the sign it didn't treat me right but I just couldn't let him go.

You waited for me he showed me what it was to love.

You have been here since day one.

I knew it was you you have been here all along. You have always been the one. Yeah who knew it was you? You have always been the one.

You.

You have been here since day one.

Never left my sight once I.

You.