I’d say only when AI can build highly custom branded mobile applications, marketing sites, and data driven backends.
Until then AI/ML would just be relegated to a function of business intelligence, data mining and a sub function of database engineering/administration.
And I’m speaking from experience; six enterprise level companies later after even remotely learning big data, ELK, and Tensorflow or supporting the teams that do work with it and each of those teams sole duty is business intelligence / database work.
Also that Bubble app looks interesting but just like SQL was meant to be a business tool or the hundreds of other visual development tools that are simple enough that a coder is not needed.
These too will still be relegated to a technical contributor role (e.g. a developer / engineer).
Its just the nature of business majors. Tech is something that business doesn’t do; even if a spreadsheet or word doc that can be delegated to someone else they will delegate.
There has been a long line of products claiming to end the need for programmers, starting with something on the Apple II dubbed The Last One or something like that. After that came Microsoft’s Visual series and Borland’s Delphi. Likewise YACC and LEX for compiler writers.
They do succeed in automating some of the tedious and boring parts of development such as much of the user interface work.
Today, it is targeting the drudgery of many parts of web interfaces.
You are right. There will always be things only a creative human can do. But some things a human should not have to do.
So . . . you’re saying that it’s just a matter of time?
How many years out is it?
Some say robots spell the end of factory workers.
There will be the need for people to design and maintain robots…
I’m sure the unemployed factory workers take great comfort in that. 
So . . . you’re saying that it’s just a matter of time?
Nope, the horse and carriage may have been replaced with more efficient petrol powered version and soon to be electric powered one. But we moving in a box on wheels while leading around something as it barrows down the road at speeds fast enough to mame or kill.
Well, same thing here. Each iteration of information tech no matter the marketing lingo just works to make that tech less cumbersome for end users. At the end of the day we can just use machine learning to model automation and enable data driven ui/apis but there’s still going to be some odd ball requirement from a stakeholder in either legal, marketing, business, or their parent’s basement that introduces an edge case which has custom code needed; thus additional developer work.
Besides; even mighty star trek’s AI driven science machines needed custom holodeck programs written by the crew (or the odd ferangi).
Well I can’t argue with Star Trek.