How do I decide which business idea to pursue?
Short answer
I run every idea through three questions: can this 10x, does the math work, and can AI handle 80% of it? Then I compare time burden, custom-work burden and whether it concentrates or fragments momentum. I favor fewer, larger, recurring customers. Ideas that don't win get parked, not deleted.
Three questions, every time:
"Can this 10x? Does the math work? Can AI handle 80 percent?"
What if two ideas both pass?
Only one gets the next operating period. I score them on:
- Upside
- How fast I'll get evidence
- How much output it gets per hour of my time
- Fit
- Cost of proof
- Downside
- My own capacity
Each idea ends in one of four calls: commit, test, park or stop, with a metric and a stop condition. And once I've chosen a vertical, I don't reopen it.
What makes an opportunity worse than it looks?
Time and custom work. I compare revenue potential, time burden, custom-work burden, how much can be automated and productized, whether it could support dedicated management when it's bigger, and whether it concentrates or fragments momentum.
A business that makes money but keeps me stuck in support or bug fixing can be worse than a more concentrated one.
Many small customers or a few big ones?
A few bigger ones, usually. My priors lean toward productized, reusable systems, automation, small human teams, larger-dollar customers and recurring revenue. I want to avoid fragmented revenue that leaves me managing lots of tiny high-touch businesses. Those are priors, not laws.
What do I do with ideas I'm not pursuing?
Park them where I can find them. I have so many good ideas that it scares me I'll lose them. I keep a Notion gallery of business ideas. Each card has the one-sentence idea and a one-sentence summary, and opens to the potential, the strategies, how AI would do it, cost per order and a small test checklist, with tags for cross-referencing.
An AI agent learned the template, so I can say "save this as a business idea" anywhere and it makes the card. Saving starts no work and no spend. Parked is the default.
How often do bets on people pan out?
Rarely, and that's fine. From working with a lot of would-be local media operators, I'd say it's not even 10% who take action and really go after it. More like 2%. You're going to be wrong most of the time. Keep giving people a shot anyway, because the one who executes can be a big win.
For what happens after you pick, see should I launch before I have a full plan.
Answered by TJ Larkin. From TJ's written operating principles, a July 2026 prioritization meeting, and team and working sessions in October 2026. Last updated .