But I think that two challenges / opportunities remain for analog. The first is that advances in digital capabilities place new demands on analog. For instance, the possibility of higher sampling rates and bit depths requires us to go back and update transducer preamps.
Second, some archaic skills can still be lucrative if the number of practitioners decreases faster than the number of remaining applications. I'm still betting my career on the need for somebody to care about physics, while also knowing how to program. At my workplace, I am possibly the last remaining person who knows how to compute a noise budget that includes transducers, analog signals, and digital processing, in an instrumentation system.
Horowitz and Hill had a chapter, "Digital Meets Analog" that discuss ways that digital designers have to keep up with analog concepts. The ways that digital systems can go wrong are often analog.
I bought the 2nd edition 15 years ago, I loved it, and I now cherish that useful book in my bookcase.
Analog only looks like it's falling in importance if you divide by the size of the digital industry.
The way to deal with that in the analog world is an expensive process called 'trimming' where you use adjustable components to get you where you want to be. But trimming makes the circuitry sensitive to being bounced around and makes it more sensitive to temperature changes if you use the cheap version and the expensive version (digital circuitry masquerading as trimmable analog components) puts the horse behind the cart if you wanted analog then you're no longer getting it (but some kind of quantized version of analog).
For some circuits this is all more important than for others.
Finally, digital reduces the component count and so increases reliability and decreases cost. Software is hard to make and expensive but once you have it the marginal cost of reproduction is next to zero.
Digital, once the bugs are worked out has none of these issues and so lends itself much more to economic mass production.
But another feature, not inherent to all digital circuits but certainly prevalent, is the ability to reprogram them. For me to change an analog circuit might take hours with a soldering iron. You can load a whole new "circuit" into a digital system at the push of a button. This in turn changes how products are designed. Analog functionality is used sparingly where needed, and gotten working early in a project. The digital system is left with some flexibility, e.g., allowing changes to made later in the project without drastic hardware revision. The mantra of the analog team, at the 11th hour of the project, is: "Can we fix this in the firmware?"
This thought that every problem is solved with a microprocessor shows no thought being applied to the solution, not brilliance by some engineer.