Transcription
All right, today we're going to do three things. We're going to instruct you on building a step-by-step ChatGPT prompt that's going to sequence and it's going to trigger ChatGPT to do more than just answer a question. This prompt sequence will turn your ChatGPT into an interactive assistant.
The sequence I will show you today can be applied across a whole variety of different tasks. The principles of this of this particular sequence are important because you're going to do a number of things. I'm going to help you with to build a prompt. They call it prompt engineering. It is the skill that you will need in order to survive. Prompt engineering, remember that, okay?
This prompt sequence, this prompt engineering, this is a two-step prompt sequence that you can use with emails or updates to clients through your case management system or any other communication tools. So let's take a look now at the sequence. I'm going to show you the prompt and let's go ahead and I've already entered this and sequenced it out. But here's the prompt:
I want you to pretend you are an attorney representing Ron, my client, and a company named My Client Co. You are negotiating the termination of services with adverse company and you're dealing with corporations, uh, adverse company corporation's attorney. His name is Mike, adverse counsel. I will provide you with a draft response to the latest emails received by Mike. And here's the key: Before proceeding, I want you to ask me to upload Mike's latest email for context. After that, ask me to upload a draft response. I then want you to provide a revised response that stresses that this is the final offer, improves the tone of the conversation as courteous and professional, and expresses confidence in our legal position.
And now I'm going to ask my assistant. Do you understand my instructions?
ChatGPT says, "Yes, I understand your instructions. For me to help revise the response appropriately, I would need the necessary context. Could you please upload the latest email from Mike, adverse counsel? Afterwards, please upload your draft response. With these, I will be able to provide you with a revised response that maintains a final stance, boosts a courteous and professional tone, and expresses confidence in our legal position."
Wow. All right, let's break that down. This is prompt engineering, my friends, and this is where you need to build your skill set. So let's take a look at this and break it down.
So, note: I am telling AI who I am or who who you want it to be, either or. You could say, you can use first person, or you could say, "I want you to pretend you are an attorney." Okay? I want you to pretend you are an attorney. And then you're going to tell AI who you're speaking with. If this is going to be a communication, you're going to be speaking with Mike, adverse counsel, Mike's adverse company corp's attorney.
The third thing you need to do in order to to do, uh, the correct prompt and to get really deep into prompt engineering, what you're going to need to be able to do is provide context. Context is critical. Prompt engineering requires you to think about all the information you would have to provide to your assistant in order for your assistant to do the job. We had a prior video, which you can go back and check out, that was about the essential skills that you need in order to succeed at AI. And one of those skills was to be able to delegate. Delegation is a skill that most people struggle with, and understandably, you always want to just do it yourself, "Only you can do it all," that nonsense. But in the AI age, to use these assistants, you're going to need to learn the language of delegation. Part of that language is to provide context. Context. I provided my prompt is, "You're negotiating the termination of services. I want you to ask me to upload Mike's latest email for context." So the context is going to be the latest email. It could be prior contract language. It could be any sort of content that you're uploading as context, okay?
The fourth thing, fourth thing, is to is to provide instructions. What do you want your assistant to do? "I want you to ask me to upload Mike's latest email for context. After that sequence, ask me to upload a draft response. Sequence. I then want you to provide a revised response that meets these criteria." So let's go back into ChatGPT and see how AI processed the who you want it to be, who you you want to be speaking with, what context needs to be considered, and what are the instructions. These are the prompt engineering skills that you need for this particular prompt strategy.
All right, so it understands my instructions and it asks me to upload my adverse counsel's prior email. I simply cut and paste this out of the email, and this is, of course, I'll pretend content. And so it provides all this background context. The key here is AI is going to consider all of this when it drafts your response. Now, I I uploaded that, and so then it continued the stepwise functions of my assistant by saying, "Thank you for providing the email from Mike. Now, please upload your draft response for me to revise as per your instructions."
So this is where we have to do a little bit of work again, and this is what do you want the AI to to say? And so let's go ahead and take a look at at at the the instructions that I'm providing about what I want this draft response to convey. Now, the first thing is, you don't, this doesn't have to be pretty. The whole point of this is that it's supposed to be saving time. So you need to spell check. You don't need to correct grammar. It's quick and dirty. AI will know what you mean and correct everything. Since we're trying to save time, it doesn't make sense to spend too much time making this perfect.
So what I did here is I just created bullet points of the different things that I wanted to convey in this response. So this is a lot easier than drafting the recent clients. I'm just saying, boom, boom, boom, boom, boom. Here are some things that I want to include in the response. Um, I want to, I want the AI to know that I already spoke with my client. We want to express appreciation for the continued conversation. We want to, um, have a, I recall that there was some prior negotiations with these numbers. We want to express that, you know, we're willing to make this new offer of 15. We hope that it's approved by the client. And then here is something that's key: If not, we will need to shut down services on June 20th. We believe that the contract terms allow us to do so. However, we respect your position to the contrary. Again, we're not looking for a fight here, right? Um, then we want to stress that they've been on notice since since March, and so they've had plenty of time to transition to a new vendor. They don't have any hardship. I'm in close contact with my client. And here I'm, I'm dealing with some of the things that that he said to me. Said, "Oh, do you have authority?" Well, dumb thing to say. That was just triggered by the fact that I shoved a bunch of cases down his throat to show that we win, probably as a matter of law. And so he responded with a bit of anger. And so we want to address those issues, but let them know that I do have full authority. Um, we're ready to go to court if necessary, but that would be unfortunate.
So here are the bullet points that I provided to ChatGPT. And again, you don't even have to make them bullet points. Just put it in whatever format you can cut and paste out of. So here is the response that ChatGPT gets me, and it is tone perfect. It is a much better response than I could do on my own, and it has been done in a much shorter period of time than I could have done it.
"Dear Mike, I trust this message finds you well. I write to acknowledge receipt of your email and respond on behalf of my client. We remain open to and appreciate the dialogue aimed at reaching a mutually beneficial agreement." This is the kind of thing that I tend to skip in my emails. I am more on the very practical, get it done, very frank, very to the point. ChatGPT, because it has this ability to predict the next word, and it's programmed to, unless you override it, to be professional and appropriate, and we've instructed it to be professional and appropriate, can put in this really great language, which really sets the tone, puts the reader at ease, doesn't pick unnecessary fights. "As to your first point, rest assured that I'm fully authorized to represent my client in this negotiation. My capacity to negotiate does not necessitate licensing in Michigan. I am, of course, licensed in Michigan. If our litigation becomes necessary, we'll engage local counsel." So I actually changed the state in this particular example. "Turning to the subject of revenue share, we note your proposition of a 35 percent share. After careful consideration, we are prepared to increase our initial offer to 15 percent, um, and this will be on a new month-to-month contract. Um, this marks our final offer." This is important. I told ChatGPT to be firm, and one of my bullet points was, "This needs, this is going to be our final offer." At least the import of my bullet points was that it was a final offer, and so it knew to make this our final offer. Um, all right, but again, we'd like to find an amicable resolution to this matter. We hope to see this proposal approved by your board. In the event the proposal is not accepted, and this, all these dates are coming out of that context that we provided, we maintain our position that the original contract has concluded, and we are within our rights to discontinue services." Wow. "Perfect. Please note that adverse company Corp has been provided ample notice and has been in a position to secure an alternative vendor for several months now. Continuation of services up till now has been a courtesy." Wow. "Perfection. I welcome a call to further discuss these matters. We hope, we we are hopeful litigation can be avoided. Thank you for your time and consideration."
So you can see this really nailed it. You do have to review and potentially edit this. This one happens to be perfect, but you need to to review it and make sure you get it right. Now, I didn't mention this at the beginning of the video, but I'm going to give you another power tip. If you want to change the email in any way, you can instruct your assistant, ChatGPT, to do so. You can say, "Let's just go to my tool here, which is, as you've seen from my prior videos, the most important tool in your toolbox, which is speech to text." AI-enabled Whisper, available through OpenAI. You can check out how to get that in the prior videos. So I'm going to, I'm going to modify this email. "I want you to modify this email and be more direct and less polite. I want the email to come across as firm and aggressive on the following points: We will litigate as necessary. This is our final and last offer. We believe we will win as a matter of law." Boom. Speech to text. It just took all of that. So easy. Cut and paste. It understood exactly. Got it. Perfect. It's not Siri. And now I can put this into ChatGPT to make a modification. You're going to see a new version here that is going to be much more aggressive and and and modified in exactly the way that I instructed. And so I'll let this come back and we'll we'll circle back to this in in a minute. See how it does.
All right, this takes me to power tip number two. We've told you about prompt engineering, the sequential step-by-step prompt for communication. Now we're going to talk a little bit about something I just referenced, which is Whisper technology. All day long, I use Whisper. I use it to dictate my time entries. I use it to dictate my draft emails. I do use it to dictate my instructions to ChatGPT. I use it to dictate my team updates in Skype. I use it to draft initial language for contracts. I use it all day long. You can speak three to five times faster than you can type, even if you type well. So you want to get three times more work done in a day. Whisper. The whole AI model is moving towards speech to text. Let that absorb. You will not be typing in the future. You will not be typing hardly anything. Maybe a little bit of editing. Everything is going to be you speaking to your device, your computer, your cell phone. Everything is going to be speech to text. If you don't feel like you're very good at that dictation, you need to get good at it. You need, you must get good at it. And one of the tips that I always share is stop thinking about it as dictation. Think about it as conversation with your assistant. Learn to speak, learn to speak to text.
All right, so here is the tip. All of this information that you see here was dictated. So when I, uh, the background email, I I cut and paste, but all of this was dictated using Whisper technology. So I have to type all this out. This took me literally less than a minute. And all I did was refer to some of the prior materials as I was dictating in order to get these key points. So I didn't have to type this out. You need to learn to speak your prompts, your instructions, as as a key way to save a lot of time. And you could always do some editing to it as necessary.
All right, so use dictation to as a shortcut and time-saving measure when you're interacting with ChatGPT. Now, the third thing, and this is really where where everything is going next, and you probably haven't heard about this unless you've watched my prior video from earlier this week on this brand new feature for ChatGPT, and it's ChatGPT Share. It's only been made available to certain people. I'm one of those early users and adopters. So I'm on that list to get all the new latest and greatest stuff first. It may or may not be rolled out on your ChatGPT yet, but it will soon. And it's right here. It is this little icon that allows you to share this entire thread. You can see the whole thread gets shared here. So remember I told you I was going to help you teach your staff, your employees, your partners, everyone else who's critical to your success, how to use ChatGPT. This is how you do it. You show them this. You copy the link. You send an email. You send it in Skype, whatever your communication tool is, and you simply invite them to see what you just did.
All right, my name is Enrico Schaefer. I'm an AI and Technology attorney, helping you level up on AI. I want you to do me a big favor right now. I want you to like this video if you got value. I want you to share it with anyone else on your team that you think might benefit from it. And I want you to subscribe to the channel so you get the benefit of all of the content that's coming your way over the next months that are all designed to 10x you using AI. Until next time.