Episode 438: Mistakes to Avoid When Prompt Writing

Intro

00:00:00:18 – (Bob Woods): Welcome to the Making Sales Social podcast featuring the top voices in sales, marketing, and business. Join Brynne Tillman, me, and Bob Woods as we each bring you the best tips and strategies our guests teach their clients so you can leverage them for your own virtual and social selling. This episode of the Making Sales Social podcast is brought to you by Social Sales Link, the company that helps you start more trust-based conversations without being salesy through the power of LinkedIn and AI. Start your journey for free by joining our resource library.  Enjoy the show!

Bob Woods 00:22

Hey, everyone, and welcome. Thanks for joining us for Making Sales Social LIVE! Coming to you from the social sales link, virtual studios. I’m Bob Woods. Brynne Tillman is off, as in, off in the wild blue yonder as she goes flying to an assignment. As they say in the news business, it’s a little different for us, but still, Stan Robinson Jr, though, is here. How are you doing, Stan?

Stan Robinson Jr. 01:05

Doing fantastic. Bob, good to see you.

Bob Woods 01:09

Good, good, good. y’all have probably heard a version of this phrase by now. AI probably won’t take your job, but someone who knows AI probably will. This is why knowing how to effectively use and leverage AI tools has become essential for anyone in business, especially sales professionals, marketers, business development experts, and those who manage them. And we’re not forgetting about entrepreneurs either; they especially need to know AI and the many roles it can take to help them out with their businesses. And there’s a lot there. AI is great at generating insightful responses based on the prompts that you provide. However, many users struggle to get optimal results due to common pitfalls in prompt writing now, Stan, I don’t know about you, but I hear people struggling to get their prompt sessions right. Is that something that you’re hearing too?

Stan Robinson Jr. 02:09

Yes, there is frustration as more and more people start to use these tools and try to figure out how to talk to them to get the best results. look forward to diving in here.

Bob Woods 02:20

Let’s dive right into it with the first thing in our agenda. Oh, and by the way, before I forget, we do have an ebook that goes along with this. It is free. It’s at social sales. Link.com/mistakes, if you’re watching us live right now, we have a QR code at the upper right I’m pointing there it is right there you can, you can use the QR, the QR code, or type that URL in. It’ll do a little redirect to a much longer URL, but that is where you will download it. And again, it is free to do . Stan, why don’t we get into our little agenda here with number one, great.

Stan Robinson Jr. 03:06

first mistake a lot of folks make is failing to give the AI enough context. in the world of AI, at least as things stand right now, context is king, to speak. You could think of AI as a super smart intern, but one who is totally unfamiliar with your company, you need to give them details about the context of what you want them to do. For example, rather than saying write a great marketing plan for my business? , it’s going to be a different marketing plan if you’re a pharmaceutical company in a highly regulated industry, if you’re a startup, if you’re a manufacturer, being specific about I’m a pharmaceutical company developing, , developing different types of medicines. Write a marketing plan for my pharmaceutical company. Our audience is X, Y, and Z. Use social media, email marketing, blogs and the more specific, the better. Bob. Any thoughts on that?

Bob Woods 04:24

For me, it goes back to an old saying: if you’ve ever been involved with BNI networking groups, it’s be specific to get terrific. It really does apply here. You need to be specific so the AI doesn’t start going off on its own tangents. Believe me, it will if you let it — you have to rein it in.

We talk about the CRISPY prompt formatting that we teach here. Whether you use that or something else, properly formatted prompts are the key to getting exactly what you need out of AI.

We’re going from not providing enough context to something closely related — a lack of specificity. Context and specificity are both essential when crafting prompts. The CRISPY framework automatically includes these elements because of how it’s structured.

Being specific is crucial when you’re crafting your prompts. General inquiries don’t work anymore — in fact, they never have. Instead of asking, How do I increase sales? which lacks both context and specificity, you might say, What techniques can I use to increase sales in a competitive online market for my [product/service] or for my company [company name]? You could even include your company’s main URL.

When you provide clear, specific language, you guide the AI to narrow its responses and give you better results. Otherwise, it becomes that overly eager, intelligent intern constantly asking, What about this? Or this? Or that? The AI genuinely wants to help, but you need to focus it on exactly what you’re looking for.

Stan Robinson Jr. 07:04

Totally agree. A lot of the work that goes into using AI effectively takes place outside of the tool itself because it’s all about clarity — knowing exactly what you want from it. You need to think about that beforehand, since one of the most powerful skills moving forward when working with AI is the ability to ask the right questions.

While AI can help by providing examples, you still need to be strategic about the questions you ask.

Bob Woods 07:46

I want to go back quickly to that quote I mentioned at the top of the show — people who know AI may take your job. If you go in saying, “Oh, I know AI,” and then type in a generic prompt that generates something basic, you might present it to your boss or, worse yet, take it to market as an entrepreneur. If it’s too general, you’re going to get buried — either by the market or by your boss.

That’s why you need to learn and embrace doing things the right way. AI is, without a doubt, a powerful tool when used correctly. But if you use it the wrong way, you’ll end up going in too many directions.

It’s similar to when the World Wide Web first started — you’d go down a rabbit hole, and before you knew it, an hour had passed. AI can be the same way. However, when used properly, it can also save you a tremendous amount of time.

That’s a good one. So, next — who’s going? Is it you or me?

Stan Robinson Jr. 09:01

I think it’s me, because we’ve already alluded to this. One of the biggest mistakes people make is giving up too quickly. That goes back to what we were talking about earlier — asking the right questions. When you ask the right questions, you’ll get answers that are much closer to what you’re looking for.

Even though OpenAI isn’t the best at naming conventions, the “chat” part of ChatGPT is incredibly accurate. The goal isn’t to ask one question and get the perfect answer — this isn’t Google. It’s a different experience. The idea is to go back and forth, like a real conversation.

Imagine you have a super-smart intern. You give them an assignment, and when they return with their work, it’s not quite what you needed. You wouldn’t give up — you’d say, “Here’s where we need to fine-tune. Could you look further into this? I also realized there’s another question we should address.”

That’s the attitude you should have when using large language models. This isn’t a one-and-done question — it’s the beginning of a discussion. And as Bob said, the tool wants to make you happy. That’s why it gives you an answer even if your question is very general. You just need to be willing to keep the conversation going.

Bob Woods 10:43

The keyword here is conversation. The first couple of times you try it, if you haven’t done it before, it’s going to feel a little strange to have a back-and-forth with AI — but trust me, you’ll get over that quickly.

Feel free to say things like, “This part was great, but this section needs some work. Please redo it.” Then let the AI adjust and refine its response. Later, you can even revisit earlier parts of the conversation. For example, after improving Parts B and C, you might realize that Part A no longer fits. You can tell the AI, “Now that we’ve refined Parts B and C, let’s go back and make sure Part A flows smoothly with them.”

These are the same kinds of discussions you’d have with an intern or a team member. You’re the one in command — you lead the AI to achieve what you need.

We all know that what’s called one-shot prompting doesn’t really work. You have to guide the AI step by step to get exactly what you want. You might think, “That sounds like it’ll take more time,” but in reality, it doesn’t. You’ll still save time in the long run by knowing how to prompt effectively and steer the conversation in the right direction.

Even with CRISPY prompting, you’ll often need to clarify or redirect the output. If it goes off track, tell it, “No, that’s not what I want — do this instead.” Remember, large language models aren’t infallible. They can still “hallucinate” or fixate on a point you didn’t intend to emphasize.

When that happens, simply guide it back. It’s all part of the conversation. The key is persistence — because persistence leads to better results.

Stan Robinson Jr. 12:49

Exactly, exactly, the models are getting smarter. They are getting better at asking you to be more specific if you don’t start that way. it that the intern in our example is getting smarter and smarter,

Bob Woods 13:12

It’s like having an intern on day one versus the last few days of their internship — where you’re thinking, Wow, I could hire this person right now.

AI isn’t quite at that level yet, but it’s definitely progressing. It’s already come a long way, especially within the ChatGPT ecosystem — from version 3 to 4, to 4o, and now to 5. The technology continues to evolve rapidly.

Now, moving on to number four — something we’ve been talking about: one-shot prompts.

Effective prompt writing is important, but the dialogue that follows is just as critical. You need to ask follow-up questions and request elaboration. If something doesn’t look right, say, “Can you provide the source for that?”

When you’re asking AI to research, you can even limit it to the information within your conversation or any documents you’ve uploaded. For instance, if you’re generating a sales projection based on previous data, you can instruct it to use only that uploaded material — and not research outside of it.

That’s one of the mind-blowing things about AI — it’s truly amazing. But again, you have to take command. You need to be the conductor of the train, keeping it on the rails. That’s probably the clearest way to describe it.

Stan Robinson Jr. 15:08

And one of the things about keeping it on the rails is giving examples to the model in your prompt, it knows what good looks for you. whether it’s a an example of a blog post or plan, or whatever you can do, as far as giving the model an example, it will help tremendously.

Bob Woods 15:32

We just had a great question — What is CRISPY prompting?

Since it’s just me running the show right now, I won’t display the question on screen, but thank you, Doug Karen, for asking. As I reach down here, we’ve actually developed a book on this topic. We also have free eBooks that go along with it, titled Prompt Writing Made Easy with the CRISPY Framework. The CRISPY framework is a specific structure designed to help you write better prompts. When you use it, you’ll find that it provides built-in specificity and guardrails within the conversation, helping the AI or LLM stay on the track you’ve laid out.

You can find CRISPY on Amazon. I wasn’t planning to promote the book today, but if you want to buy it, that’s great — and if not, that’s perfectly fine too. We also have a free eBook version that I’ll make sure is included in the show notes so you can download it.

And if you’re watching live on LinkedIn, I’ll post a link in the chat right after this session that will take you to a page with more information about getting started with CRISPY.

Doug, thank you very much for the question — I really appreciate it.

Now, speaking of guardrails, this next point is a big one. Stan, I’ll hand it over to you.

Stan Robinson Jr. 17:14

Yes, one of the most brilliant things about the CRISPY framework is how comprehensive it is. Bob is the originator of the CRISPY prompt-writing formula, and as you’ve probably guessed, CRISPY is an acronym.

It begins with C for Context — which we’ve already been talking about — followed by R for Role, meaning you tell the AI what role you want it to assume. Then comes I for Inspiration, where you include any specifics or key facets of the problem, as well as S for Scope, which defines the format, length, or style you want.

Now, P stands for Prohibitions, which I’ll focus on in a moment. But before that, the Y in CRISPY stands for You. This step encourages the AI to essentially “interview” you — to capture your point of view, tone, and unique perspective on the query.

Prohibitions are one of the most brilliant aspects of the CRISPY framework because they tell the AI what not to do. Most prompting frameworks I’ve seen — and there are many — don’t include this step.

If you’ve read a lot of AI-generated content, you’ve probably noticed certain patterns: things like overusing em dashes (—) or starting sentences with phrases such as In today’s rapidly changing business landscape… AI tends to fall back on those clichés. With CRISPY, you can specifically instruct it: Please do not use em dashes. Do not start with sentences like “As we delve into…” or “In today’s fast-paced world…”

These prohibitions act as guardrails, just like Bob mentioned earlier. You can set clear boundaries such as, Please don’t do this, this, or this. And I’d say eight or nine times out of ten, the AI will get it right.

Bob Woods 19:20

When it comes to em dashes, you can tell the AI “no em dashes” multiple times, and yet you’ll still find one sneaking through. That’s why it’s important to always read whatever content it generates — especially if you’re using it for content development. We mainly focus on content creation, sales training, and sales strategy here on Making Sales Social Live, but there are many other things you can use AI for. Regardless of the purpose, you have to read what it produces — and not just read it silently. You should read it out loud. When you read something out loud, you’ll immediately notice if it doesn’t sound like you. If you start stumbling over the words, that’s a sign it needs to be rewritten — either by you or by the AI. At that point, though, you’ll probably prefer to rewrite it yourself.

This actually leads us into number six in our list — overcomplicating your prompts. Keep your prompts straightforward. That’s another great advantage of the CRISPY framework: it helps you stay focused and simple. When you use overly complex language, you’re just going to confuse the AI — even with the latest versions available. That leads to irrelevant responses and more back-and-forth corrections like, “No, that’s not what I want. Do this instead.” Simplicity leads to clarity, especially when you’re using CRISPY.

Again, CRISPY helps you stay on track. I can’t tell you how many times I’ve been on X (formerly Twitter) and seen people share their prompts — and some of them are pages and pages long. Occasionally, that level of detail is necessary if you’re working on something complex, but most of the time, it’s not. For most tasks, you need to focus on keeping your prompts as simple and as crisp — see what I did there? — as possible.

Stan Robinson Jr. 21:45

Yes. And if you do have a process that is somewhat complex, consider breaking it down into steps. In other words, ask for the highest level thing, first get that answer back and then ask for more in a multi-step process, rather than trying to do it all right?

Bob Woods 22:06

And then what I’ve done that several times myself, in that initial prompt, let the AI know that you’re doing, that this is going to be a multi-step thing, and that you will enter in one thing at a time, and then we’ll answer that, and then we’ll go to the next part. It’ll even prompt you. It’s , , I think we’re done. Let’s, are you ready for the next step? Feel free to do all that type of stuff. But to chunk it out that is brilliant, and it does work. I’ve seen it work myself. I think we’re on the last one.

Stan Robinson Jr. 22:41

The number seven is you. What we’re calling user intent, which is only ask AI for as much as you need. Don’t expect it to read your mind. If you want five bullet points. Put that in your prompt. Don’t say, give me a list of marketing techniques that will work in in the world of marketing today. we kept we keep talking about being specific. Make sure that you let the tool know. This is what I’m looking for specifically, because otherwise, as the over-helpful intern, it will go, as Bob mentioned, in a bunch of different directions, trying to please you you can make your life easier, and then get and save yourself time.

Bob Woods 23:45

Let’s add “overly caffeinated” to our description of that intern — because that’s exactly how it acts sometimes. Which is great… until it’s not. Too much caffeine is never a good thing. But still, that enthusiasm is valuable — you just need to manage and direct it.

When you avoid all these common mistakes in prompt writing, you can significantly improve your experience with AI tools. You’ll not only cut down the time you spend using them but also get much better results.

So, examine and experiment with these strategies. When you’re using AI, talk to it. I know it might feel strange the first few times, but trust me, it becomes second nature much faster than you’d expect. We’re all human — we like to communicate, to interact, and to help others. The same principle applies when working with AI.

I’m confident that once you start treating AI as a conversation partner, you’ll see a big improvement in the quality of the responses you receive — especially when you have our eBook handy, which details everything we’ve been discussing today.

It’s free at socialsaleslink.com/mistakes

You can scan the QR code on your screen if you’re watching live. And if you’re listening to this as a podcast, you’ll find the link in the show notes as well: socialsaleslink.com/mistakes. Anything else, Stan? Or is it time — like Brynne will be in a few hours — to bring this one in for a landing?

Stan Robinson Jr. 25:17

Wrapping up. The only other thing I’d mentioned is when we say, chat with it, that can be via text, which is what we’re looking we’re used to doing, but now you can also dictate to it, if you’re more comfortable literally talking to it using your voice, you can press the little microphone button, and now you can do that on your desktop in addition to the mobile device, you can do it on both

Bob Woods 25:40

We are. We are getting closer and closer ever to that Star Trek world where’s computer, help me. Blah, blah, blah, yes. And then if you haven’t seen the Star Trek episode where that happens, trust me, we’re, we’re, we’re very close to that already, and we’re nowhere near the 23rd 22nd century, or wherever it takes place. That’s all very cool. Whether you’ve been with us live or recorded via our podcast, thanks for joining us for this episode of Making Sales Social Live. We do this every week. Join us. If you’re listening to us on our podcast, as I mentioned before, which means that we’re taped and you haven’t subscribed already, go ahead and hit that subscribe or follow button to access all of our previous shows and be alerted when new ones drop. Besides these LinkedIn lives, we also interview leaders and experts in sales, marketing, business and many more areas. catch those episodes as. https://socialsaleslink.com/podcast is where we have more information about the entire making sales social universe. You can call it, since we’re talking about Star Trek and stuff. When you are out and about, be sure that you’re making your sales social Excellent. Thanks. Everybody. Have a great week. See you next week. Bye, bye, bye. Now you.

Outro:

Thanks for watching, and join us again for more special guest instructors, bringing you marketing, sales training, and social selling strategies that will set you apart. Hit the subscribe button below to get the latest episodes from the Making Sales Social podcast, give this video a thumbs up, and comment down below on what you want to hear from us next. You can also listen to us on Apple Podcasts, Spotify, Youtube Music, and Amazon Music. Visit our website, socialsaleslink.com for more information.