Imagine that there was a government agency that evaluated and approved business licensing forms. Assume these forms are very complex, and require an entire day’s worth of work to properly evaluate and approve. And let’s say that there are 1000 businesses out there sending in forms each day, and that the government has enough capacity to support that need.
This implies that there are 1000 government agents paid, say, $100k each year, to evaluate and approve forms.
Now, let’s say that someone writes some open source software that can process 1000 forms in a day. You still need at least one person, just in case, but you don’t need all 1000. The government agency installs the software, and fires 999 people.
In terms of Keynesian economics, this is a catastrophe. ~1000 people out on the street. No wages, and therefore significantly diminished spending from those people. There is a meaningful negative impact on GDP.
But yet what normal person would think this was a bad thing? But the Keynesian model suggests exactly that – that automation is something to be avoided, especially when times are tough.