Unlock AI-Powered Efficiency for Your Accounting Firm in Just 60 Minutes!

Just a couple of bits of latest news. I usually start off with what's new. There's actually so much new right now in the world of AI. But the, the one thing that I didn't notice the other day recently, it might be a week or so ago now. So this is old news in the world of AI. But I did see that Chat GPT did a really powerful upgrade recently to their chat GPT4 and it's extremely good. But. But most people don't even notice because they just built it. They just built it in. So it's already there. If you're using ChatGPT, it's there. And I have noticed I've been using it a ton over the last 10 days and I am noticed it's giving me some really, really good responses. And so I say that because one of the things we find is that there are some various league tables on which is the best large language model to use. And consistently ChatGPT is the best. There was a moment a few months ago when Claude Claude or sorry, Anthropic launched Claude Sonnet 3.5 and for a brief moment it knocked ChatGPT off the top spot. And also Gemini by Google has been doing some stuff recently which is pretty cool, which I announced last time we did a live stream. And I know Emily does this more. She did a great livestream a little while back where she looked at the different large language models, the previous pros and cons of each. But just very, very quickly for you if you're not sure because there are so many large language models now most people are using and have heard of ChatGPT, but pretty much every major company now tech company has their own large language model. You may have seen Elon Musk with X, they've released Grok, you've got Meta who have got their llama model. Generally ChatGPT is the best for most things. Sonnet 3.5, the paid for version of Claude that is arguably the best for written. If you want to create written communication, the way it writes is generally considered to be better. Personally I get great results out of ChatGPT. But if you want to write an ebook, for example, if you want to write written content, then a lot of people rave about Sonnet 3.5 at the moment by Anthropic. And Gemini's big thing at the moment is they have the biggest, I believe the biggest context window. They recently announced a 1 million context window which is kind of mind boggling really. That's 1 million tokens, which is you could Basically feed it several books. Now we have to be careful with these claims because they're not as good as they sound because what tends to happen, and this is really interesting. One of the things that's interesting, AI, these large language models are trained on human data, they're trained by humans and they have some strange quirks as humans. So for example, if I asked you to read a book and then afterwards, a few years later ask me to ask you to tell me about what you read in the book, what happens is the law of primacy and recency. We tend to remember best the things we learned first and the things we learned last, whereas the stuff in the middle we tend to forget. Now you'd have thought that large language models AI wouldn't suffer from that. Interestingly, they do. So you can have a huge context window. You can feed ChatGPT or Gemini a huge amount of information in it still tends to forget and ignore the middle stuff. Okay, which is really interesting. Anyway, that's kind of a few thoughts on where we are latest news. Let me just see. So Randall kind of agrees with me. Randall says that he's finding ChatGPT4O is much better. I'm finding extraordinary at the moment. And let me also just say hello to again my good friend Heinrich Grove from Cape Town. I was recently on his YouTube channel. He has an amazing YouTube channel, particularly if you're an accountant in South Africa and you particularly use things like Sage, then check out Heinrich's YouTube channel, Happy Heinrich, for you to put a link to it in the chat. You have my permission if you want to do that, Heinrich, because we have a few South Africans here. Okay, let's get into, let's get into the content for today. So we're going to look at how we can be more efficient, save time in our accounting firms.

Using prompts, using large language models. And more specifically, what we're going to cover in today's session is how can we really easily create what I call prompt templates. Okay. If you have no idea what that is, I'm going to explain what a prompt template is, why they are so powerful, why you'll want to build prompt templates, and then following on from that, why you'll then want to build a prompt library. How to build a prompt library. I'll talk about, for example, why adding variables to your prompts is a great way of saving huge amounts of time. I'm going to give you a bunch of ideas. I think I said in the write up for this, I'll give you seven Great prompt template ideas. I'll probably give you more than that. We'll see. There may be some bonus ones. We'll see how we do for time. And we'll also look at a very practical example of a particular prompt that illustrates some of these ideas. And it's a prompt for enhancing email templates. What I mean by that is let's say you have very often when you in the accounting world we have these software tools, this tech stack and some of this tech stack that we have will have built into it some template emails that go out to clients when certain things are triggered. An example I often give is Ignition. If you're using Ignition and you are ready to you've done the proposal, you've done the engagement letter, it will send that out with a covering email. And the email is a template and they're okay, they're okay. But what we really should do as part of our systems is enhance them, improve them. And if you've got lots of different things in your tech stack that sends out standard emails like your cloud accounting software does, if you use Ignition and so on, there's a great little prompt you can use to improve all of those templates in seconds with your writing style, your layout and so on. Okay. If you have, of course, any questions, please do ask questions in the chat and.

I will answer any questions as they come in later on. So please do ask your questions. I can see already there is a question from Randall, so I will get to that. Randall, yours will be the first I answer because the earlier you ask your question, I'll do them in order. If I run out of time, then I won't answer all the questions. So the sooner you ask your question the better.

Right. So let's start off with this. What is a prompt template? Well, there's really a couple of types. There's two types of prompt that we can use. And one prompt is just the short little prompt where we just want to do something fast. And sometimes we can do a simple one sentence prompt may well be sufficient. In some cases, you're just doing something quickly. Perhaps, for example, it may be that you, you're thinking of writing a blog post and you're not sure what to write it on. You could just quickly create a prompt that says I want to write a blog post. Give me 10 ideas for a great blog post for my audience. My audience are small businesses that own businesses in the medical sector, for example. And it won't give you 10 ideas for what that blog post could be aimed at. People in the medical industry, small business owners, and that will do a good result. There is nothing wrong necessarily with doing short prompts. They can be really useful. Of course, you do need to make sure you build in some of the key things, the key elements of great prompts. We can improve them once we get better at prompt engineering. But the point is we don't always have to write long prompts, so we can do the short prompts. The short prompts can be fine in many situations, most situations even. But there will be some things that you do where you do something frequently. You keep doing a task over and over again. And if you have a repetitive task that you do several times, then you can speed that task up by creating a prompt. And if you're going to do that several times, that repeated task, let's create a great prompt. Let's make it a prompt that's. That gives a very predictable and high quality outcome every single time. And let's make it a template so that that means we could just reuse it. So rather than having to keep creating a prompt, let's say, for example, that you do want to, let's say once a month you want to create five or six ideas for a great blog post every month. You want to write some blog posts and get them out over the course of the month. And you want to run that idea. Ask that question of ChatGPT on a monthly basis. Then you will get a much better list of ideas for blog posts. When you create a better prompt, when you have more clarity, more specificity, when you give it more guardrails. So it gives you exactly what you're looking for. In other words, when you, when you want to create a list of great topics for a blog post, you want to make sure that you tell it. What is your expertise in your firm? What's the topic you want to speak about? Who is your ideal client? What are the specific features of those clients, like their pain points, their challenges, that if you build those into the blog post, it's going to grab their attention and get the right people reading your blog post. So there's all these things we can do to build a better prompt, to get a better list of ideas, which is. And therefore it's worth spending the time to craft a better response prompt to get a better response. Think of it as like, it's like a little system. A prompt template is like a little system. You understand the power of systems. Systems are critical in any business. And in the accounting profession we kind of get that we are good at systems. We Are as accountants, as bookkeepers. We are, our brains are wired to be systematic in our thinking. We are very good with systems, whether they are audit checklists, tax return preparation checklists, standard letters, standard forms. We are very systematic in our thinking. So generally speaking, accounting firms are more systemized than many other small businesses, many which operate on a basis of just pure chaos. So we are good at systems. And so a prompt template. Just think of it as a little mini system designed to get you a predictable, better outcome in less time. Hopefully that makes sense. So we have these two types of prompts. Sometimes a quick prompt is fine, but the prompt template is the one that's going to be a mini system that we can use over and over again.

So you'll want to start to build up, if you haven't already, a library of these prompts. Why a library? Because like any system, you're going to want to access them really quickly. You've probably got, for many of your systems you may have workflow software, which with workflow software you can access, for example, your client onboarding system very easily. Well, you also want some way of accessing these little mini systems, these prompts, and that's where a prompt library comes in. And a prompt library is simply a way of organizing and giving you instant access to the prompts that you want. How could you do that? Well, a very low tech way, you could just simply have a word document or a folder with a series of word documents. Each document has the prompt, then you copy, you cut and paste it from the word document into ChatGPT. That's okay, you could do that. In fact, we started doing exactly that and then we found we got overwhelmed with prompts and it was too hard to find the ones we want. So tech is often a better way of doing that. And the people I know who do use tech technology for prompt libraries, the two most common tools people use are some people use. And you know what? The name has just gone out of my mind. I knew it would do that because I don't actually like it. If it comes back to me, I will tell you the one that I love and I know many people in the value pricing academy love and most accountants love because we tend to be visual people. We tend to be visual learners in the accounting profession. And so the one that I like is ChatGPT. So let me just show you.

I will come back to and remember what the other one is. I just don't like it much. But some people do. If somebody in the chat knows what I'm talking About. Just put it in there. It'll help me out. Okay. ChatGPT, this is our prompt library that we use. ChatGPT is great for this because it's. Because it's visual, you can see it all. And, and so what we can do is you can categorize your prompts so you can have columns. In ChatGPT, for example, we have columns or standard categories of prompts from ideation and learning, marketing, content creation, and so on and so forth. I'll give you some practical examples in a second and go deeper on this. But essentially, Trello is a great way of organizing your prompts. It's great because.

Thank you, Randall. Trello is one of them. It's the other one I can't remember. Randall.

Trello is great because we can create categories. We can categorize our prompts through the columns.

And then we can also add things like labels, which help us to categorize. And through labels, for example, if I want to do live demonstrations, I can filter on just the ones that I've tagged as live demonstrations. And if I just close that filter down, I've got dozens of prompts that I can use when I teach people prompts. So we can use labels to help us filter to different types of prompts and many, many more things we can do in Trello. This is not a Trello lesson. I just simply want to make the point that Trello is a great tool for building your prompt library. So let me give you some examples of. Let me give some examples of prompts that you could, that you could create as templates, because what I found, I've been teaching to the profession. Now, I've been teaching ChatGPT since, amazingly, March 2023. So just a few months after ChatGPT came out to the public in November 22nd. So wait, go back. Going back to March 23rd, I've been teaching this to an accounting profession, and I've been building up in my mind or building up all these different prompts. And the two things that I've. The two things I think are the most important things that you need to learn when it comes to getting the best out of any large language model is, number one is prompt engineering. That's basically how do we communicate with these AI tools? How do we structure our communication? How do we structure the prompts? What do we build into the prompts to get better answers? That's prompt engineering, which is a really important skill. And I think the second thing is simply just coming up with the ideas, because I remember when I First sat down with this.

Back in March 2023, my brain just went, wow, this is amazing. We could do so many things with this. And then I asked myself the question, okay, so what could I do? What could I do that would be really practical? That's going to save me time. And it took me quite a bit of time to think about it. It was through a lot of playing around with it and then figuring out. And that's the same with any, with any new tool, sometimes we're not sure how we can use it. And there's so much flexibility. It's a bit like when I first came across Trello. Trello's got so much flexibility, but I wasn't sure how to what I could use it for. There's a blank board. What do I do with it? And so over time, a lot of my learning has just been through interacting with my audience in the value pricing academy and then them asking me questions, Mark, how could I use. How can I use ChatGPT to do XYZ? And I would then go away and figure out a prompt that would be really, really useful for that.

And so over the years of using it, over the months, the year and a half of using it, I just have now loads of ideas for different ways of automating things that we do both in my own business, we have lots of prompts that we use internally. I have standard prompts I use every day. And also for your business, as an accountant, as a bookkeeper, once you start getting the ideas flowing, you'll think of lots of things. The more you use these tools, the more you'll start realizing there is so much stuff that, that you can use and then you can build prompt templates. So.

Let me give you some practical. I said I'd give you seven, but I'm going to do a few a bit more than that. So here are some different types of templates that you could use that would save you loads of time. So let me just. If you're listening on the podcast, then I am showing a few things on the screen. You can check them out on my YouTube channel. Okay, so one of the great things that you can do with ChatGPT is ideation and learning, brainstorming stuff and learning. So you could create a standard prompt, which is the latest tax and accounting regulations. And when I write prompts, they are quite long, and they're long for a reason, because I want to be super specific on exactly what I want it to do, what the output needs to look like, and so on. So you could create a, a standard prompt, which is called, let's say, the latest tax and accounting regulations. And then you might decide once a month, you might want to run that prompt in your firm once a month and get ChatGPT to go and do the research and so on, and come up with an output in the style that you want. And you might want to do that because perhaps, for example, in your accounting firm, you have a newsletter that goes out to clients and it would give you ideas for what to put in the newsletter of latest tax and accounting regulations. Or perhaps you might want to write a blog post around that. And so having a list of the five or six latest innovations could be really useful for writing a blog post. So that's a prompt that you might want to build in your firm. It might not be tax and accounting regulations, it could be something else, but basically a research prompt that you could run on a regular basis. And it'll go and do the research exactly how you want it to be done, to give the output in the way that you want it, and then you can use that in all sorts of ways.

Marketing. There are so many standard prompt templates we could have with marketing. For example.

You could create a standard prompt to create each and every one of your blog posts. Every time I do a blog post, get ChatGPT to do it. And content creation. So let's say, for example, you want to do content marketing, you regular content marketing. And what you want is ideas for what might be your social media content for the month. Having a prompt that you can just grab that will, you can just run this prompt and it will come up, let's say, on a monthly basis with your full content marketing calendar and exactly the layout you want. It is going to save you a huge amount of time. So lots of things that we can do in terms of standard template prompts for marketing and content creation. I was teaching people in the value pricing academy earlier in August, for example, a series of prompt templates that we could use every time we're looking to create an ebook, for example, to save hours of time. In fact, Sarah, who's here, Sarah Bridges, told me she did the training and within a day or so she wrote her first ebook following the series of prompts. The standard template prompts, in just a matter of.

An hour or so. I forget exactly how long it was, but not a lot of time. Then what about pricing? So if you want to get better, you know, I teach value pricing and one of the things I've done is I've built prompt templates for helping you with pricing. So one of the things that I teach is that with every one of your services, you should give the client a number of different options, packages, sometimes called menu, I call it menu pricing, sometimes called tiered pricing. So let's imagine that you want to do another service. Do you want to take one of your services and split it up into three options? Let's say it's a tax turn preparation service and you don't know how to do it. You could create a prompt that will take your product, your service and break it up into a series of three different options. And then you could run that prompt for every single service that you have and it will create a series of different packages for each one, saving you a huge amount of time. And then those of you who are then more forward thinking, thinking and perhaps a bit and thinking further ahead, you could also use that prompt to help your clients with pricing. In fact, I did that just the other day. I was helping a lady in here in Ericeira in Portugal. One of our friends who has a is launching a wine club. She does wine tastings and wine tours and she was launching a wine tech club. And we talked about why not think about different packages, different options. She said that's a great idea. And I ran the prompt with her business. I just put her website in, ran the prompt and it came up with three great packages. Business advisory. There is so many things if you do business advisory if you're helping your clients build more successful businesses. This is my favorite use case, by the way of ChatGPT. There is so many things that we can do to help our clients to create and add more value through advisory services. And there are loads of standard prompts that we could use. For example, I have a great prompt that works really well for doing industry research and benchmarking. And all you do is you add a couple of things at the end of it. I'm going to show you this later actually in more detail. We can add a couple of variables and then we just run the prompt and in seconds it will go and do the research for you and give you a report which you could then use as preparation for a meeting with either a brand new client or client that you're looking to do advisory with now. It'll give you ideas about their industry that you can use in the meeting. Obviously, as accountants we love financial data analysis and there are all sorts of standard prompt templates that we could build for doing that. So what about analyzing a monthly profit and loss account where you feed it the monthly profit and loss account, and then it will give you some analysis of that. And that could be whatever you want it to be. But your prompt could, for example, give some analysis on the trends, any potential areas of concern, advice on how to improve those trends, and so much more. We can create a standard prompt and every client, all we do is we drop in the P and L account and then we run the prompt and within about 60 seconds we've got a full analysis and report. What about, here's an idea for you that's really powerful. What if you've got clients, which I'm sure you have, who have got cloud accounting systems, but they don't know how to use them, they keep making mistakes. Let's say you have a client that uses, for example, QuickBooks Online and they keep making that. They want to do their own bookkeeping. So imagine you're an accounting firm rather than a bookkeeper. They're doing their own bookkeeping. They want to keep doing their bookkeeping, but they keep messing up how they enter credit card expenses. And so what you could do is with a simple prompt, in 60 seconds you could produce a report, a step by step user guide written specifically for that client on how they can enter the credit card expenses into QuickBooks Online. And then it'll also create for you a covering letter or email that you can send to the client. With this step by step guide, imagine being able to write training manuals specifically tailored for your clients where they're going to think, wow, you must have spent hours writing this. And it took you probably 60 seconds to run the prompt and then perhaps five minutes just to check you're happy with it and send it to the client. So that's a great prompt. An idea for a prompt template and then another one that we'll look at more in more detail is I mentioned earlier is we could then also, for example improve some of our systems, like improving email templates. So there are a whole bunch of things that we could templatize. We can create these little mini systems.

Inside our prompt library that we can use with ChatGPT or whatever your preferred large language model is. It may be that they are for Claude, Sonnet, Gemini, whatever.

So let's talk a little bit about what, what do I mean by a prompt template? What would you put into a prompt template? And I'll show you a couple and walk you through a couple of just examples that we could use. So.

Some of the things that you would build into a template, there are many. When I teach prompt engineering, there are many elements of things that we should put into prompts. So when you create a prompt, for example, ask Chat GPT to play a particular role, makes a big difference in the output because we know from research that large language models have got a default Persona, if you like. In ChatGPT, it's a helpful assistant, it's built into the the systems prompt. So what that means is that ChatGPT will always act as a helpful assistant. And the answer it gives the response is as a health assistant. And it's good, it's very good. But we can get a better response, a more detailed response, a more nuanced response if we ask it to play a particular role. So if, for example, we want ChatGPT to analyze a profit and loss account and give us some insights into, for example, potential errors, then if we ask ChatGPT to play the role of a financial accountant skilled in understanding financial data, what we found in research into prompt engineering is that it will then play that role and the response it will give will be the response you would expect from a financial accountant as opposed to a helpful assistant. It's very often a better response. So giving telling ChatGPT to play a particular role or giving it a Persona improves the response. And that's one of the many areas of things I teach in prompt engineering components that you put into your prompts. Another one might be if it's if you're writing an email, an ebook, if you're writing anything that's going to be consumed by somebody else. Another component is the writing style. Telling it what writing style to use can have a huge impact on the output.

So there is a number of different elements that we can build into our prompts to get better results, to get more consistent results. If we're thinking about templates, creating templates that we can use, then three particularly important elements, useful elements that we can use that you'll want to build into your prompts, are number one, within the prompt itself. Think about giving it a template. In other words, give it a framework for what you're looking for in terms of the specific response. If, for example, you want to create a prompt that's going to create a report for you, and this is a prompt that you might run several times, it might be a tax diagnostic report, for example. Then you might want to be in a particular style, you might want to have a particular heading at the start, you might want to break it up into some subheadings, so on. So giving a template for it to follow will give you a more consistent output. So I'll give you an example of a template in a short while. Templates are really powerful in prompts generally, but specifically in ones where you want to make the prompt itself a template that you're going to use over and over again. So always think about including a template wherever possible. The second thing that I always think about when I create prompt templates is examples. Examples are really, really powerful because, and ideally you want to have at least 3, at least 3, and 5 is the optimum. We've learned from research into prompt engineering that five examples is the best. Why do you want to give examples? And what I mean by that is let's say that you want to. Let's say you want to have a. Let's say you write regular blog posts. And if you understand marketing, you'll know that with any piece of content, the most important thing. The most important thing is always the headline. In an email, it's a subject line. In a blog post, it's the headline. The headline is the most important thing. Why? Because if the headline doesn't grab somebody's attention, they're not going to read your content. So when you write a blog post, for example, the headline is the most important thing. And so if you write blog posts in your firm and you're not, and you struggle sometimes with headlines, then create a prompt to do that. Make it a template prompt. You can run over and over again and give it five examples of headlines. Give it five examples of great headlines. Why? Well, large language models like CHAT GPT, they work based on probability. They basically just predict what is the next word in the sequence. They are not intelligent, okay, they are not intelligent. They simply just have a huge amount of data and they work on probabilities. What is the next most likely word in the sequence? It's as simple as that. And so they look for patterns. They look for patterns. And giving examples is powerful because they look for the patterns in those examples. With one example, they can't really see much of a pattern. Sometimes one example is okay. If you're giving an example of something quite long, this is an example of the report I want. But if you let's say headlines, I would always give at least five. Because if you give five great headlines, it's going to start to spot patterns in there as to what sort of things make a great headline. So examples are really important. They sometimes take time to come up with examples, which is why if you're doing a one off prompt, the first type of prompt, you might not do examples. You just want to do something quick and fast. But if you're going to create a template that's a mini system, then building examples is worth spending a bit of time doing that. And a little tip for you, a little tip that I find that I use all the time is what I do is when I create a prompt prompt, I may not put the examples in. The first time I create the whole of the prompt without the example, I then run the prompt. This is assuming I haven't got my own examples yet. I run the prompt and I see what comes out. I then would put that, let's say, in a Word document, and I would tweak it and improve it until I'm happy with it, the output. And then if I'm happy, that output, and that's exactly what I want, I will then cut and paste that in to the prompt and that's then the example. And so I can then run that prompt a second time, knowing that it's probably going to be a better output. I might still have to have to do a few tweaks, but then the second time I can use it as a second example. And that's how I build up examples. Often they are, if I haven't got an example already, I can use the prompt to create the example and then just tweak it until I'm happy with it. Number three, the third element of the three, I want to teach you the third element, which is my favorite. This is my favorite element of a prompt when I'm creating templates. It's variables. It's variables because let's say, for example, you want to create a research report, an industry research report in advance of every meeting with a brand new client.

Which you can use to demonstrate to that potential client that you understand their industry. You've done some research, you've done some thinking. You could create an industry research report.

But without variables. Every time you run it, every time within that prompt you mention the industry. Perhaps tomorrow you've got a client coming in that's in the electrical contracting industry and you write this great prompt and throughout you say, write this report. So it's for an electrical contracting business. Go and look at this website for an electrical contracting business. And you may several times in the prompt mention an electrical contracting business. And then a week later you've got a meeting with a florist shop and you think, well, that's a great prompt, I want to use that again. But now with that prompt, you've got to go through it and you've got to edit every instance of electrical contracting with florist shop. That's time consuming. So we can save time with variables. Let me just show you very quickly what I mean by that and talk you through with one of those prompts I mentioned. So I mentioned the industry benchmarking. And so if we go into, I'm going to mention I've got my Trello board here. And by the way, another thing I love about Trello, this is more of an advanced feature. I know people in the valueprice academy understand this because I teach prompt engineering. Markdown is really powerful. Markdown is great because we can emphasize certain things, create greater clarity in our prompts. And so I would always use markdown inside chat GPT. And a great way of doing that if you use Trello is if you were to hit the edit button.

There is a button where you can view the prompt in markdown language. And when you view it in Markdown, you have a little button called copy. And all you do is you hit copy, go into chatgpt and hit paste. It's as simple as that. It couldn't be simpler. In fact, I'm going to get. I will show you in a second just how easy it is, but let me just go back to the prompt. So if you want to do industry research and benchmarking, then you could create a great prompt for doing exactly that. And then in this particular case, let me mention, I have in this case both a template and variables. There's no examples in this one, but there are variables and templates. So the template is here. The report should include the following sections and I've created what I want my industry research report to look like. I want to start with an executive summary as a main heading. I then want the current market trends. I want competitor analysis. I want the latest innovations in the industry. I want the regulatory changes. That's how I want my report into the industry to look like. So it'll always do that every time I run this prompt, it'll just run it with a report in that particular way. So that is what I mean by a template. Really powerful in terms of the variables. You will see straight away that it says latest innovations in. Now I could put electrical contracting, but I then have to change that every time I run it. So I've got latest innovations in and in curly brackets it says industry sector. And that's because it's a variable and throughout the prompt that variable is used. So for example, using your browsing capability, research the industry sector. So I could put in floristry, business, electrical Contracting and it will just put that into the prompt. Likewise, with this particular prompt, it will. I've asked it to analyze the client's website. So with variables. The way that I do it the way that I do it with variables in my prompts to make it really crazy easy is I always have a section at the end called variables and I put the variables at the end. And so I can have clients website equals. And then you can just put a URL to a website. So you've got a, you've got a client coming in next week, a potential client, you want to do some research. All you do is you cut and paste the prompt, you put in a website, the website of this prospective client, and then under the second variable of industry sector, you put in the sector like electrical tracting. That's it. That's the only things you have to enter. You hit the send button in ChatGPT and it's going to create this report for you in seconds. It makes it so easy to do. Let me show you another one.

And.

We'Ll work this one the whole way through. So I mentioned earlier, one of the things you could do is create better email templates. This is particularly useful if you've got email templates built into some of your tech stack. And those emails that are built into the software tools are okay, but you might want to make them better. So let me.

Run what I'm going to do for this one.

I'm going to do it how I would do it in ChatGPT, I would hit the edit button, grab the markdown, hit the copy button, go into chat GPT, start a new chat, hit Control v, Paste, and then I've got the prompt in there. And in this case there are no variables. Instead, in this case it's the original email. And I've actually put an email in the example for when I share this. So this particular original email, it happens to come from Ignition. Great software. Nothing, I'm not saying anything bad about Ignition, but like many with these software tools, when they have built into them an email that goes out automatically, in this case it's to send out the engagement letter, then it's usually fairly bland. Bland because it has to be, because every accounting firm will have a different style of how they want to do their emails. So it's not the fault of the tech stack that they put in the tech stack. They're just giving you a system, a sample email, and it's up to you to improve that. So all you do with this, the variable, if you like, is the email template that's built into the tech stack and the rest of the prompt, if I go back to it, it'll say, in this case, you're a professional email writer. Your task is to enhance an email from a professional firm of accountants, making it more friendly, conversational and exciting for the client. There's some action steps of what to do, which is to review the original email, then rewrite the original email. We have some goals and constraints, outcome expectations, and then examples of required response. So I said earlier, examples are really powerful. So if you've got some examples of the sorts of way you do your emails. So you might start off with your emails with high client names. It could be hello, client name, it could be good morning, client name, whatever it is. So you would just put into the template or the example, perhaps you start with something friendly in your emails, like I hope you're doing well. So I've got an example I've put in of an email that I would send out to people and you could put up. You could put many examples into your template. But anyway, let me just hit the button and show you how quick this is. So you saw all I did is I went to the prompt library, copied and paste copied out the. The template, the. The template prompt paste into chatgpt. You just have to then put in your. The, in this case the email that you want to improve, hit the send button. And as you know, large language models work incredibly fast. So rather than saying this is the Ignition one, please read through and sign the agreement we discussed together. Let me know if you have any questions. Otherwise, really looking forward to working with you. Nothing wrong with that. But what about instead if we change that in Ignition to say subject? Looking forward to our exciting collaboration. Hi John, I hope this email finds you well. I'm thrilled to share the agreement we discussed and I'm eager to get started on this exciting journey together. Please take a moment to review it at your convenience. If you have any questions or need further clarification, don't hesitate to reach out. I'm here to help. Your comfort and understanding are key to our service. I'm genuinely excited about the opportunity to work with you and look forward to all the great things we'll achieve together. Best regards. Now, you might think that's not my writing style, that's fine because that's based on the email example I put in the template. Personally, I'd be happy with that, but all you need to do is give an example in the template of how you might your style of Writing and structure in an email, you might want to start with Good morning or dear or hello, whatever it happens to be. You just make your own example up that you're happy with. And then you can run that prompt as quick as I've done it. And in seconds you get a longer what looks like a much nicer email than the one that is the template email provided by the software companies.

So that's improving email templates as an example of how we can templicize, if that's a word templar size our.

So the key things, what have we covered in this? We've talked about the two different types of prompts. There's the simple prompt we can use, but then we can have prompts that we spend a bit of time on because they are like mini systems. We can use them over and over again. And then what we need is some way of organizing those prompts. I call it a prompt library. I love Trello. We use Trello in our business. I think it's the best way to organize your prompt because it's very visual. The other one which I have remembered now, it just pops back into my head is notion. Notion. I know some people love notion and I know some people who use notion as their prompt library. It's very powerful. I don't find it as visual. I find it always looks confusing. But you might love notion. And if you do, if you have notion already, then that could be your prompt library. So I think they're the two best options, Notion or Trello. And we've talked through some examples of different types of things that you could have as prompts, like analyzing financial statements, things like industry research. And we've looked at those key elements of a prompt template, having a template within the prompt for it to follow, examples and variables. And we've looked at some email templates. I hope you have found that valuable. So we have questions. I have lots of questions. We have 30 minutes left for questions. Let me answer as many as I possibly can. If you found this valuable so far, if there's something of value that you've taken away from this, then please do click the thumbs up the like button. I would really appreciate that. It helps me to know whether you're enjoying this content and I can see some people are doing that already. So thank you for the thumbs up. Let's go back and see what questions. I did forget to say earlier. I really, really, I apologize for this. But the regulars know this, that when you ask a question, because there are lots of comments as well in the Chat that I have like Galile Baiini, if that's pronounced right, from South Africa. Lots of comments. So sometimes it's hard to find the question. So I always say to people, put two Q Q's in front like Randall. Randall's a regular. He knows that. He knows the drill. Thank you Randall. Because it helps me to spot the questions if you put two capital Q's in front of so have you found ChatGPT to hiccup lately? I'll start the response, then freeze and then say to retry. Yes, all language models have hiccups like all tech or all technology from time to time. ChatGPT sometimes has outages not that common. It is quite rare, but sometimes it will just go down. I do find that if it gets stuck, if it doesn't do the response, some of the things that you can do is hit the if you don't finish off the response you can hit the regenerate and it'll do it again. If you're not sure what that is at the bottom of any.

If a response isn't finished, there's one of the four icons. Underneath the prompt is the two sets of swirling arrows which means regenerate and that will just do it again. Useful if it's not finished off what you wanted or you could. Another thing I sometimes do is just refresh the browser. If you. If you're using ChatGPT in a browser that sometimes fixes it. Other things that I found Randall that do get temperamental, I find it very useful. If let's say I'm doing industry research, then one of the variables that I mentioned earlier is the client's website. So all you need to do is if you put in the client's website and the prompt says analyze this website and then tell me about their business, that's a very cut down version of that prompt. But essentially analyze this website. Most of the time it does a really good job of going away reading the website and then that gives it the context about the business and then it will give you a report. Sometimes it will say I can't read the website either, it's having a bad day.

Sometimes I found there are certain websites it just can't read and if that's the case, if it happens the first time I'll just try it again, I'll just regenerate and sometimes I'll just put in a follow up prompt that's just the website and I might say try again. Sometimes saying things like try harder next time sometimes works because strangely AI has these Strange human quirks.

So reading websites can be sometimes frustrating and if you, if it really is refusing on a particular website, the other thing you can do is take the website itself, grab a PDF version of it or screenshots and feed it that way. Another thing that is can be frustrating with ChatGPT is, and this is, this is particularly important if you are doing financial analysis. Sometimes I find that if you feed it and upload an Excel spreadsheet to analyze that, then force it to use code interpreter which is very powerful. But also if the Excel spreadsheet isn't anal, sorry, isn't formatted in the way it's expecting, it will then go through a process where it'll come back and say the sales isn't formatted right. Leave it with me, I'm going to fix that. You wait 20 seconds. Okay, I'll fix that. But it looks like this isn't right. So it's going through and tidying up and cleaning up the Excel spreadsheet, which a can take time, sometimes can time out and sometimes it then just doesn't give you the full answer. I had that earlier with one. So instead what I did is I turned the Excel spreadsheet into a PDF and fed it a PDF. Perfect answer straight away. So it can be temperamental with code interpreters, sometimes with Excel spreadsheets, temperamental with reading websites. So sometimes feeding a PDF or something can work better I find at the moment. Okay, thank you for that question, Randall. It was a great question. Clive Stewart what does Clive say? What is the difference between a prompt template and a custom GPT? Clive, great question. A custom GPT is taking it to the next level still. And I love GPTs. I talk a lot about them. In the Value Pricing Academy mastery level we have lots of GPTs. So a GPT.

The benefit of a GPT over a prompt template I would say is threefold. Number one is accessibility. In other words, your GPTs are available to you instantly inside chat GPT. So on the, the top left sidebar above your, your your chat history, you will see all of the GPTs that you have pinned to your sidebar. I've got, as you can see, I use GPTs a lot. I have one called I use quite a lot is Email Responder. Email Responder is a GPT that when I run it it just asks me two questions. Give me the email. What in summary do you want to say? And then it writes a perfect email response. I don't write emails now. ChatGPT writes my email and I use email Responder, it's a GPT. So first benefit, Clive, is you can pin the GPT to your left hand menu bar. Second benefit of a GPT is I think they work really well. If you want to do a multi step process. Now you can do multi steps inside a prompt template. Inside a prompt you can say I want to do this first, then this, then this. If you have more than three or four steps in a prompt, it sometimes trips over itself, it doesn't do it properly. And so a GPT for multiple step processes tends to work better, particularly if you want it to iterate with you. So in other words, do step one and ask me this question. When I give you the answer in step two, then do this. So the email one, for example, in my GPT step one is it gives me the instructions and says to me basically please give me the email that you want to respond to. So I give it the email. Once I then hit the send button, it knows that step two is okay. What are some of the things you'd like to put in this response? And so I give it some bullet points of what I want to say and then I hit send and then it does the email so it breaks it up into an interactive series of steps. Better if it's a GPT. And the third thing, Clive, about a GPT is you can upload a knowledge base. You can put it up to, I think it's 10, it could be 20. Is it 10 documents in there? So that's really useful if you want a GPT that's pre trained in some proprietary knowledge. Now you can upload documents into a prompt, but you have to every single time with a GPT if you, you can build that in. So for example, I have, I trained ChatGPT in my writing style. I have on my desktop a document called Writing Style Guide. So if ever I use a prompt that I want it to write something, I will say to it, write in my writing style using the attached writing style guide. And that means I have to then attach and physically drag the writing style guide into the prompt. Whereas with say my email responder, the writing style guide is built into the knowledge base automatically its instructions are read. Make sure you understand my writing style using the writing style guide in your knowledge base. It's in the instructions already. I don't have to repeat it, I don't have to drag it in. So those are the three instances Clive, where a custom GPT is even better still.

So next question, next question. Randall, if you use chat to me to write a Book should you. One, do a chapter at a time. Two, use one chapter to do the book. Randall, you're in the vpa, so make sure you check out my.

Training this month. If you haven't already, catch the recording of the August session, the Marketing Fundamentals session, where I'll walk you through exactly how to use a series of eight prompts I've written for members. You just run the prompts and you'll have your ebook done in minutes like Sarah did. But to answer the question.

Yes, I strongly recommend that you could you use ChatGPT? ChatGPT is great to give you the outline of your ebook. Now to do that you need to give it context and so on. But once it then gives you the outline for an ebook, I always then recommend that you write it one chapter at a time. Why? It's partly because the context window. If you say now write the whole ebook, it's going to be quite short. It'll be probably way too short and too brief. Whereas if you write one chapter at a time, it'll give you. And then let's say it's five chapters. You then collate those five chapters. You will get a better, more in depth ebook. So hope that. But you can do it all in one chat. So the way I taught it, Randall, is I've got, I've actually in this case got a series of eight prompt templates that you use one after another. I've called them ebook 1, 2, 3, 4, 5, 6, 7, 8. But it does the whole process because to write an ebook with one prompt template is asking you to do too much to do a proper job of it. You could do it, Clive, you could do a GPT to write an ebook. But again, I think there's too many steps. So I think with something as big as that, as big a task, I would suggest you break it down to smaller prompts that you do in one chat and then just make them template prompts.

Okay.

Next question is Randall. So Randall's got lots of questions today. I have two minutes left, so probably time for one or two questions. In fact I have. Oh, before I answer that Randall's question, I must. One thing I must say, if you're enjoying this and you want to learn more about prompts and so on, we are running tomorrow. I am running our next AI and ChatGPT virtual summit. It's a three hours of training. We are giving away free tickets. So if you have not yet booked for tomorrow. I'm going to be doing three hours. Tomorrow starts at 10:45 Eastern Standard Time, 3.45uk time. You can, Sarah's going to put the link into the chat for you. And over three hours I'm going to give you, we're going to go through a lot of prompts, practical prompts. So if you're enjoying that stuff, then come along tomorrow. Okay, let me go back to Randall's question. So not that one. Yes, it's a new one. If you use Chatbot to write a book.

That'S the same question. I think you asked that twice. Okay.

Great comment from Heinrich Chandry is great at doing translations. Okay, this is going to be perhaps my last question because we've almost hit the limit. Merrill says, as our clients typically have a short attention span and tend not to read what we send them, should we consider AI video combined with these instructional prompt outputs?

Yes. So everybody.

Likes to receive information in different ways. We have different learning modalities. Some people are visual, some people are auditory, some people are kinesthetic. And so whenever you're giving information to a client, then I would use as many of these different modalities as possible. And so I'm a big fan of video and I know a lot of people in the value pricing academy know this. I love Trello for doing quick videos and many of our members do exactly that. And so what I would do is let's say I was sending out a diagnostic report to a client by email. Well, knowing they probably wouldn't read it, I would also open up Trello, open up screen share, open up the report and do a quick three or four minute video where I walk through the key highlights of the report, the overview, the big picture of the report, hit the stop button, grab the link and put the link into the email saying hi John, I've got attached your diagnostic review but to save you time if you want, I've also done a four minute video. Here's the link to the video. So use, use as many different communication techniques as possible. Meryl, I hope that's kind of answered your, your question. I think you talked about AI video and yes, we can also do AI video and so on and other ways of using AI to speed that process up. Okay, we are at a time, we do keep these sessions to exactly 60 minutes. If I didn't answer your question, then I am sorry. Sorry for not doing that. But come along next time and I'll answer your question. Come tomorrow to the Virtual Summit and I will certainly be answering questions during that tomorrow. Or just reach out to me on LinkedIn also, if you're new. Please tell me in the comments. Please tell me what's the best thing that you've learned today? What's been most useful to you? Also, hit the thumbs up the like button if you've learned something useful. And I'm going to play it with some music for three minutes. Gives me a chance to read your comments of what you found was most useful. I will see you again soon, hopefully tomorrow. Bye for now.