I’m at a large firm of ~1000 employees, only about 25% are devs and everyone is being told they should be using LLMs/AI Agents or they are going to be behind. Even to say, if you don’t adopt these tools, we can find someone who will.

Why not using LLMs? That would be like employing hundreds of farmers and making sure they don't use a tractor and do everything by hand instead?

LLMs can be a huge performance boost, when used wisely (i.e. not just blindly using whatever they spit out)

Most people are not, in fact, wise

But the way they learn to be wise in the context of using LLMs is to try using them and fail, just like all learning experiences. Companies insisting on the use of these tools seems logical to me when the assumption is that they will, once learned, be better than previous methods of working, but only with practice.