I'm observing a process to purchase an AI Coding assistant right now, and it made me wonder, when the rubber hits the road, how much would I actually spend on the assistant?

I don't know if I'd call myself a booster or skeptic. I'm loosely speaking all in in the office, but what would I actually spend?

On the one hand, my - I dunno, thought leader-y hat would say, why wouldn't I spend 10k/head on them. These tools are amazing, they could double someone's output.

But they also are these like infinite toys that can do anything, and you can spend a lot of money trying all the things. And does that actually provide value. Actually provide like rubber hits the road monetary value on a per-head basis. If you're really focused and good, probably? But if you're not disciplined. You can spend a bunch of money on tokens, then spend a bunch of money on the new features / services in production, and spend a lot of your colleagues time cleaning up that mess.

And this is all human stuff. This is all assuming the LLM works relatively perfectly at interpreting what you're trying to do and doing it.

So like. Does it actually provide the benefit of X dollars per engineer per year? Because it wouldn't have to, it could in fact go the opposite.