Customize Your AI: How to Build a GPT Assistant For Client Emails

This is the 3rd video in this series on GPT assistants. GPT assistants were announced by OpenAI in November 2023. They are game changers for business professionals looking to 10x with AI tools. In the prior 2 videos, we talked about some of the context you need, uh, in order to understand GPT assistants and why they're so special. But in today's video, we're going to actually put a GPT assistant in action. We're going to show you how to customize and build your own GPT assistant and help you understand how to go about that process of building an assistant for a specific task. So go ahead and subscribe if you want notice of the next videos in these series. And go ahead and share this video if in fact you have any business associates who you think are interested in AI and might benefit from being part of this growing community that we've built of professionals looking to 10x with AI. All right, let's dive right in. With GPT assistance, customization of your AI tool is not just a theoretical concept. It's not just out there in the future someday it's going to happen, but it's a practical reality right now as of November 2023. A prime example is a specialized GPT assistant that I developed for handling client emails. I'm a lawyer, I'm in the legal profession, I deal with a lot of client emails. I've always felt I did a decent job in developing email responses to clients. But I will tell you, I have 10x'd my email game with this GPT assistant that I'm going to show you today. I have 10x'd my ability to draft an email that is comprehensive, that is on point, that's got the right tone, that adds additional value and builds that relationship with my client. The intangible value that is more than How much time did it take me? So this tool that I'm going to share with you today exemplifies how customization can cater to a very specific business need. Remember, you can create a GPT assistant to do any specific task, a repetitive task, one that you're doing multiple times a day, or for a specific project that you're going to be revisiting day in and day out for some period of time. Before jumping in, let's just talk again about prompt engineering. Prompt engineering, as you might recall, is the art of talking to your generative AI tool, to your chat-enabled AI tool. Prompt engineering is the, is the way that you not only ask questions, but the context that you provide the AI tool, the knowledge that you provide the AI tool, the format of the requested output, the tone of the output, the approach to the output, how you design this communication with chatbot is mission critical. So we're going to be doing a lot of videos in the future that are going to circle back on prompt engineering as a mission critical task that every professional needs to get better at in order to really dial up on AI. This legal client communication assistant is, um, is not a generic tool. It's been carefully customized with prompt engineering to align with the specific requirements of legal professionals looking to respond to client emails. It uses a unique prompt structure and specialized knowledge, ensuring that it drafts and refines emails from a lawyer's perspective. You can see some of that prompt engineering here. So first, it's going to ask for any relevant background information or prior emails, ensuring that its responses are informed and contextually appropriate. So this is really important. Just take a minute here. The more information you put into the AI tool, the more context, the more specific information, the more specific guidance. The, the more you guide it as to how it is going to assist you, the better the output's going to be. So this first step, this first step that I've put into this assistant asking, hey, do you have any relevant prior emails or information? This is crucial for maintaining the continuity and relevance of the communication. You don't want a generic email to your client. You want to respond to their specific email, or potentially there's 10 or 20. Emails in the thread, you're going to upload all of it into ChatGPT Assistant, and that's going to help that assistant really provide you great information. And by the way, you don't necessarily just have to say draft me an email. You could say, hey, could you summarize this prior email thread? You can say, um, what were the main concerns of my client after you upload that information? So before you even get to drafting the email, you can educate yourself have your assistant educate you about anything that is going to be found within the background information that you upload into the system. So the next thing that it's going to do is— and this is stepwise prompt engineering— first you're going to do this, next you're going to do that, next you're going to do that. Second in this prompt engineering sequence, it is going to inquire if there's an existing draft Of a response email that you've put together, or if you've got key points to include in the response. This allows the assistant to build on your input and the specific things you want to communicate, tailoring the message to fit the client's specific situation needs and the things that you want to tell the client. Right? So again, let's take a minute here. You can put. Down simple bullet points or stream of thought as the information that you want to include in the response. I go to ChatGPT, you know, OpenAI Playground. I use the Whisper, uh, tool, the speech-to-text AI tool, in order to just dictate into the tool what it is I want to say to the client in a kind of a free-flowing way. And then I cut and paste that back into this ChatGPT assistant is, you know, from my prior videos on Whisper technology, you can speak a lot faster than you can type, as I'm now illustrating right here, right now by talking. And so having this speech-to-text Whisper technology there always available to take your dictation and use it to input information into your assistant— boy, game changer, right? If you're a great typist Awesome. But I encourage you to go check out Whisper technology. Look at my prior videos because you can speak faster than you can type, number one. But number two, you're going to need to learn to speak to your AI tools. Typing is not going to be the mode of communication in the future. Let me say it again. You need to learn to dictate speech-to-text instructions using these AI-enabled speech-to-text tools. Such as Whisper technology. All right. So let's go down to the 3rd thing in this prompt engineering sequence. And this is going to set the tone. So you want the GPT assistant to consider the desired tone that you want for the email, whether it's a formal tone or informative tone or casual tone or friendly tone or any other style, this assistant can be adapt— Adapted and can adapt its output to suit your particular preference. And if I've got clients who I've had longstanding relationships with, I'm not going to want a professional tone. I'm going to want a friendly and casual because that's the way I communicate with them day in and day out. And as you're going to see, the GPT assistant, by the way, is going to spit out emails that you could never dream of doing yourself. GPT's special gift to the world is it can draft really appropriate, really nuanced, balanced, perfectly tailored emails for virtually any situation. So in this case, it's going to ask the user and suggest to the user, here's some possible tones that you might use. Once it's got that information, it's going to give you a draft email. And in many instances, that draft email is going to be 90, 95%, 98% of what it needs to be. I'll add some edits. I'll put it into the response email, cut and paste, boom. And again, play with this tool, play with this GPT assistant I'm going to share with you at the end of the video, because you're going to see that it's not just a, oh, I saved time drafting a response email. That is not the point. Time is not the point. Saving time is not the point. The point is much bigger than saving time. Sometimes it might take you more time, but you're going to— the quality is going to be incredible. The additional educational value either to you and/or that you're conveying to your client is going to be 5x of what it otherwise was with great emails and client communication in the age of, of, of okay or K as the response, right? The letter K. As everything gets shorter and shorter and shorter and shorter and shorthand and abbreviations and all the rest, we've lost the art of communication. ChatGPT has not, right? So it's going to give you this great email. You can use it, or you can do what you do with everyday assistants, with real people. You say, that's good, but I want you to go back and do this, or I want you to add that, or could you prioritize this, or the tone is still A little bit too formal. I'd like it a little bit more casual. I'd like you to shorten the email. Whatever instructions that you would normally provide an assistant doing a task, you are going to provide to the GPT assistant. You should learn to iterate with these assistants. Hopefully it's not going to give you a perfect output so that you can learn how to continue to work with your assistant to get the right output that you want. So. It's going to actually, you can see here, invite feedback and suggestions for improvement. It's going to emphasize to you that it's only a collaborative tool. And the collaborative nature of this tool is to work with you, not for you. And so this last step is going to ensure that the final output is going to align with your personal standards and expectations. Boom. Let's talk about the next important thing. Enhanced knowledge. In the world of GPT assistants, customization is everything, and it reaches new heights that didn't exist before GPTs or GPT assistants, however you want to refer to them, because of the ability to provide enhanced knowledge to the GPT assistant that becomes foundational to the assistant. So you can integrate specific documents into the assistant as part of its knowledge base. Let's take an example of a legal client communication assistant like we've been talking about, which not only follows a tailored prompt structure but also utilizes uploaded documents for enhanced performance. For this particular assistant, I have uploaded a document that instructs on how to write professional emails. So this particular assistant has this document, these instructions of how to, how to draft really great emails in a professional session, um, in a professional way. And it's got this document as part of its knowledge base, and this document has been uploaded into it, and it's there for it to access at all times. It has it in its memory. It is building this knowledge base, and you can add to the knowledge base. These documents provide, um, you know, knowledge that's going to shape the assistant's understanding of the output that you're desired. And so this is crucial in a field like law where you're not looking for a generic communication. You need the communication to be clear, precise, and adhere to professional standards. So by incorporating this particular document, the GPT is equipped to draft emails that reflect best practices in professional legal communication. The integration of specific documents is infinite. Up to 300 pages of documents can be uploaded into any GPT assistant. And uploading documents, uploading knowledge into the GPT transforms it from a generic AI tool to a specialized asset, an assistant for legal professionals or any other professionals. It becomes not just an assistant, but a knowledgeable partner with the context and the expertise necessary to do a specific assigned task that you have asked it to do, that you've created, that its purpose is to do, that you've created for that special purpose. So, um, you know, the example that I've given you here with this ChatGPT assistant, this GPT for client email responses, is just an example. You can create a GPT assistant to do any specific job. So the ability to upload and integrate specific documents as knowledge sources transforms these AI tools into, you know, very specialized tools. Let's just go through some examples, and they're, they're infinite, but let's just take one in the legal field. Let's say you wanted to create a GPT That could utilize, that could help you work on a litigation case. You can upload into that GPT for that specific client, all of the prior pleadings, the key communications, the client's goals. If you've documented the client's goals, you can upload that into the knowledge base. And then you can tell it what, if the client's goal is settlement, that's going to move the assistant in one direction. If If the client's goal, goal is, you know, protracted litigation, that is going to lead your assistant down a different path. So you have to think about the types of information that you need to create to upload into an assistant for something like litigation assistant for a specific case. All right. Another example, and I'm sure you could think of them as well. Let's say I'm negotiating and drafting contracts, right? My client's on one side. There's another party being represented by an attorney that's handling the negotiations for the other side. I can upload all the prior communications, the instructions from my client, my advice to that client, prior versions of the contract, all the negotiation history into a GPT assistant designed specifically to help me with that particular matter. Now, I could design it so that it can also help me with emails to the client on that contract or with emails to adverse party on that contract or to help me negotiate the contract. Or compare the contracts. I can create a single assistant to do all of those things or specific assistants to do very narrow things within that project. So you can see now that the possibilities really become endless. This knowledge upload ensures that the assistant is not only informed by the general principles, but by specific context and details of the case at hand. It's more than prompt engineering. It's adding to the knowledge and the training of that GPT. And keep in mind, it now has that available every time you interact with that GPT, pulling it out of the sidebar. Every time you pull that GPT for that task up day in and day out, it's going to have all that knowledge available to you, but it's also going to have the context of all your prior communications so that it continues to learn and get better, just like a real-world assistant hopefully would. So by leveraging these tailored knowledge sources within these GPTs, you're, you're not only gaining efficiency, but you're, you're gaining expertise. You're aligning with the unique demands required by professionals and by experts within industries, within the standards that are expected within a particular profession. These GPTs can become invaluable partners in your workflow, capable— really, you're going to be blown away— handling really complex industry-specific tasks with a high degree of confidence and precision. That's why GPTs are, from my point of view, a game changer. That's why we're spending so much time in all these videos talking about GPTs, why they're important, how you interact with them, giving you examples, helping you get to a point where you can get the most out of your GPT. Keep in mind A lot of people can interact with ChatGPT. Maybe they're maybe a 1.5x, maybe they're going backwards because they're playing all the time, not getting anything done. If you want to go 2x, 5x, 10x, go ahead and subscribe to this channel. Go ahead and subscribe to this channel. Make sure that you participate in this community. Our videos have done over 100,000 views. We're just getting started. We're creating all kinds of great tools for you as a business professional in the future. To learn to maximize your amplification using these tools. All right, so go ahead and do one more thing for me. If you got value out of this video, just share it with someone that you, that you care about— a business partner, an associate that's interested in AI, uh, someone in your company, someone outside your company. Let's build this community together so that we are all in it together and we can all get that 10x that we're looking for in 2024. All right, my name is Enrico Schaefer. I'm an AI attorney. I'm an expert in AI tools. We'll see you in the next video. This next video is going to be video number 4, and it's going to talk about the evolution of AI in business. Like, how did we get here? How did we get from the general tools to the GPTs?