Building an AI Strategy

1. Use it

Imagine two people doing the same office work. One of them uses AI; the other doesn’t. Invariably, over time, the person using AI is more productive.

2. Choose frictionless

People often ask “which AI?”, which seems a reasonable question. The answer is straightforward: the one that makes you change your process the least. Any accommodation you make, or new thing you do to accommodate AI, won’t matter a year or three from now. The friction gets worn down, and the effort you spent accommodating something won’t matter, because it will have changed.

3. You can improve the quickest

Picture yourself using AI in six months. You’ll know more, and the AI will be better. If you could hand today’s self the prompts you’ll be using then, you’d already be farther along.

The AI improves on its own — you can’t speed that up. But your view of what AI can do is the part that can move fastest, because it is the part you control.

4. Requirements are more valuable than code

Any code written by a coding agent today is likely to be improved by the languages and coding agents of the future. The essential thing to capture is the need — the requirement. It isn’t the code you create now that is valuable; it is your understanding and formulation of the requirements. Requirements are often the hardest part of any project, and they are the ones most uniquely tied to the domain knowledge.

5. Embrace Change

Given that AI is rapidly improving, the process or tool you are using now won’t be competitive a year from now. I’d guess about 20% of your time should go to learning the newest and best tools, so you stay in pace with the progress and can leverage the new capability.