I will share my experience. AI is speeding up the bootstrapping phase of the project or prototyping. However, when you are working on a project with millions of lines of code, it is approximately helping 5-10% only with some autocompletion or googling. In addition to that, since AI is not retrained on up-to-date data, it sometimes uses old APIs or old information, old methodologies. That is one of the downfalls, I think, at the moment. The biggest downside is that we will face the result in the future, where new graduates and junior developers will use it heavily at the moment; they will not be able to learn the details or nuances properly. That makes me scared about the future because overall software quality will collapse/drop in the future, in my opinion.