Everything I learned running a one-person AI business

When I recorded this in March, it was just me. One person, a day job, and a product with dozens of customers that grew faster than I ever expected.

It was the last video in a long free series, and I used it to tell the truth about my mistakes. Not the highlight reel. The stuff that cost me time, money, and some customers' goodwill.

If you're running a business by yourself, or about to, read this before you learn it the hard way.

Mistake one: I made myself too accessible

When I started, I put my business email out there. I made a Telegram group where customers could chat with me directly. It felt great. Real connection.

Then my day job got a whole lot busier. I couldn't answer every email. I couldn't do the support I'd promised. And instead of the four or five AI employees I run now, I ended up with about twenty, because I needed something covering everything I couldn't get to.

The customers who got used to direct access got angry. "Richard, you said you were accessible. Answer this email right now." That's just not possible when you're one person, especially once you have dozens of customers instead of two.

So here's the fix. You're the CEO. Act like it from day one. Use a no-reply address. Route people to support. If your business is technical, build an AI support specialist to handle the first line. You're only going to get busier.

Mistake two: I got the release timing wrong, both ways

Sometimes I overbuilt. I'd have a version ready to ship and I'd hold it back. "Let me add this. Let me change that." Don't be scared to release.

Other times I shipped something that wasn't ready. One early version was a full program on its own, and people loved it, but it broke. Constantly. Every break became a support ticket, and the tickets buried me.

I showed that version to a gentleman in a meeting, and he told me, "This is good, but I don't think this is your final product." He was right for two reasons. One, I hadn't picked an audience. I was selling it to everybody. You have to niche down to win. Two, it was good, but it wasn't great, because it kept breaking.

What fixed it was a decision. I stopped trying to compete with Claude and started building on top of it. Now when something under the hood needs fixing, that's on Anthropic, not on me.

So find the balance. Don't wait forever, and don't ship something that breaks. Make sure it's simple enough to keep working.

And a warning if you build with code. If you're not a trained software engineer, be careful how much code you put in your product. Claude is good, but it is not a human. You change one line and things can fall apart. If you're in the trades, real estate, or you make websites for folks, you have it easier. What you need is not intricate code.

Mistake three: I chased every idea

I'd see something on X and think, "I've got to build that." Then the next thing. Then the next.

So I did something simple. I turned my X feed over to sports and the news. Got the AI stuff out of there. AI moves too fast to keep up with, and if you try to build everything, you'll be building forever and competing with billion-dollar companies you can't match.

That's another reason I build on Claude. When Anthropic ships something new, I get it. I don't have to chase it. And yes, I keep backup plans in case that ever changes.

Mistake four: I promised forever

Early on I offered a one-time lifetime price that included every future update. I thought it was a differentiator. I will never do it again.

Lifetime customers expect lifetime support from a single payment. And when you build something new and better, they expect that for free too. I got enough angry emails that I gave my lifetime members the new product at no charge, even though it was a totally different product from what they bought. I had to eat that money, because my whole reputation is that I'm honest and I keep my word.

Don't put yourself there. Use a subscription or pay as you go. If you're in the trades, a one-time fee per job is normal. I'm talking about software and services.

How I actually brought in revenue

I kept it simple. YouTube, X, my website, and Google Workspace for email.

YouTube goes real far, especially if you're willing to put your face on camera. A lot of my customers bought because they trusted me, not because of the product. You don't have to show your face, though. In the trades, just post photos and videos of your work.

Then personal outreach. Cold email works if it's personal and not spam. Word of mouth works. Talking to people in your community works. Face to face is still something. AI is great, but it ain't got nothing on a real person.

How to set your prices

Don't guess. Have your AI research your market and what your competitors charge. Build tiers. Think about what your customer is already paying for. If they pay $20 a month for Claude, factor that in.

And here's the rule I'd tape to the wall: raise your prices when every client says yes immediately.

What this means for you

Running a business alone is doable now in a way it never was. But you can't do everything by hand, and you can't figure it all out by trial and error the way I did.

That's what the RMDW Program is for. You get a mentor every week, your AI gets put to work alongside you, the standard builds are included, and it comes with a 12-week guarantee.

Learn from my mistakes. Then go build.