Does AI actually make you more productive, or just make you switch tasks faster?

I've noticed something interesting while using AI tools.
It's definitely faster to get started on things. I can ask for an explanation, generate a first draft, debug something, or try a completely different approach in a few minutes.

But there's another side to it: because switching tasks is so easy, I sometimes switch tasks too much.

One minute I'm debugging something, then I ask AI about a different problem, then I end up reading about a library I wasn't even planning to use. An hour later, I've done a lot of things but haven't actually finished the thing I started.

So I'm curious how other developers handle this.
Has AI genuinely improved your productivity, or has it made context switching easier and therefore more tempting?

Do you have any rules for yourself when using AI while working?

Comments

[IndiaInfraNotes]

Same experience here. The rule that helped me most is one open question per task: if a new curiosity comes up, I write it in a parking list instead of asking the AI right away, then batch those at the end of the session. AI cuts the cost of starting things to near zero, so the scarce skill becomes deciding what not to start. Do you track finished tasks per day, or does it just feel scattered?

iin1005ev