When should Developers use AI?

By Tucker Bichsel

5/25/2025

There is no denying that AI has the possibility to greatly increase the productivity of a developer. But there is also the risk of being too reliant on it. If used too much you are skipping the learning stage during the development process. So how do you thread the needle to get the productivity boost without the negative side effect of hampering your growth?

It comes down to when and for what you are using AI for, so here is my list of good times to use AI.

Inspiration

Getting started is sometimes the hardest part. I often like to use chatGPT to bounce ideas off of or just straight up asking what i should work on next. If I do have an idea I might ask for some questions to round out my vision.

When you don't know what you don't know

I personally believe this is one of the biggest benefits of LLMs. Before I often had to know what I was looking for to google it, but now I can describe what I am wanting to do and I will get an answer telling me what I can use to achieve that goal. CSS is a good example of this, I don't have a great grasp on CSS classes but if i can describe a style I want then chatGPT can give me the classes I should be using. Then I can dive deeper into how that class works and work on understanding it so I can get the effect I want. Before I may not have ever even found that class, just because I didn't know I was looking for it.

I like to think of ChatGPT as someone who knows about everything. But not necessarily how to do everything with everything. So it can help me find the what then the how is more my responsibility.

Implementing a 3rd party dependency

I don't really see value in spending time to fully understand how to configure a tool I am using to achieve my goals. It's none distinguishing to implement a public tool. Instead i can just get the config created for me for my use case by just asking.

Fixing Cryptic Errrors

I wish i had a dollar for every time I read an error message and just had no idea what it was saying. But now I can just paste the error message and chatGPT translates it into actionable steps I can take remediate the error. No more praying someone else has encountered the same archaic error and got an answer already.

Improvements

Finally I would say using AI to improve your implementations is a good use case. There are degrees of this you could do, ranging from not using AI until you think you are done then asking for help cleaning it up to basically just asking AI to clean up it's own code. And with that there is a range of the benefits. If you skew to the former then you are still building skills and training your brain to think like a developer, then AI can come through and help iterate on your design and implementations. Bonus points if you take the time to understand why a change might be a good idea.

Summary

I have a complicated relationship with ChatGPT. It's very helpful and does a lot of work for me, but it's also an idiot (I mean this in a friendly kind of way, like how you would call a friend an idiot. Please don't target me first.). The trick is to know when it's being helpful and when it's blowing smoke, since chatGPT is confident sometimes it's hard to tell without some knowledge on the subject yourself. Which leads to the problem of relying on it too much. It may make it faster to slap something together, but if you don't learn from the process then the next time you want to build something you aren't any better at it.