ML can't learn to understand the business needs of a company, that's something that even external human developers are always struggling with, and it can affect

ML can't learn to understand the business needs of a company, that's something that even external human developers are always struggling with, and it can affect usefulness of application much more than occasional bugs. On the other hand non-developer employees might not be able to produce optimized code and they'll make rookie mistakes, but they know how their company works, they understand the business needs and processes. IMHO the win-win approach would be to have apps designed by employees with drag&dropping UI and wizards for setting the logic rules, and then to let ML analyze it and generate the high quality code based on that.