The Marketer's Guide to Chain Prompting with AI

Dan Sanchez [00:00:05]:
Welcome back to the AI driven marketer. I'm Dan Sanchez. My friends call me Sanchez, and I am on a journey to master AI in 2024. Today is part 3 of AI fundamentals course that I am releasing on the podcast for free, and it will always be free. I'm putting it on YouTube, and I'm also gonna be stringing all these videos together in a personalized course powered by AI to come in the future. So once all these videos drop, stay tuned for hearing about how I make that, because I'll be documenting that whole journey as well. But I'm trying to get the videos out there as fast as I can because enough people are asking me about how to how to level up their AI game. So I'm trying to make a step by step instructions on how to do that.

Dan Sanchez [00:00:44]:
So today, like I said, is part 3, and we're gonna be going into a topic called chain prompting. Now this is a big topic in the AI community, and lots of people have talked about this and how to do it and where to do it. I'm gonna be showing it based on the topic that we've been talking about. 1st, we talked about super prompts. The last video, we talked about the step method for getting better results with Chat GPT 4 specifically. Today, we're gonna be talking about how to use chain prompting to level it up even still, because, like, you thought we were done with the step method. You're like, oh my gosh. These prompts are getting so freaking long.

Dan Sanchez [00:01:18]:
There's a reason for it, isn't there? So if you haven't checked out the super prompting or the step method episode, go back and watch those. They're right before this one in the feed or close to it, and let's get started into this one. Now, again, if you are listening, I'm gonna do my best to explain everything that I'm doing, but this is a video. It is gonna be available to watch on Apple Podcasts and on YouTube or on the website, aidrivenmarketer.com. But if you're on Spotify or one of the other podcast apps, then hang with me. If you're on a run or something, I'm gonna try my best to explain what's going on. But this is a video podcast, at least for this episode. Not all of them are as visually intense as this one, but I am looking at chat gpt.

Dan Sanchez [00:01:59]:
So let's dive into it. I am here in Chat GPT 4, in OpenAI's interface here, and we are looking at the very end of the last episode that we had, where we had a long mega prop using the step method to load in a PDF that HOD as a template and an example of what we wanted this post to look like. But there's a problem with it that I didn't tell you about in the last episode. And that in this one, you could see I even broke down the prompt into multiple steps. I have a PDF with that example. It has the transcript from the episode we're trying to convert into a LinkedIn post, and it says very specifically, you are an expert sales coach. So it has the role. You take a deep breath and follow these step by step instructions.

Dan Sanchez [00:02:40]:
Step 1, based on the transcript from the episode, find one solid idea in the episode, then move on to the next step. Next step 2, turn that idea into one actionable post for LinkedIn using the template and explained in the attached an example on the attached PDF to guide your writing. And then there's a bunch there's a little bit more written here with other instructions to keep it more concise and all that kind of stuff. But there is an issue when you're giving multiple steps in one prompt, and that is currently at this moment with chat gpt 4 or 3.5 if you're using the free version of AI. And this goes for all the language models, not just chat gpt. AI currently struggles to break complex tasks down into multiple steps by itself. Because of that, it often does much better if you break it down the steps for it. And I don't mean just in one prompt and making it clear.

Dan Sanchez [00:03:35]:
I mean, like, actually giving it time and space to execute step 1, let it go. And then based on the results it gives you after step 1, ask giving it another prompt. If you can baby step it through and link them all together, that is what we call a chain prompt. In fact, I posted about this back in January because I turned, one of my custom GPTs into a chain prompt in case you weren't paying for, Chat GPT Plus to be able to run the the faster way to do it. I turned it into a chain prompt so you can get preproduction done for your podcast with just a chain prompt. You could see a lot of people are putting out slide decks like this on LinkedIn, and I'm tuned into LinkedIn now where I'm looking at the slide deck I made. It's like your AI show runner. Here's the chain prompt to make it happen if you scroll through the pages as step 1, put this prompt in and it has some places for you to copy and paste it in there, some information about your show.

Dan Sanchez [00:04:27]:
So you prompt that, and then you you let it load and do its thing, and it gives you some information, then you give it a second prompt. Prompt. And then after it loads that, you give it a 3rd prompt, and then you give it a 4th prompt, and then a 5th prompt in order to walk it all the way through the my showrunner process. And the results that I get from linking together the prompts like that are freaking amazing. You can get better results, too. In fact, let's go back to our original example here. We're gonna keep going on this one train. Turning Brownlee's Brownlee getting lots of airtime here, turning your episode into a fantastic LinkedIn post.

Dan Sanchez [00:05:01]:
So let's see what happens. Let's see the difference when you turn, a super prompt into a chain prompt. It even becomes better even using the step method. So let's look at let's take a look at what happens when you do this. So I took the same prompt we had, and I just didn't put it all in there. I broke it down. Remember, we have this super long prompt, and I'm, like, highlighting the last one in the from the last video. It's really long, and, of course, the whole transcript's embedded in there.

Dan Sanchez [00:05:26]:
It's got the PDF, and I'm just gonna break it apart. Break it up a little bit so the AI can, you know, just handle one thing at a time, much better at that. And let's take a look at what I did. I still have the role. You are an expert sales coach and former VP of sales, Matt Brownlee. Based on the transcript from your recent podcast below, find 5 solid ideas that can be turned into a LinkedIn post and let me choose from the ideas. So now instead of having it then move from it selecting 1, because who knows, there's a couple of major points Matt made in the episode. If we turn it into a chain prompt, we could take a quick look at them and pick one for ourselves.

Dan Sanchez [00:06:00]:
Right? So we let AI do less and us do a little bit more. That's what's possible, chain prompting. And then, of course, I gave it the episode transcript. So let's take a look at the results it gave. Scrolling down. It's a long transcript. And here it is. Based on Matt Brownlee's insights from the podcast episode, here are 5 potential LinkedIn post ideas, and it gives us and outlines the 5 ideas with, like, a major title and a few sentences describing the idea.

Dan Sanchez [00:06:23]:
One is discipline and focus in sales and sports. 2 is the importance of coaching and achieving success. 3 is team dynamics and sales success. 4 is results matter, learning from success and failure in sales in life. And 5 is strategic focus in sales, narrowing down to amplify results. And I was like, I really like this coaching idea. So let's make a post just about that one. And I said, can you break down number 2 into 3 distinct actionable points? Do not add and this is important.

Dan Sanchez [00:06:52]:
See, this is a prompt. We don't need to do the the full super prompt because parts of it are already in there. We already gave it a role. Right? And now I'm putting do not add any major point that was not mentioned in the transcript, which is important because AI can get off on its own path and start adding in extra stuff, And you have to remind it to go back farther up the chat to go and reference the episode transcript again, which it has access to, but you have to reference that again in order for it to stay on track because it will literally just take, what is it, idea 2 here on coaching and achieving success, and then kinda go make up some other stuff about coaching. In fact, before before I recorded this, I ran a demo and then had to go back and add that that modifier at the end, that limitation so that it wouldn't actually make up stuff because I ran it, and it did. It just made up stuff. I'm like, he didn't say that in the transcript. What the heck? So you have to keep it on track and add little details like that in order to keep it from just making up stuff.

Dan Sanchez [00:07:49]:
And it's not like what it made up was wrong either. It's just not what was talked about in the episode. So here it goes. Certainly and I'm doing this because we all know if you watch this last episode, the breakdown that I have for this post template actually takes 3 snippets, and I wanted to be able to see what it was going to do before it actually went to the LinkedIn post. So it actually gave it coaches provide structural feedback, coaches help with strategic thinking, coaches as motivate motivators and accountability partners, which were all topics he talked about in the post. So then I do the last part of the prompt that we had already created in the last video, which is loading in that post template and example, and then I start with the rest of the prompt. Turn that idea into one actionable post for LinkedIn using the template and example in the attached PDF and a quote, so I quote the PDF LinkedIn post post template dot PDF to guide your writing. The goal is to entice readers and listen to the episode.

Dan Sanchez [00:08:42]:
Other instructions. I actually modified this from the last time because last time we noticed the post was a little bit too long, too many words. LinkedIn is very short, very choppy, much more concise. So I added one bullet in from the normal one to say keep the post extremely concise. Guess what? It caught it, and it probably made it even a little bit too concise for my liking for LinkedIn post, but it was still way better. It was really good. So let's take a look at what this LinkedIn look LinkedIn post now looks like after adding in our super prompt, after adding in the step method, after adding in a chain prompt in order to really narrow it down. What do we have? Let's take a look.

Dan Sanchez [00:09:18]:
And here it is. Ever feel like you're running in a place running in place in your own sales career? You're not alone. Here's what coaching can do for you. And I love that it like, just the little metaphor at the beginning because a lot of the conversation, even as concise as this post was and it's a little too concise, it still adds in some of the flavor because the whole thing was about, like, running triathlons and races compared to sales. So it's still putting that in there. You're not alone. Here's what coaching could do for you. 1, and then it has it broken down to 3 points structure and feedback, strategic thinking, motivation, motivation, accountability, and then breaks it into, like, 2 super each point gets an extra 2 sentences, but they're, like, 5 forward sentences.

Dan Sanchez [00:09:56]:
Structural feedback keeps you on track, shows you where to improve. They're fragments, but like it still works. It keeps it short and concise, probably a little bit too concise. I would have liked to flush those out a little bit more. And I could go and ask it now to, like, flush it out just a little bit more, a little bit too concise. And we could do that. But I just wanted to show you what this post can now look like because this is now looking like a really fantastic post. It just needs a little bit more modification.

Dan Sanchez [00:10:22]:
Let it breathe a little bit, and it even finishes with the call to action that we want. Find the full episode here. Close mode dot media fan freaking tastic. So you can see the power of the chain prompt allows you to get a little bit more control at it because you can give input along the way, and it optimizes for what the AI can accomplish. It's really good at very small steps. 1 at a time. It can analyze, it can think, it can do research, it can synthesize a lot of data. But if you give it a multi step project all in one go, it'll try.

Dan Sanchez [00:10:55]:
But the more you can break it up into smaller pieces, especially with some guidance along the way, the more likely it is to deliver you better and better results. So that is chain prompting in the nutshell. But stay tuned for the 4th part of this series where we're going to talk about, how to simplify this whole process. Because believe it or not, like, you're probably looking at this be like, Dan, I am not gonna go through with this freaking amount of mountain of work of having a super prompt and a freaking template and example for every time I need to turn a podcast into just one LinkedIn post. What about the blog post? What about all the other stuff? I got you covered because that's what they've created custom GPTs for. Because believe it or not, you can if you're doing this over and over again, you have transcript after transcript and you need to turn it into a LinkedIn post every time, we can automate that whole process. Yes. We can we can build this this process out once at a chain prompt and and a template and example on the super prompt.

Dan Sanchez [00:11:50]:
We can build that all once, refine it, and then just let that run over and over and over again. Do you wanna see this turn into a custom GPT? Wait for that next video on how to build simple custom GPTs and then complex ones after that. So stay tuned.