Developers' key value is their mindset for analyzing problems, and their ability to identify vagueness, contradiction, and mistakes in the given task, go back to the client, and talk through the edge cases and issues. AI might replace code monkeys who never even attempted to improve themselves, but as with every no-/low-code solution, management will quickly find that a) it's harder than it looks, as they don't have the mindset to clearly communicate the task in language the tool understands (this includes using domain-specific business jargon that the AI won't have trained on, or references to concepts prevalent in that specific company's internal email discussions), and b) a dedicated programmer has a time-efficiency bonus that makes it cheaper for them to do the work than a manager, so might as well delegate to the specialist anyway and free up time for managing other aspects of the business.

Thing is, developers are constantly creating new languages and libraries in an attempt to more concisely write their intentions in a way the computer can understand. Dropping back to human grammar loses a ton of specificity, and introduces a new sort of linguistic boilerplate.