Etzioni on AI: Ten Commandments for AI Startups

For my tenth column in the “Etzioni on AI” series, I want to share ten commandments for AI startups, preceded by timeless classics that still apply. They draw on my work with founders at the AI2 Incubator, Madrona, and my own experience as an AI founder from Netbot (1996) to Vercept (2024).

To see what’s different for AI startups, let’s warm up with ten commandments for startups in general, which fold in the wisdom of startup stalwarts such as Vinod Khosla, Reid Hoffman, and Eric Ries. If you already know the classics, jump straight to the AI commandments.

1. The real risk is regret. It’s sitting in the stands at 60 and realizing you never stepped up to the plate.

2. Choose your co-founder as carefully as a spouse. You may end up spending more time with your co-founder in the early years. And founder breakups are also very painful.

3. Maximize your odds of success, not your ownership stake. Don’t end up owning 99% of nothing. The right investor, incubator, or hire will add far more value than cost.

4. Raise on the story. Investors are buying the company you could become. As Khosla puts it: “Don’t subvert your story in service of logical order.”

5. Solve a real problem. Is it a painkiller or a vitamin? Startups sell painkillers.

6. Talk to your customers early and often. Your assumptions are wrong until a customer proves otherwise; an ounce of data is worth a pound of intuition. A scrappy experiment engaging users settles arguments that could otherwise take months to resolve.

7. Focus, focus, focus. The hardest word for some founders is “no.” Say it to good ideas so you can execute the one great one. If your product is both a dessert topping and a floor wax, it’s neither.

8. Build the rocket while you’re flying it. You launch without every answer in place and iterate quickly to figure the rest out.

9. Be ready for a rough ride. A startup can seem like a series of near-death experiences. The winners are the ones who refused to quit.

10. Beware of consultants. Remember the ham-and-eggs adage: the chicken is involved, but the pig is committed. You want a committed team. Anyone running a meter has incentives misaligned with yours.

Here are a few commandments that didn’t make the cut: hire carpenters, not architects; hire slowly, fire fast; know your numbers cold, especially your burn rate.

Other ten-commandment lists worth perusing include: Reid Hoffman’s, Howard Tullman’s, and Shlomo Kalish’s, as updated by Glilot Capital.

These classics still rule, but it’s time to add ten AI commandments.

1. “We’re an AI company” is no longer a differentiator. It’s table stakes. Tell the story: what’s the pain point? Who’s the customer? How do you make money? Why now?

2. AI technology is not enough. As Madrona’s Matt McIlwain puts it, “the most important AI model is the business model.” And you have to deliver against that model — vision without execution is hallucination.

3. Don’t put lipstick on a model. If your company is a skin-deep gloss over someone else’s API, the frontier labs will eat you alive. They have the model and the distribution. You have neither. Build where they won’t (or can’t) go.

4. Own your data, but don’t mistake it for the moat. Proprietary data helps, but it’s not the promised land. What matters is the flywheel: Waymo learns from every mile its cars drive, and the learning makes the next mile better. Build that loop.

5. Velocity is the new moat. AI has collapsed the cost of being smart. The edge belongs to whoever is smart faster. As a16z’s Bryan Kim writes, “momentum is the moat.”

6. Embed in the workflow. The application your customer opens at 9am and closes at 6pm is the one they can’t switch off. Become that. Or quietly take over the one they already use. AI coding makes it cheaper than ever to recreate the application layer from scratch.

7. Distribution is the scarce resource. Building an AI product has never been cheaper. Getting it in front of buyers has never been harder. The thousand companies competing for your customer won’t lose on features. They’ll lose on reach. Distribution deserves equal billing with product from day one.

8. Don’t marry a model. The frontier model that wins your demo today will be third-best in six months. Build the stack so you can make an easy change if the chemistry fades.

9. Inference is the new COGS. Every query costs real money, and the cost scales with every user. Bessemer puts it bluntly: “If the math doesn’t work at 10 customers, it won’t at 1,000.” Know what your product costs you before you grow.

10. Personal relationships still matter most. AI doesn’t earn trust. It won’t take your call at midnight or defend you to the board, and it won’t be there when you need a bridge round.

Seattle is a phenomenal place to live by these commandments: AI House has opened on the waterfront; new funds have been raised; and hundreds of startups are thriving.

The time for AI startups is now. Go build.