Many programmers are negative about the impact of code generation on code quality and on their profession in general. I believe most of those people try to integrate code-generation into old workflows, which forces the AI to guess about how it can best serve you, but results in surprising and sub-optimal decisions. To use code-generation successfully, we have to rethink our workflows.

Below I will briefly go into the critical elements that determine code-generation outcomes. Next, I will describe how we can design our AI-enabled workflow to produce better results. Finally, I will reference some resources (that I created) which will help you to put this approach into practice. By adopting the TDD process that is described below, you will probably see improvements quickly, but I expect that a few weeks of experimentation are needed to create the full setup and workflow that suits your personal preferences.