Most AI advice is tactical: this prompt, that tool, this workflow. Tactics decay in months. Strategy is the layer underneath — where to stand, what to bet on, what to ignore. Here are nine principles that have held up so far.
1. Compete on what does not commoditise
Anything a model can do, everyone will soon do for free. So do not compete on production. Compete on taste, distribution, trust, proprietary data, or physical presence. If your advantage is \”I can make things,\” it has already gone.
2. Own the demand, not the supply
Supply of content, code and design is heading toward infinite. Attention is not. The person with ten thousand people who trust them wins over the person with a better tool, every time. Build the list, own the relationship.
3. Specific beats general
General models are excellent generalists, which means the value moved to specificity: your niche, your data, your customer\’s exact workflow. \”AI for X\” where you deeply know X beats \”AI\” every time.
4. Speed of iteration is the new moat
You cannot out-resource a large lab. You can out-iterate almost anyone, because a small operator with good tools can test twenty ideas while a company approves one. Structure your work for iteration count, not perfection.
5. Be early to interfaces, not to hype
Every platform shift creates a window where distribution is cheap: early search, early app stores, early social. New AI-native interfaces open the same windows. Be early on the surface where your customers will arrive, and ignore the token of the week.
6. Barbell your risk
Keep most of your position boring and safe — stable income, low costs, cash. Put a small, capped portion into high-variance bets you can afford to lose entirely. Never the middle, which is where you get all the risk and none of the upside.
7. Skills compound, tools do not
Mastery of a specific tool is worth six to eighteen months. Judgement, taste, framing, verification and relationships compound over decades. Spend your learning budget accordingly — and see the resource guide for where to learn the durable things.
8. Build in public, carefully
Showing the work is the cheapest distribution available, and it compounds trust. But publish outcomes and methods, not credentials, keys or client confidences. The people who benefit most from building in public are the ones who are visibly useful, not visibly busy.
9. Decide what you will not do
The abundance of possible actions is the actual problem now. Write down what you are not doing this quarter — not that niche, not that platform, not that tool. Constraints are the only thing that turns unlimited capability into finished work.
The uncomfortable one
Most people reading this will do none of it. That is precisely why the advantage is available. The gap between knowing and doing has never been more profitable to close.
Next: the action plans turn these principles into weeks, and AI for business applies them department by department.