I'd argue the bigger counterweight dragging down expected performance gains is lost agility.

The more parties in a system, the more ossified it becomes. (Hello, healthcare)

Inevitably, the world changes... and now because there are so many intermediary layers the system as a whole is unable to adapt.

Then you're left with a system that can't be changed, that very efficiently does something different than what you need it to do.

Or, in a nutshell, most enterprise software older than 5 years.

Have gains not been accounting tricks since the 90s?

I would say that almost all of it is, eg, disassembling our manufacturing and shipping it over seas - which ultimately eroded the middle class and jeopardized national security. But neither of those is on the balance sheets of the relevant company.

Anti-social short-term metricized business is the ultimate form of Taylorism — and in three generations, we can see that it’s an abysmal failure.

Sprinkling math on top doesn’t make reckless greed a good idea.

>It's a net loss of efficiency

Add to it the time lost because software tends to be less reliable than its counterpart because multiple software interfaces tend to increase complexity. There are some things that software is wonderful for improving. But I don’t need a IoT stick of deodorant.