Hello and welcome to another episode of the Working Conversations podcast where we talk all things leadership, business communication, and trends in organizational life. I'm your host, Dr Janel Anderson.

Getting generative AI to help with your work is all the rage. But do you ever feel like talking to AI is a bit like playing 20 questions with a really smart but easily confused Genie. You ask for a great idea and somehow end up with a lukewarm suggestion for a spaghetti recipe. The secret sauce to getting what you actually want lies in writing killer prompts.

In today's episode, we're diving into the art and science of coaxing AI into giving you the answers, insights and content that you're after, no random spaghetti involved. Let's talk about how to turn your vague ideas into crystal clear directions that AI can follow so that you can finally get those magic results.

To make the best of generative AI, you need to know how to write great prompts using ChatGPT, Perplexity, Claude, or any of the tools that are out there, and there are new ones coming online every day, using them, using prompts is radically different than doing a Google search.

Let me give you an example, if you wanted to know something about sous vide, well, it's a cooking technique. I've talked about it before on the show in random places. But if you wanted to know about sous vide, it's spelled S, O, U, S, V, I, D, E, which is a French cooking technique, and you went to Google, you would just put in sous vide and see what comes up.

But generative AI is not like that at all. It doesn't just want a term to define. It wants you to give it a set of directions. So for example, just to see what would happen, I went to chatGPT and I typed in sous vide, and it came back with sous vide is a cooking technique where food is sealed in a vacuum bag and then cooked in a water bath at a precise temperature. This method allows for very accurate control over cooking temperatures, which results in food that is evenly cooked and retains its moisture and flavor.

And it gave me another sentence or two, but then it did its best to help guide me towards what it thought maybe I really wanted. I probably didn't just want a definition, although maybe I did, in which case I should have typed in a prompt that said, Please give me a definition of sous vide. But at the end of its response, it asked me back, do you have any specific sous vide recipes or techniques that you're interested in?

And then I went a little bit further. I asked, How do we pronounce it? I asked if I needed any additional equipment or specialized equipment, or what the process entails, and then it gave me step by step instructions. In fact, it gave me a six step process to follow, and told me exactly what tools in the kitchen I would need in order to make that in order to follow the process.

So what do you want to know? You need to build that into the prompt. Is it how to? Is it general research? Is it to draft a document or write some code? You have to let it know what you want it to do in order for it to do it well for you.

Now, I have an analogy I like to use, which I'll get to in just a second. I recently heard that the people who write the best prompts are those who are over 45 and I fall squarely into that category myself. Now who knows if it's actually true. I didn't see a study, a study cited, or any specific resources that went along with that age 45 but here's why, it might be true, the older we get, the more experience we have asking questions, a lifetime of experience, often first asking questions that are too general, too vague, or maybe asking questions that assume a certain line, a certain baseline of knowledge that somebody doesn't have. So we have more experience asking questions and refining our questions than somebody who might be younger.

You have to educate generative AI as you go now, here's the analogy I was alluding to. I like to think of generative AI, whichever tool I'm using, and it what we're talking about today. Writing great prompts doesn't matter if you're using ChatGPT or Perplexity or Claude or again, any of the other tools that are out there.

Here's how I want you to think about interacting with them. Think of them as your whip smart college intern or new employee. They are brilliant at some things, like doing research, but they don't have a lot of street smarts as it relates to what you want them to do or how you want them to do it again. You have to build the output into your prompts and educate it along the way. Remember, generative AI is just an algorithm. It still is just an algorithm. That's it. It is not inherently smart yet, so we have to help it along.

Now, if I was hiring a brand new intern or a brand new employee, of course I'm hiring them because of some certain subject matter expertise that they have, or I hope they have, but I'm not going to assume that they know how to do exactly what I want them to do in the way I want them to do it for my business or for the project I've hired them for. I'm going to have to ask them some questions. Do you know how to do this? Have you ever done one of these before? What would you need to know from me in order to be able to carry out these instructions, so you have to really engage it in dialog again, as if it were your whip smart college intern or a brand new employee.

So I'm going to get into a couple of examples, but before we do that, I want to give you a handful of high level guidelines. First of all, pay for a subscription. When you pay for a subscription, you can put your own content into the tool, and that content does not get indexed with everybody else's content. So if you had a draft of something that you wanted it to revise, and you have paid for a subscription, you upload your draft into your own little corner of Perplexity, or chatGPT or whatever tool that you're using, and then it doesn't get disseminated or worked into the broader algorithm and searching that broader base of knowledge that the tool is going to use. So you're carving out your own little corner with your content. So anybody who is working with their own or their company's own intellectual property absolutely must do that.

If you're just looking for some advice on how to write a better message in a sympathy card. Well then maybe you don't need a paid subscription, but anybody who's unique to do substantial, substantive work absolutely needs a paid subscription. So that's the first key tip.

The second key tip is to teach it about you. So if you have a specific writing style, or, you know, career path, a trajectory that you're on the line of work that you do, let it know who you are. Let it know how many years in experience you have, let it know what size company you work for, let it know about you, because the more it knows about you, the better it is going to customize and tailor its output specifically to someone just like you. So you do need to teach it about you.

So I have a little bit of a I mean, I like a contrarian point of view. Once in a while, I'm a little bit sardonic from time to time. Oh, my chat GPT knows me. It knows exactly. And even if I don't ask for it to give me a contrarian point of view, sometimes it just will, because it knows that I'm usually looking for that. So teach it about you. That's your second tip. It's going to do better for you if you teach it a little bit about your writing style or your art style, or whatever it is you're asking it to do for you, and a little bit about your personality as well. And the more you have it do work for you, the more it will catch on on its own. But the more explicit you can be, the faster you're going to get there.

And then your third tip is this, each new conversation is a brand new conversation. So once you've asked it for some asked it some questions, and let's say the sous vide thing. Okay, so when once I'm done getting everything I need to know about sous vide. Now if I want some, oh, let's say a catchy subject line for an email that I'm writing, I need to start a new conversation. I need to start a new thread, because otherwise it's bound to write me some email subject lines about sous vide.

Okay, so when I want to treat each conversation as discrete. It's just like if you had a conversation last week with a friend about going to a concert and now you wanted to ask them about looking after your kids while you go away for a weekend. You wouldn't pick up the phone and start the conversation with them from the point of view of you've just finished making concert arrangements. No, that was a discrete conversation. Now you're starting a new conversation. You're calling up your friend and saying, Hey, I have a favor to ask of you. There's this thing, and it's totally separate from that other conversation you were having.

So make sure, again, whatever tool you're using, you're always starting a new thread. Usually that's up in the upper left hand corner of just about any software tool you're going to use is basically like to start a new document or start a new conversation. It's absolutely critical in getting it to focus on the specific topic at hand and not bringing back into this new conversation themes and content that you were talking about at a previous in a previous conversation. So that's your tip. Number three is each new conversation is a new conversation.

And then tip number four, and this is the last tip I'm going to give you before we get into some specific examples. Give it feedback. Recently, I wanted a clever line to open a paragraph on something I was writing, and I was working the line, the clever line myself, and I had it like 75% of the way there. And that this is a great thing to do with chat GPT. Perplexity, any of the tools, is you've got something in draft form, and then you ask it to help you make it better. So I gave it my draft, and I told it what was missing. So it gave me, then a draft with the missing link filled in and three more alternatives. So I got four actual discrete, separate examples of what it was I was looking for.

Now, then what I did is I told it which one I liked best. I asked it this, then give me three more that were similar in or fashion or sense of humor or whatever. And it did that. And it turns out I like that first one best, so that's the one I ran with. But it is iteration. It's an ongoing conversation. Again, going back to like the new employee or the whip smart intern, they're not going to get it right the first time. So when you ask them to write something and they come back and you just most of the time, you're not just gonna go, like, okay, great, thanks. Here's your next work assignment. No, you're gonna say, Okay, well, I like, this is what I like your first draft here, and now I'd like you to take it back and iterate it. And there's this whole line of research that was missing entirely. So let's add some of that in and so on. So it's a back and forth iteration. You really need to think about it that way. It is not going to get it right on the first try, even if you write amazing prompts.

So now let's start talking about how to write some amazing prompts. So and I also want to encourage you to think about using generative AI for a larger project, so something that's got multiple steps, and again, you want to be in dialog with it.

So here's something I'm currently working on. I am currently working on a competitive analysis of my business against my competitors, so that the people who work for me can better understand who I am in the marketplace, what I deliver, and how that's different from my competitors. So when they're positioning me, whether that's in website copy or sales call with an association about bringing me in for a keynote, that they really understand how I'm different from the competition. So this is a really important piece in my business right now.

So here was my prompt to chat GPT again, thinking of it as my whip smart college intern. I literally wrote this I'd like to do a competitive analysis. Is that something you can help me with? If so, what questions should I ask you so you can do the best research. So really, I'm asking it to me questions, in a sense, so then it spit out 15 questions and some questions that would be helpful for me to share with it so that it could give me back the best competitive analysis.

So it wanted some obvious things like, what's the main purpose of the competitive analysis? What's the aspect of the competition that you're the most interested in? And it also asked me a handful of things that just weren't that helpful to my business as a keynote speaker. So one of the questions it asked me was about market share. Now I don't really know market share.

Now, if this was, if I was like a chewing gum company or a soda brand, then maybe I would have data on market share. But the world of keynote speaking is a little bit elusive, and I don't know exactly how many keynotes on the same topics that I speak on our give each year. So to come up with market share really wasn't a useful category for me.

Also, it asked about financial performance, so it just if I was interested in financial comparisons. Again, much like the market share, I don't necessarily know all the exact same price points. I mean, I know I don't command the same fee as a former president or a Hollywood actor, but likewise, I don't command the same fee as somebody who's just starting my business either. But so financial performance is was not necessarily a key metric that I thought that chat GPT would be able to give me feedback on or find the right research for now. It also concluded with, you know, by answering these questions, I can tailor the competitive analysis to your needs, ensure that it provides you a valuable insights and so on.

But then, you know, as I got thinking about it, I thought, well, it doesn't know from what I've given it so far if it is writing a competitive analysis on chewing gum or soda or real estate or automobiles. So I needed to refine my prompt in order for me to tell it more explicitly what I wanted it to research. So I followed up with this prompt again in the same thread, because we're still having the same conversation here. So my refined prompt was this more specifically I'd like to do a competitive analysis of keynote speakers who speak on the future of work like I do. Are there different questions you would want me to answer in order to provide the best competitive analysis possible?

And again, it spit out a series of questions, and a number of them were different. There were a few of them that were the same as the other competitive analysis because some of them were just generic enough that they applied. But let me just give you a couple examples of different ones that it asked then that were different from before. It said, What is your unique value proposition as a speaker again, for me to let chatGPT know what my value proposition as a speaker is it can then better find other speakers that have a similar value proposition than as I do.

So that it can make an accurate comparison, so that it's comparing more apples to apples instead of apples to horses. It asked about target audience and market positioning. It asked about speaking style and delivery. So again, for it to find my best competitors, it needs to know my speaking style and my delivery style and what's different about me on stage.

It also wanted to know about marketing and online presence so that it could find other speakers, let's say that have a podcast or that are active on various types of social media. It wanted a client base and testimonials. It wanted visibility and thought leadership. It wanted content formulas and channels. It wanted content formats and channels. So are we active on LinkedIn? Do we have a podcast? Are we on Instagram? Do we create reels, all of those kinds of things? And then I also thought this was very key for me.

Specifically, it asked about my geographic focus and my market focus. So do I want associations? I'm based here in the greater Minneapolis, St Paul area. Do I want work that is closer to home? Or am I willing to travel across the United States or maybe even worldwide. So it asked a much better set of questions when I gave it more information about who I was and what I was looking for.

Now here's how this gets even, even better. So I answered the questions that it asked me about my business, and now it can go out, taking that information and then cross referencing that with all of the data that it searches, it's going to do a much, much better job at giving me an accurate competitive analysis of my competitors, of my actual competitors, instead of just general, you know, speakers or general anything, it's going to give me a much, much better picture of who my competition is, and that's going to help my employees do their jobs better.

Okay, so you see how this is going. I'm not just asking it one question and expecting it to spit out a competitive analysis report for me, I'm going to engage in dialog with it, much like if I have that whip smart college intern, I might ask them, Hey, have you ever done a competitive analysis before? And they maybe would say, No, I haven't. And then I would describe to them, well, a competitive analysis generally looks at these types of things. Is this research you'd feel comfortable looking for and finding for me? And they would probably say, Oh yes, thanks for the guidance. Now I'll go find that.

Okay, then when you when you found that, then, then let's come back and talk about what the format of the document look would look like, and so on. And so likewise, you're going to be in an iterative conversation with your generative AI tool in order for it to do the best work for you, okay? And in fact, this one once it asked me all the questions that it wanted me to answer for it, it said, By answering these questions, I can create a detailed competitive analysis that will provide actionable insights, helping you refine your strategy as a keynote speaker in the future of work space. Okay?

So in short, I got a much better set of questions for me to answer, to provide it with the information it needed before I asked it to even write anything. Do you see how this works? And once you fought, once you sort of get into this mindset, it's magical. So let me give you a second example.

So this example comes from a recent keynote that I was giving an in person keynote, and somebody asked me after the presentation. And I, you know, I often do, like book signing or Q A or whatever, and I had made some passing reference the keynote itself was not about, uh, AI, although there's one of those in the work. So if you need a keynote on AI. Hang tight. I've got one coming for you. But in the meantime, I had made some passing reference to using generative AI in the process of your workflow and how it can really enhance your productivity and all this stuff.

Well, this person pulls me aside on the break or after my keynote on their break, before they went off to do the next thing, and asked me about how I would go about having generative AI be like a ghost writer, and write a book. So he wanted to write a book, and he wanted this to be an autobiographical work of fiction. So he had a specific incident, and he didn't tell me what it was, so I'm not at liberty to say, but he had a specific incident that happened to him when he was a teenager. The impression I got was a very major event in his life that really developed his character into who he is today. And he is, I believe, in his late 50s or early 60s at this point. So generations ago, something happened, and he wants to write this, this, you know, novel based on his own life, about it. And so asked him, What approach would you take?

And he said, Well, he thought he would just sit down, and he does have a paid version. He told me that, and he thought he'd just sit down and say, like, you know, write a novel about a boy who has this specific experience when he was 15. And so I took great pause, and I told him that was probably not going to get him the result he wanted.

So he asked, How could he go about it instead? And so I said, this was my like, verbatim advice to her. I said, instead of asking it to write the whole book, start by asking it to write a plot synopsis. Write a plot synopsis of a 240 page book, novel in which the main character, a 15 year old boy, has this life changing experience, and then tell it what the life changing experience is. And then when it spits out the bot synopsis, and I also told them, like, give it a specific number of words. I want a 500 word plot synopsis of a novel in which the main character, a 15 year old boy, has this life changing transformative experience. Then take that 500 word Plot Synopsis and ask it to revise it, because there's going to be some things in there where it missed the mark, and there's going to be some things in there where it's spot on.

And so you're going to tell it, I love the part about the role the family played, but there was no formative teacher in his school, so take the teacher part out and double down on the role the family played, whatever I mean, whatever it is, but so you give it the feedback, and I told him, do a bunch of iterations on the plot synopsis until the plot synopsis is really nailed.

Once that plot synopsis is really nailed, now you're going to ask it for the synopsis of chapter one, give me a 250 word chapter synopsis of chapter one, and maybe there's a specific thing you want to have happen in that chapter where the 15 year old boy meets the new neighbor who's moved in next door, or where the 15 year old boy gets reprimanded by his father, whatever it is that there must be some piece that's really important to get there in chapter one that sets the tone for the novel and the character development and all the things.

So tell it specifically what that is, and then you're going to revise again. You're going to say it's, you know, so revise the synopsis a few times after you've had the tool revise the synopsis a few times, then you're going to ask it to actually write the chapter, and you're going to be specific you want a certain number. And probably I'd go word length, not pages, because depending on how the book is formatted, the page numbers is going to be different. So I would give it a word length.

And so you could look up, you know, and also, is this young adult fiction? Is this adult fiction? And you can find what the average chapter lengths are in word count that's readily available on the Internet. In fact, you could use Google or chatgpt, one of the generative AI things to do that part of research about the format of the book and generally the pacing and so forth, and also the arc of the plot all of that.

But then you would repeat this process with chapter two. Plot Synopsis, revise, revise, revise, okay, now write the chapter. Okay, this worked. I want more dialog. I want less dialog. I want more action. I want more suspense. So you're maybe it's humor, but you're going to tell it the tone of the writing as well. Again, once you have the chapters written and rewritten and rewritten within chatgpt, then it's time for you to put the finishing touches on so you are going to take whatever chat GPT spits out, or, again, whatever tool you're using as its final product, and think of that as your then you're going to go through it, and you're going to continue to do some revision and put on those finishing touches so that it is truly your intellectual property.

Okay, now in the process of giving it your story to begin with, and again, remember, in a paid subscription, and we're working in your tiny little corner of the universe inside of that paid subscription, so that it's not necessarily putting what you're creating out there for others to index into their work as well. So that's going to help create it being yours, but also to really make it yours.

Once it is done, you still will edit and draft and revise and so on to really, really make it yours. Now, if he were to go about creating his novel in this process, I think he might be able to come up with a really good novel that is he and chatGPT or perplexity, or whatever tool he ends up using in combination with each other in dialog. Now for the book, remember, I was saying new conversations for the book, if one project with one thing, I would keep that all within one conversation, so that the character development, the plot development and so forth does not get lost from a different thread.

So this would be one big, long conversation that maybe spanned many weeks, perhaps even months, as you work on crafting that book. So that's how I would go about writing a novel. I don't have a novel in me about something formative that happened to me at 15, so it is not my work to be done. It’s somebody else's work, but that's how I would go about it.

So there you have a competitive analysis and a novel based on a formative event in one's youth. So again, I really want you to think about your generative AI assistant as the new employee, or the whip smart intern who is excellent at researching things but doesn't have any street smarts whatsoever about work life, or your industry, or what your boss is looking for in the report that you're asking, that you've been asked to write, and that you're asking AI to help you with. So you have to see what they know, so that you know what kind of directions to give them, and then you can give them very, very specific directions and and feedback to get better and better iterations, all right. So there you have it.

Whether you are trying to generate new ideas, crafting engaging content or just having a little fun with AI, it all comes down to the prompt. A well crafted prompt is like giving AI a clear map instead of letting it wander through a maze. So the next time you're typing away, remember, the more specifics and creative you are, the better the output.

Keep experimenting, keep refining, and who knows, you might just find that AI is a better creative partner than you ever imagined. Happy prompting my friends!

Remember, the future of work is not only about technology, it's about the values we uphold, the communities we build, and the sustainable growth we all strive for. We need to keep exploring, keep in, keep envisioning the remarkable possibilities that lie ahead as always, stay curious, stay informed, and by all means, stay ahead of that curve.

Tune in next week for another insightful exploration of the trends shaping our professional world. Now, if you learned something in this episode or you simply enjoy this content, please subscribe to my channel on YouTube, subscribe to the podcast on your podcast platform of choice, and follow me on social media, or do all three of the above. These are all excellent, no cost ways for you to support me and my work over on YouTube.

Make sure you hit that little subscribe button and knock the bell so that you get notified every time there's a new episode out, you'll find @ youtube.com/Janel AndersonPhD, wherever you're listening or watching, please leave me a review or share a comment. It helps other listeners find me, and it lets me know you're along for the ride. Until next time, my friends. Be well.