Speaker 1: 0:00

The art of prompting for chat GPT. We talked about it. You got the handout If you listened to the last episode. If you didn't, well, no worries, there's one down below. Same thing right down below. Just click the link.

Speaker 1: 0:12

So you're an agent like me, pnc agent probably, maybe a life agent. Either way, that's awesome. Are you ready for some cold, hard fact bombs? The robots are coming, my friends, and they're not just here to take your jobs, they're here to steal your clients too. That's right, the AI revolution is upon us, and if you're not prepared, your agency could be the next casualty. But fear not, because in today's episode, I'm going to arm you with the knowledge and the tools you need not just to survive, but to thrive in the face of this existential opportunity. Are you ready? Let's go.

Speaker 1: 1:00

So in today's episode, we're going to dive deep into the art and science of prompting. We're going to explore eight crucial steps that are going to help you get the most out of interactions with your AI tools, from chat GPT to custom-built solutions whatever's out there right? So grab a pen and paper, because you're about to learn the secrets to becoming a master prompter. Now grab your sheet. Click down in the bottom, if you haven't done it already and you can hit pause, that's fine. Hit pause and just go get it. But number one, step one, is be clear and specific. Now, a lot of these things are really going to help you in your marriage too, but in your dialogue with any of the LLMs, especially chat GPT, you want to make sure you're clear and specific. The first step is really crafting to crafting this. An effective prompt is to be super crystal clear. You're not dealing with something with emotions. You're not dealing with something that can infer like think Spock, you're talking to Spock. Ambiguity is the enemy of you getting results. So you want to make sure they're not going to make any assumptions. And if they do make any assumptions, they're going to make the wrong assumptions or weird assumptions. So we just want to. We want to manage it right. It's going to get us the best data possible, as long as we are putting the right information in to extract it right. You're not going to go try to put the nail into the coffin. Maybe that's not the best analogy. You're not going to go try to put the screw into the wood over here using a hammer. Right Wrong tool. You can use a screwdriver or a drill. But same thing here, right? Ambiguity is the wrong tool for this.

Speaker 1: 2:39

We're not trying to make somebody interested or make them curious, right? We're trying to extract information from something that has it. So we have to ask the right questions in order to get it. So, as an example, you want to learn more about the benefits of life insurance. If you said, tell me about insurance, you're not going to get a great answer, right, it's going to be pretty crappy. But if you say, what are the benefits of insurance for young families? Now you're going to get some bullet points. Right Now you have something that you can write an article about or that you could put in. You could then take those three bullet points. We'll get down to that when we get there.

Speaker 1: 3:10

But the key is being specific. So don't leave room for any interpretation, any misinterpretation, because that's what it will likely be, and the more targeted that the question is, the more targeted the response is going to be. So garbage in, garbage out, precision in, precision out. So that's great. Provide context. Context is so, so important here. Context is king when it comes to prompting. So the more background information you can provide, the better equipped the AI is going to be to understand the nuances of the request and give you that golden response that you want. Request and give you that golden response that you want.

Speaker 1: 3:50

Think about asking a question of somebody that is like in the middle of a process but not giving them the steps prior to that question. Right, they're not going to understand what to say. So it's the same thing here. Every time you talk into this thing, for the most part, it doesn't have much memory, so it's not going to remember that prior conversation. So you want to make sure that you provide it enough that it can infer from that, right? So, man, my nose is bad news today. Sorry about that Again.

Speaker 1: 4:14

Like, think about, like, term life. What is term life insurance is going to give you a response, right, but if you add some context like this explain the benefits of term life insurance for a 30 year old nonsmoker, now you're giving it really, really detailed context. So think about the situation that you're asking it in and for and you want to make sure that you supply all of the details. Now, there's some of them. Like some of the graphics generators are a little different where it wants shorter, more concise, but for all intents and purposes, with ChatGPT, I believe you can put in like 120 000 token prompt, which is like 100 000 words, something like that. Um, all right, so, oh, yeah, oh, number three is great use simple language.

Speaker 1: 5:03

When it comes to prompting, simplicity is your friend. So you don't want to use jargon, you don't want technical terms, you want to use complex language, like it could come up with complex language, but don't don't give it to it. So it has to like unwind what you're writing, right? Think second, third grade level, fifth grade level, something that's really clear and really easy to read, because the easier you make it for it, the more effort it can put on the output instead of trying to understand it. So, and and then miss the mark. Right, you want to get it. You want to hopefully have it. Get it in one shot, right, the first shot.

Speaker 1: 5:38

But you often, and like many times we're going to ask again or ask again, ask again, and keep making it more, keep whittling it down, because you're not going to lose it. So, for example, here, can you elucidate the operational mechanisms of life insurance? Why would you say that? I don't think anybody ever would, but if you did, it's probably not going to get you the best response. Right, because it's a little. Yeah, so how does life insurance work? That'd be great. Right Now, in relation to the prior questions, you probably might be thinking, hey, that question is a little too short or not clear enough, and you're right.

Speaker 1: 6:16

But for this example and for this point, it's just to help you out there. So step four ask one question at a time. I'm bad at this. I do it. I have a lot of questions in. I will number them if I do cram multiple questions in. So be careful with that. What I've found most recently that's become more effective is if I type out the questions or I voice to text the questions out so that I have them. Then I'll copy and paste each one individually and it is just going to put more energy into each question. If you gave it five questions numbered, it's going to answer them Likely. It's going to answer them Likely. It's going to answer all of them. But you're going to get much less meat on that bone. So you want to make sure you go one at a time.

Speaker 1: 7:01

Again, another example I wasn't very creative when I was coming up with my stuff, but it's just to prove the point. So, one question at a time. What is whole life insurance. Now, if we're going to follow the format, I probably would say what is whole life insurance and how does it help a 23-year-old? We don't want to ask what is whole life insurance, term life insurance, universal life insurance and what are the accounts that are associated with it. We don't want to do all that. If we want to ask all those things, totally cool, you have unlimited questions. It's not the genie, right? You're allowed to ask more than three questions, so just do it one at a time, no problem.

Speaker 1: 7:40

Specify the desired format. So, if you have a format in mind for your response, put it in there. So, whether you're looking for a bulleted list, a summary, a paragraph, a detailed explanation, specify what you want. Now you also could set in your settings. You could have like a generic system prompt, so a tonality that it always talks in, or a way to address you. Like I haven't called me dude, so it says dude. I said act like a surfer, so it talks all kind of surfer to me. I told it to throw F-bombs, and every now and then, so it makes it fun, right, but you can organize the way that it speaks to you in any way. So, however, you like to have that done. So that's really helpful. Right, you're going to specify that.

Speaker 1: 8:22

So, oh, another thing that's really awesome, and this just came out. So, if you're new to it, great, you came at a great time. Back before we weren't able to do this, but a couple of days ago, when the new update came out, I just asked it hey, can you put that in a PDF? And boom, it outputted a list that I made. Oh no, it was um, oh no, it wasn't that. I thought it was the eight, the eight steps, but it's not. So, um, but I could have, like what I did was I told it to to create a Google doc and it did. So that was pretty cool, and I didn't know it was going to do that.

Speaker 1: 8:55

Like, that's new. So every day it's almost every day or every week, something new is coming out. And, uh, just think, you know, if you wait four weeks, that's four new things that you missed. If you wait eight weeks, that's eight new things you missed, and you're going to be even more behind and you're going to be more like Blockbuster, and I hate it when people are like Blockbuster. So, let's not be like Blockbuster, let's be like Netflix.

Speaker 1: 9:16

So if you were to say something like what are the benefits of whole life insurance? Probably not going to be that great, but if you said, list the top five benefits of whole life insurance, that you're going to get something right. And you could say give me bullets of the top five and then it's going to do, do, do, do, do. And if there are long bullets, you could say make them shorter. Right Now you can talk to it like mid conversation. Number six step six examples. I love examples.

Speaker 1: 9:42

It really, really this is kind of new for me to add the examples. I this is kind of new for me to add the examples. I didn't used to put it, but I was at a bot summit event where we were doing some builds on some new bots for the agency and what we found was that using examples, huge, huge. So if you're trying to understand how term life insurance works, you might prompt the AI with something like explain term life insurance with an example of a 40-year-old policyholder. This will help the AI frame it. So you're creating that. In this case, you're asking it to create the example. You could also provide examples for additional context, and it works out very well. So I'll say something provide examples for additional context and it works out very well. So I'll say something like for additional context.

Speaker 1: 10:32

Think of it like this here's an example A magician pulls a rabbit out of the hat. That's a trick, right, if we're talking about tricks or whatever. So that is a really great way to make it relatable. You can. You can think about it as if you're talking to a kid. You're trying to explain a concept to a kid who's going to give you really valuable information, but it's hard to get them to understand it first. Right, so really smart on the back end, but we have to first communicate.

Speaker 1: 11:00

Number seven be patient and iterative. Yes, you may yell at your computer, but that's okay. I was having to help me with some code yesterday that I was working on. I was learning Python and I was doing some stuff, and I got stuck and it just kept running me in circles. But what I noticed was I was, instead of being patient and iterative, I kept demanding the same thing and I was putting all caps and I was cussing at it and that wasn't getting me anywhere, right, it just kept giving me the same answer. I partially did it because I wanted to have that happen so that I could tell the story. I'll be honest, but also it's a good lesson, right? There's no sense in getting mad. There's no sense in thinking it's stupid because it's not and it doesn't matter, right? It just means I'm asking it correctly and if I keep doing it this way, it won't help me now and if I keep doing this way in the future it probably won't work. So we want to stick to things that work. So that is number seven. Let's go to number eight. We will bring this baby home. We're at number eight. There we go.

Speaker 1: 12:04

So use follow upup questions. Don't be afraid to use follow-up questions that can help you dig deeper. So, just like a human conversation, follow-up questions are a great way to keep the discussion flowing. Right? You're going to ensure that you're getting the most value. The thing will come up with more stuff.

Speaker 1: 12:20

You can do something like, let's say, you have it. Help, you write a post, right? So the first shot's not going to be great. Definitely, never copy and paste anything that is created without reading it first, but I would give it two or three iterations first, right, so give it. Ask for the post. Hey, I want to do a post about the great rates we have in Alabama. Awesome. So it creates something. Now take that, put it back in there and say is that really the most ideal post that you could make to attract people? And then it's going to give you a bunch of reasons why it's not. And then you can say apply them. And then you do it again, and then you do it again until it doesn't find anything, and now you have yourself a really good post that you could personalize. You're going to go in there and change it around because you don't want to just copy and paste garbage. So just be careful, but that's a really good way to iterate that I've found extremely useful. Yeah, so that I think my folks is a wrap. I appreciate you jumping in here.

Speaker 1: 13:17

Make sure to download that sheet If you don't have it. It has the eight simple steps, everything I went over today so that you can get on ChatGPT. Make sure you pay your $20. It's worth it. It is a no brainer. You get access to that for oh, I don't work for them. I don't get anything by saying this. It's just. It's that important because the folks that don't do it are the ones that are going to be sitting at home pretty soon. So thanks a lot, folks, and appreciate it. We will talk soon. Be good.