Using AI Like a Human: Strategies for Content Creation.

Hello. It's 11 AM on the West Coast, 2 PM on the East Coast, and wherever you are, thank you for joining just me. It's time to grab your favorite caffeinated beverage because we're gonna have a coffee break.

I'm your host, Alex Jonathan Brown, and this is normally where I introduce guests, but not today because there are no guests. It is just me.

We are talking about how to use AI like a human. Did I already minimize all of my windows and have to open them?

Yes. Was I able to do it like a cool smooth operator? No.

1 of the things that we are going to do today is I'm gonna give you a walk through of how I use AI, and then I'm gonna walk through an example that 1 of y'all give me. So if you have any ideas for the kind of stuff that you wanna do with AI, throw them in the chat right now.

We've got people from the Iron Horse team in there. I'm gonna keep an eye on it.

And if you are in there, say hi. Let us know where you're coming from, and let me know that I'm not just broadcasting to the void because I have no 1 to do comedy bits with today.

As I said before, it is just just little me. Hi, Lauren from Peru.

We are talking about using AI as a human. If you've been on the Internet and doing anything in the marketing space for the last year plus, You've heard probably the ongoing conversation about artificial intelligence and, large language models and and all this stuff, whether that's chat GPT or any of the, I would say, 100, but really 3 to 4 other big main competitors.

Leah in Minneapolis. Rodney, thank you for trusting my comedy. And if you're watching this on the replay and why?

Why aren’t you joining us live? This is where the fun happens.

What I wanna do today is really walk through how I would suggest that you use these tools. There is an ongoing conversation about whether or not these tools are good, whether or not you should use them, whether or not they're going to bring about the downfall of civilization.

I'm not gonna answer any of those. This is a conversation that assumes that you were going to use them, whether that's your boss saying, hey.

I've heard about these. This could save us money. Let's get on it.

Or whether you're a person who just likes tech and likes to play around. That's kind of the table stakes at this point is eventually you're gonna use some of these if you're not using them already.

Let's look at how I use them and how I think you should use them too. Like I said before, I'm Alex Jonathan Brown.

I'm a senior content strategist here at Iron Horse. That's my face.

My face is also down here in the corner. But if you ask AI to look at that image and describe it, you get something like this.

Right? It's a portrait photo, portrait photography of a bald man with a full medium length beard showcasing hues of black and gray. It's frankly rude AI.

The background is out of focus, textured appearance, etcetera, etcetera, etcetera. You can read a little bit, for yourself there.

But then if you take that prompt and add it back into AI, you get this guy who's, like, in the ballpark of being me, but very much not me. This is what if I had become a yoga instructor instead of a marketer version of me.

And if I would take this AI generated version of me and, you know, update my Slack profile picture and my LinkedIn photo and say that this is me, that would be weird because it's obviously not me. Right?

My skin's not that great. I wear this hat way more than that guy does. And so that's kind of where I wanna start this conversation.

You, dear listener, viewer, are exceptional at being a human being. You're so good at being a person.

It's what you do almost all the time. And AI is real trash at it.

Computers are good at so many things, but, like, being a person, very low on that list. There's a quote from Arthur C. Clark.

It says, any sufficiently advanced technology is indistinguishable from magic. Right?

It's like if you think about the first time that, let's say, someone encountered, like, a telephone in the wild, the idea that you could pick up a receiver and talk to somebody who was on the other side of the country, like, that felt like magic. But it's not.

It's science and wires and, like, all stuff that we can make make sense. Kat, do I have the quote wrong?

If so, that's great. I have my own version of this quote. Selling your technology as that magic is a great way to make money.

So it's not my own version. It's like a little old director's cut bonus addendum.

But this is where I think we are a lot of the time when we're talking about AI. Like, the companies that are selling AI have a literal vested interest in getting you to think that it's magic because this is the way that it's set up to work.

Right? You have a prompt at the beginning. That's what you type into your little text box.
Some magic happens, and then you get an output, which is, according to the marketing anyway, the exact thing that you were looking for. Right?

I asked it to write me an email, some magic happened, and then I got an email out on the back end. The problem with that is at least in this context.

Magic does not exist now. The first time I gave this talk, did I have a lot of people who were like, actually, my crystals would like to have a word with you. I know I spelled it without a k for a reason.

I will entertain that other forms of magic exist, but specifically this magic in the middle of this output, or in the middle of this model, I'm very skeptical of. Also, magic of Christmas, I'm 100% on board with.

What's really happening in this magic space that isn't magic is math, which on 1 hand is great because AI and computers excel at math. And I also, like, I also don't want this to come off as me saying that because it's math and not magic, there aren't use cases for it.

There aren't any ways we can use it. We absolutely can.

This is a wonderful image of a great barnyard family. I assume the barn's right behind the camera.

Right? And this is AI generated, and, like, it's pretty good. The prompt that I gave, chat gpt that then used the DALL-E image generator to generate that image was to generate an image of a horse, a goat, a sheep, a dog, a chicken, and a calf.

And from that prompt, we get back to this image, which again, we got a horse. We got a dog.

We got a sheep. We got a chicken. Maybe not a calf.

Maybe there's a calf in the background. But how tall is that dog, or how big is the sheep?
And this is where we start to get things this is where things start to get a little tricky. Because when you start looking at the AI image more, you start to realize it's more than just, like, some weird sizing that's happening on this barnyard.

That rooster's face is not that's not what a rooster's face looks like. That sheep down there has no eyes.

There's some kind of Lovecraftian nightmare in the background that's neither, well, it's not a cow. It's something else.

And down in the bottom left corner, it's either a fluffy 4 legged chicken or a sheep that's been decapitated, and either way, I don't like it. This happens for a really simple reason.

And, again, I think this has done a pretty good job. It's done, like, an 80 80% job of giving me the barnyard team that I wanted.

The problem is that the AI model doesn't know what a chicken looks like. And so when I ask it to give me a chicken, it doesn't even have, like, a kindergartner's understanding of a chicken.

Right? What it knows is from my dataset, I've seen that things have been labeled chicken 73% of the time. Again, it's math, and I'm, like, I have a minor in creative writing.

Math is not my strength. Right? But it's also unfair for me to say this is only an AI problem because it's not.

Great British Bake Off 2022 episode 4. If you are a big fan of the series, you may remember it as Mexican week.

There were 3 dishes they had to make on Mexican week. They had to make, pan dulce, traditional, Mexican tasty baked goods.

They had to make a tres leches cake, another traditional Mexican baked good. And they also had to make tacos.

And if you're wondering, that doesn't seem very great British Bake Off of them. I do largely think it was because the producers could only come up with 3 Mexican dishes, and so they had to do tacos.

And as part of the recipe for making tacos, if you're not familiar with the show, they always get very bare bones recipes and kind of have to do their best with them. As part of that recipe, it said peel an avocado.

And so as we're watching now, let's all take a second. Just visualize the process of peeling an avocado.

What does that look like to you? What do you you're in your kitchen.

You're peeling. Okay. Great. Did what you visualize look like this?

I'm guessing not because that's not how you peel an avocado. So what this woman did, I'm assuming, was think, “Okay, I know what peel means, and I've peeled a lot of other fruits and vegetables before. An avocado is just the same thing. So to peel an avocado, I'm going to do the same thing that I've done before, just to this.” The problem, of course, being that this is a terrifying way to peel an avocado.

And the way that you do peel an avocado is you cut down the thing and then you pull it up and then you slap the pit with the little corner end of your knife and it pops out and it's very satisfying. This is a good way to lose a finger or your hand.

So humans make this same mistake too. And, yes, Kat, you do always have to have to immortalize that avocado for the gram, as the kids say.

But this is why it's very important that we realize that, like, the magic's not always magic, and that means we have to add another step to this process too. So it's so important as you're working with AI to remember that you're really good at being a human being and AI is very bad at it.

And that means you have to find the places where you can inject the things that you know and the parts of you that are human into this process. So, normally, everybody's on board with the idea that that happens in the prompt.

Right? You write a prompt out for AI, and then you just let it go from there. What you should be doing at the very least, before we get to any of the cool stuff that I'm gonna show you here in a second, is you should be adding an edit stage to that process.

Right? So you should be looking at that farmyard photo before it goes out into the world and be like, hey. That chicken is a nightmare, and it's gonna scare children if they see it.

We need to do some more work on this. That's a weird way to peel an avocado.

Let's do some more work on it. This email feels like it's been written by a robot, and it says unlock 6 times.

We need to do some edits. Right?

And so this is, I think, it like, this is, for me at least, the minimum of how to work with AI. You do have to add that edit process.

You can still use it in this way to get some stuff done faster. If you're really, really stuck, you can just be like, I need some help, get started, and then I'll do some edits at the end.

But I don't think it's the most interesting or exciting way to do it because there's still a big chunk of magic in there. And if you'll remember from a few slides ago, magic does not exist.

So what we can do instead, and this is where if you've been waiting for the cool AI hacks, close your other windows and tune back in because we're diving into it, you can push the math part of this model further up, and you can create all this little space down here. And what we're gonna do with that space is just drop the whole model back in again.

We're gonna use AI to tell us how to use AI better. So this is an example that may or may not come from a very real life situation.

I'm hosting a webinar tomorrow, and I need to promote it. What information would I need to give you to write the best email possible, And what outline would you use to write it?

I am asking these AI models to give me the info that it thinks it needs to write me a better email. Before it does any work, I wanna see how it's gonna do that work.

I am asking it, tell me how you would peel an avocado. So if you're gonna do anything weird, I can stop you before you do it, and I can guide you in the right direction.

And when I gaveChatGPT that prompt, this is what it gave me back. A whole list of things.
It wants the title. It wants the date and time. It wants the platform link.

It wants the speaker information. And that's a lot of stuff and more stuff that I'm gonna wanna include in an email. So just going through this real quick, I'm like, these 6 things are good.

I'm gonna want those in there. The 6 on the bottom, I'm less interested in, but I can fill in the top half.

I can give it that information, and then we can start working. We can start getting close to an email.

And this is the outline it gave me. So subject line, we're gonna do a greeting.

We're gonna do an introduction. There's gonna be a detail section where we talk about some details. We'll do a CTA.

We'll do a postscript at the end because marketers love a postscript. Throw a link in there, and your CTR is gonna go through the roof if you ask, frankly, all of us.

It's something we say a lot. But I don't want it to do all of these things either.

Personally, there are times when I really, really hate using the, hi, Pulls the name from chat. Hi, Charles.

It feels weird. Honestly, Charles, that probably felt a little weird too. Sorry about it.

There are times when having content personalized to you, the vibes are just bad. A lot of other stuff I don't wanna know too.

Don't talk about networking. Although, if you wanna network in the chat, go nuts. Somebody already threw their LinkedIn in there.

It's Alex from Indianapolis. What's up from an Alex from Fort Wayne? Include technical requirements, postscript, etcetera.

I don't don't do that, please. So then I I get back in the chat gpt window, and I just say, remove the additional information and PS sections from the outline.

Before you write an email, get rid of those. I'm not interested.

Additionally, don't use personalization. Don't create a sense of urgency. Ignore anything I didn't give you and remove it from your list.

So just letting it know I'm not doing anything. I didn't forget anything by accident.

I forgot things on purpose. If I didn't give it to you, don't do it. And then we get this email back.

Right? So some of it's great. That's kind of the stuff that I have in green.

I like saying we'll explore dynamic ways to leverage AI. Right?

Some things I don't like, it's like, is it a unique session? I hope so, but I don't think that's necessarily a selling point.

Thank you for your attention. Will be hosted. Feels a little funky.

Anyway, so if you've ever done the thing where you just ask chat GPT to write you an email, I think this is much less meh than what we would get without it. Right?

We're giving it the specifics that it has asked for in order to get closer to the outcome that we want. But, again, magic doesn't exist, and we can keep pushing this model further up.

So now based on that first kind of second version of the email, we can go back to it, and we can start to get really creative. And we can say it's 2:48 on a Wednesday, and I'm really in a flow state or losing my mind.

Here's a little bit about me. Write it closer to the tone that I write in.

Right? I'd rather die than be called bland. And so then it says, alright.

I'm gonna I'm gonna math that, and this is what we get back. And I don't know.
Like, there's an there's an emoji in the subject line. That's never my favorite, but there's a line there that I highlighted.

I'm Alex Jonathan Brown, but around Iron Horse, they just call me the guy who somehow makes content strategy as exciting. It's a double shot of espresso.

Now a couple things about this. 1. No. They don't. 2. Anybody who knows me knows it's like a large McDonald's cup of Diet Coke way before it's ever a double shot of espresso. But, also, generally, this is just really cringe.

And, like, I think if somebody's reading that in an email, they're gonna know it's really cringe. Right?

There's lots and lots of cringe. It says come to this shindig.

Right? I'm not gonna I'm not gonna call a webinar a shindig almost ever. So maybe we went too far with tone.

Right? But then, like, as I read it and, like, really got into it, there's this sentence. We're talking about leveraging AI not just to churn out content, but to do it in a way that keeps you sounding like the vibrant, intriguing human being you are.

I mean, I don't hate it. Bring your curiosity, your passion, and maybe a snack.

Like, alright. Like, I mean, it is a marketing email. That's not that bad.

And how to keep our humanity in a world run by algorithms. If know me, this is the most me thing I've ever said, and I didn't say it. ChatGPT did.

And so I look at, like, those 3 lines that came out of this process. And if you're like me, there's always some uncertainty about whether or not anything I've been involved in is good, but I'm like, maybe this is good.

But I know it's not so embarrassing that I can't ask my boss about it. Right?

And it's so much more interesting than where we would have started if we would have just asked ChatGPT to write me an email. And so you can keep working this process all the way down.

You can always ask ChatGPT, what do you want from me? And then give it that.
Or you can ask, what do you want from me? And then I'm gonna tell you why you don't want that.

And the cool thing with ChatGPT is you don't have to ever explain yourself. You can just say, no.

We're not doing that, and moving on to the next thing. This is also the time if you have an idea, if you want me I mean, literally, if you want me to write an email for you, throw it in the chat in the next 30 seconds because we're wrapping this part of the presentation up.

But I think the thing that's so important to keep in mind, right, is that we are at a stage where so much of this AI content is out there and is being used and is being used in a way that doesn't follow this model and maybe doesn't even have an edit on the end. Right?

If you wanna see what it looks like, ChatGPT is just allowed to make all the content at once on its own with very little guidance whatsoever, Like, check out Twitter or threads now. So much of that content is, we'll say, AI aided in a way that feels a lot like being AI aided.

And as that content grows and as we get more of that content, people start to understand what it feels like, and they start to know a little bit more of, like, oh, you didn't write that. That's AI.

And now here's the thing. If it works, amazing, and I don't care. Right? 

If you're sending out AI emails that are converting, if you're using AI to to shorten the amount of time that you have to spend thinking on a thing, that's amazing. And if it works, it works. 

I think we're maybe getting you afraid or to a phase where the amount of stuff that works is getting lower, and the amount of work we have to do to get AI generated content to work is raising a bit. But we have a suggestion from the chat. 

So because the future is magic, I'm gonna click a couple buttons here, and we're gonna swap over to a chat GPT window. The future. 

We did it. Okay. So let me pull my chat around here. So the suggestion we had was, from Jonathan, and its takeaways and next steps after our AI webinar. 

So let's just start with that. So I know that there are a lot of people who think you don't need to talk to the robots like they're people, And I think that's probably fair. 

I think you don't need to. But, also, what if we do? 

I would rather the robots think I'm nice, which I guess means I hope they didn't read some of the promo emails that we sent out for this webinar. But can you write me an email with takeaways and next steps after, webinar about AI. 

Okay. So this is the base. This is just a stock model on the showroom floor. 

Do what you're gonna do, ChatGPT. So let's look at what we've got. 

Right? At the top, we've got the subject line, thank you for attending our AI webinar, key takeaways and next steps, a spot for the recipient's name. Thank you for joining us for the webinar on unlocking the power of AI in business. 

Now it made that up because I didn't give it a title yet. This is why editing gets really important because that's just a lie. 

We appreciate your participation and hope you found the session insightful and valuable. Below are the key takeaways. 

Sure. Understanding AI's potential, AI integration strategies, overcoming AI adoption challenges, ethical and responsible AI use. Maybe that's a really good, maybe that's a really good, recap of what you talked about in your webinar. 

I'm not sure. ChatGPT also isn't sure. Just made it up.

Next steps, schedule a consultation. So you can book a 1 on 1 consultation with our experts. 
You can access webinar resources. If you wanna download those slides, we can throw you a link to do that. Join our AI community. 

If you've got 1 of those, maybe that's great. Provide feedback. 

But, again, we're at language. Like, we're excited about the possibilities AI brings and are here to support you in your AI journey, which is maybe fine. 

Right? And maybe it's possible that we go through here and we're like, oh, if I just update this and I change this to reflect, like, what I'm doing, like, with 15, 20 minutes of work, you can start from this core and get to a place that's interesting. Right? 

1 of the questions that we get a lot is, like, if you're gonna have to go do this editing or if you're gonna have to do all this prompting, like, wouldn't it just be quicker to write the email in the first place? And it would. 

1 of the mistakes that I think people make and let me go let me go back to my drama. We'll go back to the ChatGPT team in a second. 

1 of the mistakes that I think people make is assuming that AI is gonna save you time. It's not. 

What AI is really good at doing is saving you brain cycles. So if you've been at that place where it's 2:48 in the morning, like, when I was working on those emails, or you're just like, my brain is fried.

I've been in meetings all day. I don't I don't know how to write an email right now, basically.

AI is really good at saving you those cycles and saying, here's some garbage, but you can make it better if you want. And sometimes that's all it takes to get you going again and to get you started.

But back to our example. So we've got this, and now I'm gonna just open up a new window. The thing about ChatGPT is that it doesn't, you can open a new window and start fresh.

So we're just gonna ask it. I'm working on an email after a webinar.

What info do you need to help me write it? Okay.

Title topic, target audience. This stuff should look pretty familiar from our last 1. And we're just gonna plow through these super fast.

So, title. Yeah. It's dope, and you should use it. Maybe.

Target audience. Marketers. Or we can get really specific and say marketers at B2B enterprise companies. Key points.

Let's see what it gives us for key points. I'm gonna leave that update.

Call to action. Contact our sales team. Additional resources, we'll say none, and contact let's say contact me at ironhorse.io.

And so already, we're getting stuff that I think is more interesting. And I didn't do anything about tone.

I didn't do anything about vibes. I didn't do anything about how I wanted to sound, but thanks for joining that title.

Let's talk next steps is better than what we had before, which were key takeaways and next steps. Let's talk next steps is more compelling instantly just because we gave it a little more information.

Right? We hope you found it insightful. This is all still pretty boilerplate.

So we're just gonna ask it to rewrite it again. The tone that's going to appeal to elder millennials.

As an elder millennial, I'm always very interested in what the world thinks that we think. Okay.

Let's chat next steps. I would like to talk better. But thanks a ton for joining our webinar.

We hope you're as pumped about the future of AI and marketing as we are. And now we're at the stage, like, just with doing those 2 steps, saying, give me a boring version, here's some info, and then here's a little tone.

We're getting there so quickly. Right? Like, whether or not you think it's a good idea to use, we hope you're as pumped about the future of AI and marketing as we are.

Like, that is very much elder millennial language. Like, nailed it.

Right? Elder millennials are in their forties now. It's a problem.

AI isn't just a buzzword. It's here to make your marketing smarter and your life easier. Hit up our sales team.

Again, maybe that's not the fit for your brand, but it is, like, a choice. If you're using these prompts correctly, if you're injecting the parts of you that are human into that big chunk of math, right, All of these little purple spots are your chance to say, you're gonna do this better.

I'm gonna make you. Right? Like, we get closer to sounding like a human and still using this technology.

Again, do I save time? Do I like the email better if I turn on my writer brain and I write the email?

Yeah. Of course. I'm a big fan of my own work, but there are times when it just doesn't happen, and now there are things from this that I can steal. Like, we're all about helping you level up your marketing game with AI.

Like, that's perfectly usable, and I can update it if I want. I can make it match my brand tone more if I want.

Looking forward to helping you crush your marketing goals with some AI magic is a little too far for me. Right?

But that's the edit stage. That's going back and going through it. So we're just gonna push it too far, and then we're gonna be done.

Keep that tone, but let's say we're working just in the waste management industries. Right?

And let's see what this comes up with. Right?

So the only and this is where I'm like, AI, you're letting me down, man. The only change it's made is to say we're gonna say waste management goals.

Right? And then so guess what we're gonna do? If you're not aware, you can always do this.

You can come up here, hit this little pencil. And I think probably what happened is that I told it to keep the tone, and so it just changed the content.

So we're gonna say adapt that tone and see where we come up with here. Memory updated, which is always a very dramatic thing.

You know what? It's not gonna give me trash metaphors, which is what I want. And this is a life lesson, and then we'll be out of here.

I know I've ran way long. It is unfair of you whether we're talking about a person, a piece of technology, somebody who works for you, your life partner, whatever.

It is unfair of you to ask for help from someone who doesn't know how to do the thing you're asking them to do. So in my head, I'm like, what I think this webinar that we're doing now needs is some great trash puns.

So I'm gonna do this whole waste management thing. It's gonna be awesome.

ChatGPT doesn't know that. It might know how to do it, but it doesn't know how to read my brain. So I'm just gonna add another sentence.

Include some hilarious trash puns. Again, don't don't do this for your emails. This is an example.
Okay. Okay. We got 1. Thanks for joining our webinar.

Let's trash talk next steps. Horrible. Horrible.

We're all about helping you turn trash into treasure with AI, and, like, that's the light bulb. Right?

Because, I mean, that's sort of what this whole talk has actually been about, but that's, like, good enough to use. And that's the thing, I think.

We're gonna wrap up. The only way this works, the only way that AI is good and helpful and useful to your marketing efforts is if you figure out the places that you can put your humanity into it. If you can say at every 1 of these little light purple blocks, right, that I'm gonna massage this and make it better.

That's when it gets better. And so, no, you're not, you're not probably writing to waste management professionals. Right?

But you are writing to an audience. TradGPT will never know your audience as good as you do.
It's up to you to find those places that you can add that, put that in. I ran incredibly long.

Sorry, Ellen. Ellen's 1 of the people who helps produce this. I'm Alex Jonathan Brown.
The next time we do this, we're gonna have guests. I'm gonna have someone else to talk to.

But thank you so much for coming. Thank you to all of you who stuck around.

We'll get the recording link, I think, comes out via email, Kerrigan. So we'll get that to you shortly.

It doesn't exist yet because I'm still doing it. But this has been delightful.

Thanks for listening to me chat for a little bit. If you have questions about more of this, if you wanna follow-up, if you wanna have a conversation, my personal email, alex.brown@ironhorse.io, email me directly.

If I can't answer it, I'll get you to the right person on Iron Horse who can. If you wanna learn more about what we do, including blogs, infographics, a bunch of other stuff about AI, that's at ironhorse.io.

It's our website. We're doing this again. Ellen, if you've got it, throw the info and maybe even the registration link for next month's coffee break.

I'm so excited about it. Our guest, Josh, is 1 of my favorite guests. There it is.

It's magic how this works. But until next time, I'm Alex Jonathan Brown. Click that link, register, and reopen Slack, I guess.

Coffee breaks over. Let's get back to work. Bye, everybody.

Thanks again.