I like this xkcd comic for a great representation of this: https://xkcd.com/1629/
How, exactly, do you quantify the productivity of someone who makes some obscure library for a tool that provides monitoring services for some other utility that...
Hell, my own job is hard enough to quantify. I do defect analysis at Intel, looking for defects in chips with a scanning electron microscope. Sure, my output of samples is pretty easy to track, but what's the actual productivity of my work? I am one very, very small cog in a very, very big machine that somehow shits out fantastically small computer chips at the end. How important is my role? Er, I'm not quite sure. All I know is that I'm paid good money for what I do, and I work 12-hour shifts and am frequently asked to work overtime. But not once have I been told, "Hey, thanks to the defects that you and your team found, we got the information needed to increase wafer yield by X! You're doing great work!" Nope. Chips come in, pictures go out, engineers do Stuff with those pictures, and that's about as far as I can tell.
How do you measure the productivity of my work? The engineers who use the data I create to tell other people to make changes in the process? The guys who make the tools that I use to do all of this? The janitor who cleans the wastebaskets of the guys who make said tools? We keep going farther and farther down the rabbit hole, and it doesn't get any clearer.
Increasingly I'm tending to agree with a group I would previously have dismissed as Ludicrously Optimistic Socialists -- though the S word may not be quite right there -- in that pretty much everything is going to have to change, in huge ways, if we're to retain any kind of stability in the developed world.
I also work in the tech industry, as a consultant, but about 75% of my output is creative / people-related -- but that still means my part of the industry is going to shrink by at least 25% in the near future as smart machines are brought online to do the non-people/creative bits. It's fascinating to talk to people with various jobs, and ask them how long they reckon it'll be before what they do will be done by a computer / machine. Most people are hugely optimistic about the level of sophistication they bring to the table, and their consequent relative safety in the workforce.
Like, what if farming becomes primarily robotic. But, in the process of that transformation, each of us ends up with solar-powered, open-source driven backyard robots that tend gardens where the produce did not require pesticides nor chemicals to grow. Perhaps even raw materials for 3D printing and manufacturing of some household items could be raised as backyard crops.
You'd have less need to buy things that you need to survive, and so less need to work. Society probably will need to trend more socialistic, but perhaps automation will wind up allowing unemployed people to live more independent and self-sufficient than they can today.
It's a very abstract sort of "atta-boy" we need to be giving ourselves here. And companies need to do a better job of tracing success stories back up the chain and celebrate individuals for their specific contributions to the "big wins" when they occur to humanize the whole endeavor.
As far as the janitor goes, I am less enthusiastic. That role in the marketplace will be automated away because there's very little craft to that job where human participation is useful. Comparing what you do with the waste and dust collector doesn't seem right.
The happy / neutral / depressing theories (more hypotheses) don't cover all the possibilities, of course.
Weirdly, while the 'depressing' scenario is badly described (is the placement of a computer in front of ever office worker an example of 'innovation'?), the initial premise -- that productivity is / has peaked -- may be accurate, iff we try to measure productivity today the same way we have done for the past hundred years. And that's kind of the problem -- we don't currently have appropriate metrics for productivity in a market increasingly based on information, let alone where information components of products are approaching zero cost.
Anyway, I'll pass-on the hive mind's recommendation of Mason's book - I'm yet to finish it, and even further from being convinced of the soundness of all his arguments, but it's the most interesting book on economics I've read.