It relates to a lot of things. One thing like telephone adoption "at this rate, everyone will need to be an operator" and then we had self-dial phones instead of switchboard operators. So for simple often done things people can do it themselves. For other rare/special things we can still call for an operator (we still have those right?)

Another way to look at it is like computer architecture. We have technologies for holding data and impedance mismatches between them, so we make intermediary caches, integrated/hybrid devices and devices start blending to look like each other 3D XPoint/NVMe. On another scale a Facebook datacentre can look like a macroscale computer architecture. My point is that 'programming' and programmers started out being one thing, but there will be many different forms of this at many levels, in various vertical areas. We already have specialized platform (Shopify) programmers, and self-serve (Zapier/Airtable) and even among those there could be third-party helpers as has happened around Google sheets data collectors/analyzers. Lo-code/no-code is so many things. It comes back to 'any sufficiently capable configuration schema becomes a lisp'. People want to tweak, so they learn to program.