Use of AI at work

I mostly use AI to get a basic code snippet or a template, like converting lower cased text to title case, or a function to parse a value and return me something simple. In general, things that don't need any codebase context and I am aware of how it can be done as well, but get it done quickly with AI.

Other times, I have a healthy discussion on system design and patterns, trying to sketch out the best possible solution that can be used for my problem statement. It's not like I give jt a prompt, it gives me response, I make changes and we are done. It's more like it gives me a response, I point out why it did that way and not this, helping both of us understand our POVs, discussing more on other patterns, design references and so on. Helps me plan my feature development well so that I don't get stuck too much while I'm coding out the implementation.

So, just like you, I use it for very obvious and simple things. Nothing too fancy.
Reason being, I've tried models to be my pair programmers and AI assistants (claude, gemini, open source models etc.), but most of them start hallucinating when they start taking in too much context. And codebases always have a gazillion imports which simply brings in too much context for these AI models to help us out. The most context that I was able to see it go through, even to a paid online LLM was 12 files containing 300 lines of prettified code. It just couldn't take anymore and kept generating nonsense tbh.