Most founders want to be needed by their business. I've spent nine months trying to make mine need me as little as possible.
Memolio is a personalised illustrated life-story book for grandparents. A customer comes in from an ad or a recommendation, lands on a page that explains the product, maybe tries a sample, and decides to buy. From there, ideally, I don't touch anything.
The order comes in and the book generates on its own. The customer reviews it and fixes any page they don't like, themselves, with up to 50 free regenerations. When they approve, the print partner produces the hardcover, sends the invoice, and ships it to the right address. No step in that chain is waiting on me.
That design is deliberate. The whole point of building this with AI was to remove myself from the day-to-day running of a business, so I can spend my time on the things that actually compound: gathering feedback, improving the product, talking to customers. Not pushing orders through a pipe.
Where I refuse to automate
There's exactly one place I keep myself in the loop, and it matters.
When a book comes out fundamentally wrong, not a single page to regenerate but the whole thing missing the mark, that's a decision a machine shouldn't make. Do I let the customer redo it, or do I offer a refund? That's a judgment about a person's disappointment, and it stays with me.
I learned this the hard way. Early on, a tester's grandmother, who had been a teacher, was drawn as a schoolgirl. The AI took "we met at school" and picked the wrong reading. No amount of better prompting fully fixes that, because the gap isn't technical. It's that a person knows their own story and the model is guessing. That single piece of feedback is why the edit feature exists at all, and why I'll always keep a human at the point where the product meets a real person's memory.
The actual skill
Running a business like this is less like coding and more like being a CEO who happens to have no staff. You're constantly deciding what you can hand off and what you have to keep. Big companies solve this with processes and org charts. I solve it with automations, tracking, and a short list of decisions I never delegate.
The honest version: I haven't fully automated myself out, and I never will. The goal was never zero involvement. It was to make sure that the only time the business needs me is when a human judgment is genuinely required. Everything else should just run.
Memolio is launching now. I'll find out very quickly how much of "it just runs" survives contact with real customers.