Depends on the business. Bespoke logic maybe, but rules based logic such as tax considerations can be abstracted away by APIs etc. Look at the difference in LOE of starting an ecommerce business in 1999 vs 2022 for example.

Also when you consider something like Docusign, you can have a pretty much feature complete product that requires little labor to maintain. You can always expand the scope of any product, but we started from 0 with web based/app software just a decade or two ago. Much of the common and widely useful programs will be largely perfected and require less active development as the years go by.

Same thing with github. You can continue to add features, but the core value of the product is already written and in a serverless world, with appropriate automation, can be maintained by a small team.

I really don't understand how somebody can think labor needs in software will never contract... the main point of debate should be on what a realistic timeline is.