In the Iran War, AI appears to have helped with operations rather than strategy

Overheard by a career diplomat whose country is not a major player in the US-Iran conflict, paraphrased from memory (probably imperfectly):

This was a war full of mistakes. The United States underestimated Iran’s resilience, overestimated the chances of regime change, and failed to anticipate Iran’s countermoves.

But the USA probably used AI in its strategy. It looks like AI helped with operations but wasn’t good with strategy.

Why should that be so? At first glance, I can think of at least three reasons, all of which should have been obvious before the bombs were dropped.

First, more than, say, writing requirements memos, strategy requires a broad, deep understanding of the world, but generative AI never had robust models in the world.

Second, strategy (particularly in a situation like this) requires the ability to project beyond past data into new situations. Generative AI was never particularly good at it.

Thirdly, there is the well-known tendency…