AI already writes texts, codes, draws, edits video, and builds entire apps — I wrote about this in more detail. And it does all of it better and cheaper (than me).
So the question comes up. What should you even do when most familiar skills stop being a competitive advantage?
I put together ten directions that seem to have decent prospects.
Ten directions
1. Physical business. Coffee shops, auto repair, home repair, cleaning, small-scale manufacturing. AI will help with ads, orders, and bookkeeping, but you still can't fix a clogged toilet with a prompt.
2. Licenses and partnerships. For example, launching a service together with a licensed law firm, or becoming an approved contractor for a major bank. Permits and contracts don't come bundled with a Claude subscription.
3. Niche media. A channel about used cars, a newsletter for dentists, a trade publication for designers. Content production is getting cheaper, but you still have to earn the audience and the trust.
4. Early entry into new niches. Spot a trend before everyone else. For example, be the first to launch a specialized AI service for architects. The product itself can get copied, but the accumulated data, client base, and reputation can't be copied nearly as fast.
5. Paid communities. Hampton, for instance, gathers entrepreneurs into closed groups with in-person meetups. You can copy the website and the chat, but not the relationships between the members.
6. Offline events. Niche conferences, club trips, meetups for specialists, professional retreats. Nomads, for example, regularly runs digital-nomad meetups in different cities.
7. Marketplaces. Cars & Bids runs car auctions for enthusiasts. Writing a similar platform is now fast. Getting enough buyers and sellers onto it is a lot harder.
8. Unique data. Nomads sells companies access to a database of cities rated by real travelers — starting at $2,490 a year. You can't just ask a neural net for data like that. It takes years to collect.
9. Curation and selection. For example, Hotelist helps you search for hotels by very specific criteria — down to whether there's a full-on weight room and blackout curtains. You could build catalogs of vetted specialists, restaurants, services, whatever. In a world of infinite content, good filtering can be a business in itself.
10. Selling the finished output. For example, an automated agency where AI agents run the actual business and produce landing pages, logos, or decks to order. The client pays for the work delivered, not a designer's hours. Though you could also spin up a thousand competing agencies in an evening.
Nothing is guaranteed
None of these businesses is bulletproof, of course. A coffee shop can go under, a community can fail to form, a marketplace can end up with no buyers.
But all of these directions require something beyond generating code and content: permits, proprietary data, trust, connections, or actual work with real people — which echoes the price-of-error rule I wrote about for tasks you're better off not handing to AI.
Shipping an app in an evening is pretty easy now. Getting people to actually use it and pay for it is still its own profession.
Have you already figured out what you'll do once AI takes over your job?
Based on this breakdown of business models and a post about the moats that are left.