Podcast: Play in new window | Download
AI moved fast in 2025 – and not every shiny new idea was worth chasing. In this solo episode, Rich Brooks shares real-world lessons he learned while experimenting with AI across content, workflows, and client work, including why being “AI-first” is often the wrong goal. If you’re a business owner or marketer trying to use AI thoughtfully – without sacrificing quality, trust, or your human edge – this episode offers clear, practical guidance you can actually use.
Stop Chasing AI Hype: 7 Lessons That Actually Matter for Small Business Marketers
I spent 2025 deep in the AI trenches—building custom GPTs, testing every image generator I could find, and burning through roughly $200 a month in subscriptions. (Don’t tell accounting!) I ran real campaigns, served real clients, and made enough expensive mistakes to save you from making them yourself.
These lessons aren’t theory; these are seven lessons that actually changed how we work at flyte new media.
Lesson 1: Be Human-First, Not AI-First
I started 2025 wanting to make flyte an “AI-first company.” By mid-year, I realized this was a terrible idea.
Remember when every company rushed to add “.com” to their name, then quietly dropped it after the crash? I expect the same to happen with AI.
Recent research shows most consumers don’t actually want AI in their products. They’re less likely to buy when you promote AI features. Dell recently admitted consumers aren’t interested in AI PCs. What do people crave? Doing business with other humans; a human connection.
When every company has an AI chatbot, what separates you from the competition? Human relationships. Real connections. Those are exponentially harder to replicate than any chatbot.
The shift I made: flyte is now a human-first company where every human feels empowered to use AI to create better outcomes for our clients. The AI serves the humans, not the other way around.
Lesson 2: Perfect Prompts Are a Myth
Every week, some YouTube guru promises the “perfect prompt” that will solve all your problems. ROLE and GOAL, RICCE, CLEAR—about fifty frameworks that all claim to be revolutionary.
Here’s what I learned: they’re mostly just repackaging what already exists.
Yes, frameworks are useful starting points. But what gets you better results is your own trial and error. Perfection is the enemy of getting things done.
Sure, watch those gurus; I still do, but approach them with healthy skepticism. You might already be better at this than they are. What matters most at the end of the day isn’t the framework—it’s the quality of the output.
Lesson 3: Test Your GPTs Across Multiple Platforms
The exact same AI assistant may give you different results on different platforms.
I built a podcast assets producer with the same instructions and knowledge documents, then tested it on ChatGPT, Claude, and Gemini. The outputs varied significantly in quality and style.
Recreating a custom GPT on another platform takes almost no time. When quality matters more than speed, I’ll drop the same prompt into all three platforms, review the results, and sometimes even combine the best parts from each.
Lesson 4: Match Your Image Generator to the Task
I entered 2025 loving Midjourney for its control and sophistication. Then Google launched Nano Banana, and everything changed for product photography.
I found stock photography of a woman holding a generic silver can, uploaded a photo of a Moxie can (it’s a Maine thing), and asked Nana Banana to combine them. Even zoomed in closely, the result was flawless—ready for actual advertising.
The other tools couldn’t match that accuracy for product work. But they excel at other tasks. I use Photoshop’s generative fill constantly for aspect ratio changes. Canva’s AI works well for consistent branding.
The lesson: different AI image generators excel at different jobs. Match the right tool to the specific task instead of forcing one platform to do everything.
Lesson 5: Create Better Content, Not More Content
Years ago, I said: “You can’t beat the internet on volume, but you can beat it on quality and perspective.” In the age of AI, this is truer than ever.
Don’t use AI to generate more content than you currently create. In fact, maybe you should use AI to create less content…just content that’s more valuable to your audience.
Here’s a question I’ll ask AI before jumping into creating a piece of content: “What are the content gaps in the top-ranking articles for this topic?” or “What are popular articles not covering that I should include?” I use it to edit and improve flow.
Think about AI for content as an enhancement tool, not a replacement engine.
Lesson 6: AI Can’t Write Your Final Draft
At flyte, I often say we never send out AI-generated content that hasn’t been touched by a human. That’s not exactly true—there’s commodity content like social posts from your business page the algorithm will show to virtually nobody.
But outside of that? I don’t trust AI to write final version content, even when it’s been trained in my voice, knows our ideal client personas, and has absorbed our brand guidelines.
Case in point: I used AI to draft this blog post based on my original podcast recording. It’s decent, but I’ve been editing every paragraph as I go through it…and adding paragraphs like this one.
AI doesn’t have the real-world experience we do. It hasn’t run actual client campaigns. It lacks our perspective.
Even if you think AI writes better than you do, here’s the problem: it writes similarly for so many brands. A recent New York Times article explained why AI sounds the way it does: it follows best practices like the Oxford comma, the rule of threes, and adds em dashes everywhere. (I just removed one from this paragraph!)
AI is great for creating frameworks, identifying content gaps, and editing your work. But if you’re publishing AI content as-is, how is that higher quality than a Google AI Overview? And if it’s not higher quality, why would Google rank it? (It wouldn’t.)
Use AI to enhance your writing, not replace it.
Lesson 7: Use AI in Your Personal Life
Don’t limit AI to work. You’ll get stuck in a rut and stop exploring new possibilities.
I recently inherited woodworking tools with no manuals. I opened ChatGPT, took photos, and asked what each tool was and how to use it. Problem solved.
I found an awesome 3D cutting board design on Pinterest shot at a dramatic angle. I dropped the image into Nano Banana and asked for a top-down view, then used Photoshop to isolate the repeating pattern.
I also use AI for gardening—identifying plants, distinguishing weeds from keepers, checking plant health.
Finding personal use cases sparks imagination for professional applications.
Three Things Worth Remembering
$20 per month is nothing. I recently taught a $500 all-day AI workshop. Half the attendees refused to pay $20 monthly for premium access. They spent $500 for one day with me but won’t spend $20 monthly for an AI assistant that works 24/7. The math doesn’t add up.
I don’t trust “10x productivity” claims. Either these people were doing nothing before AI, or they’re producing the AI slop ruining everyone’s newsfeed.
It’s getting scary hard to differentiate real from AI. I follow two subreddits dedicated to “Is this AI?” When I see what people question, I realize I wouldn’t have suspected them either. That’s concerning for trust.
What This Means for You
The businesses that win won’t be “AI-first.” They’ll be human-first companies that strategically use AI to deliver better outcomes. They’ll focus on quality over quantity. They’ll maintain their unique perspective instead of letting AI flatten their voice into generic sameness.
Stop chasing perfect prompts, test tools properly, and remember that AI is an enhancement tool—not a replacement for human expertise and relationships.
That’s what I learned in 2025. That’s what I’m bringing into 2026.
What about you? Let me know in the comments.
Because the best marketers aren’t just keeping up with change. They’re driving it.
Transcript from Rich Brooks’ Episode
Lesson #1: Don’t try and be an AI-first company. I started 2025 with this goal in mind for flyte new media, wanting to be an AI-first company. But over the course of the year, I realized that this was really a terrible idea for those of us old enough to remember the .com boom and bust.
There was a rush for companies to add .com to their name because they wanted to seem forward thinking and futuristic. Then as the internet imploded, at least for a while, a lot of those companies – the ones that didn’t go out of business – changed their name dropping the.com.
On top of that, there was some research back a year or so ago that showed that most consumers don’t want AI in their products and are actually less likely to buy products that promote the fact they come with AI.
And then I was reading just recently that Dell admitted that most consumers are not interested in AI PCs either. People want to do business with people. So you want to put your people front and center.
My current objective is to make sure that flyte is a human-first company, where every human feels empowered to use AI to create better outcomes for our clients.
Now, I’m sure you’ve seen some news stories about people with AI girlfriends or boyfriends, and people treating their chatbots as BFFs, but to me that speaks to a desire for human connection, maybe a connection those people can’t seem to make in real life. People may want to self-serve, and there are times and places for an AI chat bot or interface, but what’s going to separate you from your competition when every single company has an AI chatbot?
Human beings and human relationships are a lot harder to replicate than an AI chatbot. So really, think about that mind shift. Stop trying to be an AI-first company and instead be the human-first company that your customers really want.
Lesson 2: Perfect prompts don’t exist. Now, I follow a lot of creators on YouTube and other channels, and very often they’ll come out with some video or post about creating the perfect prompt.
I’ll be honest, it always gets me super excited. I’m like, yes, that must be the perfect prompt that will get me the perfect output. And of course, this framework of role and goal, or “ricci”, R-I-C-C-E or “clear”, C-L-E-A-R is going to solve all of my problems. But the truth is that these frameworks are often just regurgitations and renaming of things that have existed before, and chances are you’ve already tried them just under a different name.
And yes, it’s great to start with a framework for a prompt to make sure that you get everything in that prompt that should go in so the AI can create a better outcome. But it’s your own trial and error and your own process of iteration that’s going to get you better outcomes. And remember, that perfection is the enemy of getting things done.
So yeah, check out the gurus and AI experts on YouTube or TikTok or whatever your favorite platforms are. But watch those videos and read those blog posts with a healthy dose of skepticism, as you may be better at some of this than they are. You’ll find that the gurus you can trust over time by testing their frameworks and suggestions and seeing if they work for you and your business. Ultimately, it’s about the quality of the output in your opinion or your client’s opinion that matters most.
Lesson 3: Run the same GPTs on multiple platforms. 2025 was a big year for me in creating custom GPTs or AI Assistants. And I’ve discovered that it’s very easy to create a GPT in one platform, like creating a custom GPT in ChatGPT, and then recreate it in another platform like Claude Projects or Gemini Gems. Very often the exact same AI assistant will give you different results depending on the platform it’s run on.
So once I create it, it’s very easy. I mean, an AI Assistant, or a custom GPT as it’s often called, is really just a set of instructions and some knowledge docs. And so to recreate that in multiple platforms takes almost no time at all. So if you are able to create GPTs on different platforms, and often you need to have a paid version to do this, so you might need to be running multiple paid versions of your favorite LLMs. If so, I strongly recommend that you do go ahead and create them on multiple platforms. When you have the bandwidth, which isn’t always, run those processes on multiple platforms and determine which one’s better.
I do this all the time. I have a 207 script writer and a podcast assets producer, and I have them on all three major platforms. And very often I just run head-to-head competitions. I put the same prompts in, I read through them, and I determine which one is best. Very often I’ll take some of one and then add in some of the points from another one as well. And it does seem to change over time, or maybe it’s the topic that I get different results. So I’m not as much about speed as I am about the quality output.
So I recommend if you do have the bandwidth, this is something you want to try out. Now, I don’t always have the bandwidth to run this in multiple platforms and compare and contrast each one, but if you do have that time, it’s a good investment to again, get that best outcome.
Lesson 4: Different AI image generation tools for different jobs. Along those same lines, different image generation tools like Midjourney or Nano Banana or Dall-E, those are best used for very specific tasks, although I entered the year loving Midjourney for its control and sophistication and in my opinion, better output than Dall-E, which is part of Open AI, or ChatGPT when Nano Banana was introduced, which is part of Gemini. I found that its ability to stay true to the original images made it perfect for product shots.
So for example, I found a piece of stock photography that had a woman on a beach holding a silver can that could be replaced with anything, any brand. I uploaded a can of Moxie, which is a surprisingly popular drink here in Maine, don’t ask me why. And I asked Nano Banana to combine the two. Even when I zoomed in really closely on this image, Nano Banana created a perfect version, they could have been used in Moxie advertising if they so wished. The other tools just aren’t as good at that, but they might be better for other types of work.
So again, I know this is additional time and perhaps additional money as well, but different AI image generation tools are going to be best used for different tasks. And that also includes things like Photoshop, which I use their image generation or the generative fill in Photoshop all the time, especially if I have a square image and I need it to be 16×9 aspect ratio for presentation. I use it all the time. And although I don’t use Canva’s AI as much, I know that Canva’s AI has a role as well, especially when you want consistent branding in your AI generated imagery.
Lesson 5: Don’t try and create more content, try and create better content. Years ago I was on stage and I said, “You can’t beat the internet on volume, but you can beat it on quality and perspective”, and that’s even more true in the age of AI.
I had said this because when I was first starting out, it was very easy to rank well for almost any content you created. But over time, big media publishers started creating content at such volume that it was impossible to keep up with them. They had entire teams of people creating content every single day. You just couldn’t beat them on volume, at least for most businesses, but you could beat them by having a unique perspective or unique experience or having higher quality content than what those content mills were churning out.
And this is even more true in the age of AI. With content creation and AI slop, you shouldn’t be using AI to generate more content than you currently are. That is not a good outcome. In fact, maybe you should be using AI to create less content but make it better and make it more valuable to your audience. And I do this through asking AI questions like, “Take a look at the top-ranking articles for this topic and what are some of the content gaps? What are the popular articles not writing about that I should include?” Things like that. Or asking you to edit to improve the flow of content. Those are the ways that you should be using AI, in my opinion.
Lesson 6: AI is not ready to create a final draft on its own. I often say at flyte that we never send out AI-generated content that hasn’t been touched by a human. Which is not exactly true. The truth is, there is some what I call “commodity content”, such as social media posts, on our business pages to promote a recent blog post or podcast episode. I know that the algorithms are just going to show this post to virtually nobody, so I really don’t care about it. And I’m just going through the motions because it’s a box I have to check.
But outside of those posts, I do believe that there is a value hierarchy of content and AI. Even AI that’s been trained in my voice and knows flyte’s ICPs (ideal client personas) and have absorbed all of our brand guidelines. I just don’t trust it to write a final version content. It doesn’t have the real-world experiences that we do. It doesn’t have the experience of running client campaigns the way that we do. And it doesn’t have the perspective that we do.
Even if you think that AI writes better than you do, the problem is it writes similarly for so many people and so many brands. I won’t say for everybody, but there was this great article in The New York Times earlier this month, maybe last month, that explored why AI sounds the way it does. And part of what the article is all about is it sounds the way it does because it follows best practices, many of which I use, like the Oxford comma or the rule of threes.
It also explains why it loves the M-dash, because it was trained on a lot of content that uses M-dashes, and M-dashes are pretty awesome. They’re just not used by the general population as frequently as AI uses it.
I think AI, when it comes to content creation, is great for creating frameworks. I think it’s great for identifying content gaps, like I was talking about before, and identifying things that you may have missed. I think it’s great for editing your content when you’re done. But if you’re just asking AI to create content on your behalf and publishing it as is, how is that any higher quality than the AI overview you’d find Google search offers? And if that’s the case, why would Google ever rank your AI generated article at all when it could do that and does it better than you can do it yourself?
Use AI to enhance your writing, not to replace it. And along those lines, you can’t trust AI. We’ve all heard funny stories about how AI recommended gluing cheese to your pizza so it doesn’t slide off, or the number of rocks you should eat each day to stay healthy. But it’s less funny when AI makes a mistake on the content that you are creating. This is why it’s important for you to have your own expertise, even if you’re leveraging AI to help you create content.
How many times have you asked AI for some content creation or the answer to a question, and it gives you an answer you know is wrong. So what do you do? You call it out and you say, “Hey, that’s not exactly right.” And it says, “Oh, you’re completely right. That’s on me”. For me, that happens way too often, and I only catch it because I’ve got enough internal expertise, enough real-world experience that I know what it put out isn’t correct. Depending on how critical the content is for your brand and your credibility, you should definitely consider double checking anything you’re sure of and even running it through either a new chat window or a different AI tool and ask it, “Is this accurate? That’s often a great way of uncovering some hallucinations that the AI may have generated on its own.
And last but not least, Lesson 7: Use AI in your personal life. Don’t limit yourself to using AI only for work, because you’ll just find yourself kind of getting into a rut and not exploring using it in new ways. So instead, explore it in your personal life as it’s going to open up new ideas and new approaches that you may not have considered, that then you can use in your business or professional life.
As an example, recently I inherited a bunch of woodworking tools, not all of which I knew what they were or what they did. Many of them were already out of the box, maybe were stuffed into a box with a bunch of other tools, and didn’t come with manuals. So what did I do? I opened up ChatGPT. I took a picture of the tool in question and asked ChatGPT what this was and how I might use it. And ChatGPT was great in explaining, “Oh, this tool is for sharpening”, or “This tool is for sanding”, or “That tool is for putting up drywall”. Which meant that it went into the Goodwill box, because I’m not going to be putting up any drywall. So that was really helpful. And this is just another way that I start to use AI and see what it’s capable of so that maybe I’ll find a business use opportunity as well.
Another use case that I just have, and I know that these are all woodworking, but hey, I’m back into the woodworking thing as well. So another woodworking example is that I decided that for my next project I wanted to do kind of a complex ingrain cutting board. And I went on to Pinterest looking for inspiration, and I found this really awesome looking 3D style cutting board. I could tell it was ingrain, and I knew from having built other style cutting boards that it was just going to be a repeating pattern. And all I needed to do to be able to build this was basically uncover what that repeating pattern was.
Unfortunately, the photo of this image was taken at this very dramatic angle, which meant uncovering that repeating pattern was nearly impossible. So I took the image and I dropped it into Nano Banana, and I asked it to give me a photo version of this from the top down, which it was able to do. And then I was able to take that image, drop it into Photoshop and isolate the repeating pattern so I could create a template to reverse engineer and make that cutting board myself.
Now honestly, after doing all that work and doing the reverse engineering, I realized that this may be a little bit outside of my skillset today, but half the joy for me is figuring out what needs to be done. So again, just a great use case and it gets me thinking about other ways I can be using AI in my business.
Also, when the weather’s nicer and I’m gardening, I often use AI to identify plants telling me if it’s a plant or a weed, or if this plant is healthy. So again, I’m using AI to help my garden grow. Find some ways in your personal life that you can use AI, and that may spark the imagination within you to start using it more creatively in your professional life.
All right, let’s move on to “quick hits”, just a few quick ideas that I had that really weren’t exactly lessons. But the first one is $20 a month for an AI tool is nothing, and you’re a fool if you’re trying to save money by using the free version of ChatGPT or Gemini or what have you. I taught a $500 all-day workshop recently about AI, and half the people there didn’t want to spend $20 a month. I’m like, you just spent $500 to spend the day with me, and you don’t want to spend $20 a month for an AI assistant that’s going to do all this work for you and help you grow professionally? The math just doesn’t add up for me.
Now, the flip side of that is that seemingly every AI tool is $20 a month, and they add up fast. I, myself, have about $150 to $200 a month habit. I’m kind of afraid to look, but again, I’m an extreme case, and part of my goal this year is to really position myself as more of an AI authority. So this is an investment in myself.
Next quick hit, AI optimization and SEO are almost the same thing in my book. If you’re doing one, you’re likely already doing the other. There are definitely going to be nuances, and certainly this may change as the technologies advance. But at this point, I’m not sure that AI optimization for your website should be keeping you up at night.
Third quick hit, I don’t trust people who say they’re 10xing their productivity or output by using AI. Either they were doing nothing before AI, or they’re the ones producing all that AI slop that’s ruining my newsfeed.
And the fourth and last quick hit, it’s getting harder and harder to differentiate real from AI. In fact, I follow two separate subreddits in Reddit dedicated to answering the question, “Is this AI?” And when I see the photos or videos that people are asking about, I realize that if I had seen that video in the wild, I wouldn’t have even considered that it could be fake. That makes me worried about disinformation, and it also makes me worried that it’ll be hard to get people to trust us as content producers. And by ‘us’ I mean brands and marketers. So just something to keep in mind. And again, kind of loops back around to building a human-first company rather than an AI-first company.
And lastly, three AI resolutions that I’m going to be working on this coming year. First. I’m going to adhere to my blocks of AI research time. I put blocks in my calendar about five hours a week for working on AI. And very often I blow them off because I’ve got paying work or I’ve got something else going on. But I’m going to adhere to those blocks of AI research time that I put in there. And when I can’t because I need to do a kickoff meeting or a coworker needs to sit with me, I’m going to reschedule the block of time for later in the day or later in the week.
Second resolution is I’m going to talk to my AI more often and I’m going to use Whispr, W-H-I-S-P-R, often when I do. I actually used it to create this podcast script. I installed Whispr on my computer, and then I used it to transcribe my thoughts into text that then I could manipulate and clean up for today’s episode.
And lastly, I’m going to roll out more AI deliverables that make sense for our clients in 2026. I actually have started a beta program right now that we’re calling BrandGPT, and these are custom GPTs that understand our customer’s brand voice and understand the buyer personas. I’ve already built out a few during this beta phase that we’re doing right now at a reduced cost for some of our clients. Our beta testers can use this brand GPT to develop blog posts, write emails. E-commerce descriptions, and so on. Always using the dual lenses of brand voice and buyer personas.
Anyway, that’s what I learned in 2025 that I’m going to be bringing forward to help me through 2026 when it comes to AI and how we’re going to use it as a company and how I’m going to use it both personally and professionally. I hope you got some good ideas out of this, some things that you can put to work. If there’s something that you learned or something that you are planning on doing in 2026 that I didn’t mention, let me know. I’m easy to get in touch with.
You can use the contact form at theagentsofchange.com. You can find me on LinkedIn. I am @TheRichBrooks there and everywhere else on the internet, and I look forward to hearing what you are doing with AI this coming year.
Show Notes:
Rich Brooks is the President of flyte new media, a web design & digital marketing agency in Portland, Maine, and founder of the Agents of Change. He’s passionate about helping small businesses grow online and has put his 25+ years of experience into the book, The Lead Machine: The Small Business Guide to Digital Marketing.