I can see a viewpoint where it would be beneficial. It allows humans to develop things and once developed they transition into a stable operating state predominantly automated.
This, of course, is considering only the case that they achieve the ability to actually make the automation.
If a product can generate a steady return on investment by reducing costs. There is incentive to maintain that service without enshitification. The automation lacks the design ability to endlessly modify the product to add 'features' that exploit users.
The idea is that humans can do more things and are more expensive, because they can do more than just operate an existing product, if you are paying for them anyway you can put some of their efforts maximizing all those things that generate more revenue at the cost of user experience
The counterpoint is what happens when AI is a capable enshitifier.
A world with more capable AI might require more of a paradigm shift. Instead paying for a service like this, you pay for a single human who can manage an AI to deliver a bespoke service.
You are still going to need a human to understand what you want and communicate that to an AI. It'll be quite some time before people are comfortable with an AI deciding what you want and giving it to you unprompted.
Again this is predicated on AI better than we currently have.
If we do get to that stage perhaps we see AI maintained open source projects.
If AI cans do it all, why wouldn't people just ask an AI to make them their own sass.
The truly all-in AI sass would just have all API calls forwarded to an AI and it deciding what to do on a call by call basis. I could see people trying that, but I suspect when AI is capable enough to do the job, the world would have changed sufficiently that we can't predict from here what it is that we would want them to make for us.