STOP Letting AI Think for You (Seriously, Don’t Do This)

Watch the Episode on YouTube

AI isn’t coming for your solopreneur job, but it will happily take your tasks. In this episode, Carly and Joe break down how to turn AI into the world’s most over-eager intern: fast, enthusiastic, occasionally clueless, and totally coachable. From using AI without losing your voice…to avoiding “hallucination traps”… to building a GPT that actually remembers what you told it, this episode shows solopreneurs exactly how to manage AI, not get managed by it.

FAQs From The Episode

How do I actually ‘train’ AI for my business?”

Treat it like onboarding a new team member. Upload or paste your brand guidelines, audience description, business plan, examples of your writing, and list of do’s/don’ts into a custom GPT. Then, add instructions that reinforce your tone, values, target audience, and preferred outputs. Revisit the training regularly, because AI can “forget” if everything stays in one long chat thread.

What tasks should I delegate to AI, and which ones should stay 100% mine?

Delegate:

• Drafting (blogs, emails, social posts)

• Research summaries

• Brainstorming ideas

• Rewriting in different tones

• Organizing messy text

Never delegate:

• Strategic thinking

• Topic selection

• Opinions or point of view

• Final edits

• Any part of your brand voice that requires nuance

AI can assemble the pieces, but you must decide what the puzzle looks like.

How do I keep AI from drifting, forgetting instructions, or making weird mistakes?

Use a custom GPT (not just one long chat). Put your recurring instructions directly into the system prompt so they don’t fall out of the context window. Reinforce boundaries often (“never use profanity,” “don’t use dashes,” etc.), provide examples, and correct mistakes as they happen. Think of it as ongoing performance management. The more feedback and clarity you give, the better it performs.

So, if you are lacking direction, having a hard time generating leads, or are having trouble keeping up with everything you have to do, or even just lonely running a company of one, click here to check out LifeStarr Intro!  

Episode Transcript

Carly Ries: AI isn't here to take your job. It's here to take your tasks, but only if you know how to use it right. In this episode of The Aspiring Solopreneur, Joe and I break down how solopreneurs can treat AI like an overeager intern, fast, enthusiastic, and a little clueless, and turn it into a true teammate. We cover how to onboard your AI so it understands your business and brand, what not to delegate, hint your thinking, why Neil Patel says human ideas still outperform AI content, and the danger of letting your intern bot hallucinate on your behalf. If you've ever wondered how to use AI without losing your personal touch, this one's for you.

So tune in and learn how to manage AI, not be managed by it. You're listening to the Aspiring Solopreneur, the podcast for anyone on the solo business journey, whether you're just toying with the idea, taking your first bold step, or have been running your own show for years and want to keep growing, refining, and thriving. I'm Carly Ries, and along with my cohost, Joe Rando, we're your guides through the crazy but awesome world of being a company of one. As part of LifeStarr, a digital hub dedicated to all things solopreneurship, we help people design businesses that align with their life's ambitions so they can work to live, not live to work. If you're looking for a get rich quick scheme, this is not the place for you.

But if you want real world insights from industry experts, lessons from the successes and stumbles of fellow solopreneurs, and practical strategies for building and sustaining a business you love, you're in the right spot. Because flying solo in business doesn't mean you're alone. No matter where you are in your journey, we've got your back. Joe, in our last episode we talked about our experience at Inbound, and obviously a big topic for this year was AI, as is the topic for most things this year, not just at that conference. But I think the biggest fear that people have is that AI is here to take their jobs.

And I would say, it's not here to take your jobs, it's here to take your tasks. And the only reason you would lose your job, I think, is if you would lose it out to somebody that is using AI for their tasks that has the same position as you, and you are not. Would you agree with that?

Joe Rando: Well, I think I mean, if we're talking about solopreneurs, yes. If we're talking about in general, I mean, the news out today, we're on the 10/29/2025, and major companies are laying off tens of thousands of white collar employees because they're basically being pressured by investors, and the c suite's just saying, we're gonna replace these people with AI. I don't predict that's gonna go well. I suspect that they're gonna be a little surprised as to how that doesn't live up to their expectations. Eventually, it will because AI is getting better all the time.

But yes, I think as a solopreneur where you're in charge of your own destiny, you don't have to lose your job because AI starts doing something that you do. So I'll go there.

Carly Ries: And for solopreneur's sake, we think you should be looking at AI as your teammate. I think of AI as kind of like an over enthusiastic intern. Like they're fast, they're eager, but they are wrong, sometimes clueless. AI is not perfect. And so I see it as a, like how good I can do, look how good I can do, but that training still needs to be there.

So I wanna talk about AI as your teammate, especially as a solopreneur, and how you can kind of coach that teammate today, and to help you understand how to manage AI, not be managed by AI. Does that sound like a good direction?

Joe Rando: It does. I wanna start with just kind of some framing, which is, and I'm guilty of this as much as anybody, is that we like to think that AI thinks, but it's not thinking. It really isn't. It just feels that way. It predicts. It's a somebody described it as a plausibility engine, which I thought was a brilliant way to describe it. It's predicting what the next word should be based on the inputs, and, it's surprisingly good given that that's what it's doing. You kind of think about it.

It's doing that, and you go, wow. It's pretty darn good. But it is by no means, as you said, perfect or even close to perfect and can really go off the rails. And if you wanna know, just how weird it can get, I have a GPT I created called You're Wrong. And You're Wrong basically has a bunch of prompts to basically think more, critically and to be less nice about critiquing.

And so I'll put something in it, and it'll kind of rip me apart. You know? It really roasts me. And sometimes I'll take chat GPT generated material and put it into your wrong, which is also chat GPT, and it will roast it. And so you just kinda look and go, okay.

And the thing with you're wrong is it's actually better than chat GPT at coming up with good ideas. It's much more critical of its own thinking. But, anyway, the point being, you do not want to treat this as anything other than a highly eager, coachable, but pretty pretty naive intern.

Carly Ries: I'm so glad you said coachable because I think a common misconception that people have is that the AI knows their business. Like if you typed your website in, that it would automatically know the ins and outs and it just doesn't. you really have to take time to train AI so that it works the best for you. So you can't just say, let's say you're writing a social media post for your company, and it's like write a social post for LifeStarr. But it wouldn't really know what to do.

So you have to say, write a social post for LifeStarr that targets solopreneurs and helps them with their productivity. you need to train it. And granted over time it'll start picking up on it, but if you wanna start getting it right immediately, you need to put in that effort just like you were training another employee, another contractor, whatever. It needs that training. Even that onboarding.

We talk about onboarding a lot.

Joe Rando: Yeah. There's a lot to it. But one of the problems that I've seen, I don't know if anybody else has experienced this, is that I will train up a GPT, not necessarily, doing anything, you know, under the hood, so to speak, but just kinda go in and give a bunch of instructions. And over time, if I'm using this thing for a lot of work and using the same feed, it'll start to forget. And it's got to do with something called the context window that can only hold so much information.

And so you start off with some initial instructions. You get deep down into it, it starts to forget the initial instructions. So you need to if you really wanna use it and train it up, you've gotta go in and create a GPT and put in a bunch of training materials so that it doesn't ever lose that. Does that make sense?

Carly Ries: it does. And I mean, even if you have a business plan, if you have like a brand guidelines document, those are the kinds of things you should be inputting and retraining. And don't rely on it like you were saying, to always remember. Like, maybe have a Monday morning meeting with ChatGPT and be like, hey, just a reminder, this is my business plan. Let's focus on this for the rest of the week as we're putting together the content and all of that.

Joe Rando: Well, like I said, you should create a GPT for that. So the business plan is in there in the core training as opposed to just trying to make it remember in the base chat GPT.

You don't want to do that because it won't.

Carly Ries: Yeah. But I kinda look at AI as like an executor, a brainstormer, and a researcher, whereas you are the visionary, the decision maker, the editor, editor, editor. And that's how I define the two different roles. Again, kinda with that intern mindset and framework.

But Joe, if people are treating it like an employee, because, I mean, AI really is a game changer for solopreneurs. It brings up so much time. What would you say you should delegate, and what should you not delegate?

Joe Rando: Well, at the core of it, do not delegate your thinking. just as an example, write a blog post about the three biggest mistakes that solopreneurs make without telling it what those three mistakes are. Don't let AI think through. I mean, you wanna go ahead and brainstorm and say, throw out, the biggest mistakes that you think solopreneurs make based on a search of the web, and then go through those and go, oh, wow.

Yeah. I didn't think of that one, and there's this one. And then take that, but you need to feed it your thinking. Right? And you it's nice to feed it your thinking, your writing style, and a lot of context so that you're getting something that isn't just AI dribble. And nobody wants to read AI drivel. Neil Patel said that non AI generated content generates 4.7 times more views than AI generated content. And that should tell you that, and I'm not saying not to use AI, but if you can use AI in a way that it's your content, your ideas, your thinking, people are gonna respond to that better than this kind of, kind of generic AI thinking average of the entire inter interwebs, stuff that it's gonna generate on its own.

Carly Ries: Because it does hallucinate things.

Joe Rando: Oh, that too. But even when it's not hallucinating, on its own, it's pretty boring.

Carly Ries: Yeah. Absolutely. But with the hallucination thing, I think it's also really important to identify when AI is leading you astray. When you're working with it as, I'm assuming you keep saying intern for consistency purposes.

Call it whatever you want. But if you can identify the mistakes it makes time and time again and train those mistakes out of them like you would an intern, that'll also help with your strategy, your content, everything that you're using AI for. Joe, didn't you just do one that replaced Emdashes?

Joe Rando: It didn't work. I created a GPT, and the entire prompt was never use Emdashes, period. Never use Emdash. I'd put it in about it must have been 50 or 60 times. And then I asked to write an Emily Dickinson poem, a poem in the style of Emily Dickinson, and it used a bunch of em dashes.

It's nobody other than, your English professor uses em dashes. It takes, keystrokes that nobody does. How did it get trained up on em dashes?

Carly Ries: I know. it was almost like bring back the em dash.

Joe Rando: But nobody wants to use them now.

the woman that was editing Solopreneur Business for Dummies, she was an English major and very, literary. And she sent me some content. I said, oh, we can't use the em dashes. It looks like AI. She goes, what? I always use em dashes.

I heard semicolons are becoming the same way, and I love semicolons, been using them for decades, and I'm gonna have to give up semicolons.

Carly Ries: so when I have used em dashes in my writing, I've always had to copy and paste it from other copy or other content because I couldn't figure out how to do the em dash myself. Which I suppose that's a quick AI search, but I just always put them to different other content. Anyway, Joe, any parting words of wisdom for having AI as your employ and training it and onboarding it that you think solopreneurs should know about?

Joe Rando: Well, I'm gonna reiterate this idea of, there's a difference between creating a GPT and putting training in the GPT and just kinda going into ChatGPT and saying, do this and do this and don't do this because that context window and I'm no expert. So you know? But basically, what it seems to me is that context window that contains kinda what it's using as a reference point, the old stuff goes off eventually, and it starts to drift. So, if you do this, you're gonna wanna build out a GPT trained to do certain things and put certain documents in there and certain instructions. At least that's ChatGPT.

That's my go to. Not so sure on, things like perplexity and anthropic. But, the main thing with all of these is just check the work. Make sure it's not crazy. You know, it's just so easy for these things to go off into delusion.

And if you let that go out, it's not gonna reflect on ChatGPT or some other, large language model. It's gonna reflect on you.

Carly Ries: And you need to train it, correct it. You know what I've started doing, Joe? I've started and I don't know if this does anything. I think it does. I started celebrating it.

So when ChatGPT does something really well, I say, that's awesome. That's exactly what I was looking for. And then I'll iterate and be like, can you also do something like this? And it'll say, thank you so much. And then I find that everything from that point on, the results are better because I gave it the positive reinforcement.

I don't know. This might be in my mind. Before an AI takes on minds of their own, I want them to know that I was grateful.

Joe Rando: Yeah. And it's true. And, yeah, it's good to give it you know, there's a little thumbs up. You can click on chat GPT as well to let them know it did a good job or a thumbs down.

But, but, yeah, and the other thing I guess the one last thing I wanna mention is that because they're updating the model all the time, you know, it was four, then was four o. Now it's five, and they got all the different five versions and, deep research depending on which version of ChatGPT you're buying if you're buying it. But it's weird, the behavior changes, right, when they make these changes to the model. Like, I've got this you're wrong one that I talked about that I use all the time, and it's great. But all of a sudden, it started using profanity.

It never did before. And now all of a sudden, it's using and I remember they made some change just now to loosen the thing up a little bit. And, it's not my style. I don't use profanity, so I'm not really inclined to want it to do that. So now I've got to go in and change the training, say don't use profanity. but just there's a lot of weirdness. You always you build these things, then you have to keep an eye on them. It's kind of like an intern that, develops a drinking problem or something.

Carly Ries: Or something. Well, listeners, hopefully you found this episode useful. I know there's a lot of AI talk out there, but not always as it relates to solopreneurs specifically. So hopefully you learned something new. And as always, leave us that five star review.

We would so appreciate it. Share this episode with a friend. Subscribe to our show on your favorite podcast platform, including YouTube, and we'll see you next time on the Aspiring a Solopreneur. You may be going solo in business, but that doesn't mean you're alone. In fact, millions of people are in your shoes, running a one person business and figuring it out as they go.

So why not connect with them and learn from each other's successes and failures? At LifeStarr, we're creating a one person business community where you can go to meet and get advice from other solopreneurs. Be sure to join in on the conversations at community.lifestarr.com.