Both of those comics assume you are automating your own manual task. It gets even more tenuous when you are fielding stakeholder requests to automate _their_ manual tasks, when developers cost substantially more than the entry level folks that end up being assigned these manual tasks, and you don't know how long the business process that requires this manual task will even persist in its current form.

That's where the opportunity may lie with no-code type platforms, to empower other departments to automate and maintain their own idiosyncratic workflows.

What is often overlooked in these sort of discussions, is the fact that there is more to take into consideration than "efficiency".

Certain tasks are in theory executable without specialised tools, if one is willing to pay for "the inefficient ways", but in practice they are just not feasible because the lengthy / convoluted workflow that results means that you are always having to focus on the minutae, which prevents treating all but the simplest issues.

For example, I could in theory write the space shuttle code in pure assembler, using a pen and paper, and huge teams. "It would just take very long". In practice you would never get it working.

I'm willing to agree that some problems aren't worth automating, especially when they have nasty or absent API surface, or when the task changes too much. On the other hand, looking only at time spent is also a very incomplete picture; I've written scripts for tasks that I was fine doing manually just because it allowed me to add free sanity checks and cut down on human error. (Story time: We had a regular process to clean up dev environment data. Coworker ran rm on prod data. We now have a script that sanity-checks the environment parameter and does the rm commands for you.) There's also value in speed beyond the pure time saved; as an extreme case, consider deploying a fix for a 0-day with ansible vs by hand to hundreds of hosts.

Part of the payoff of automation is not just time saved, but uniformity of results. Everybody makes errors when constructing reports by hand, and most of them are never even discovered, unless you start writing programs to check them.

Also, doing things quickly and easily can have all sorts of benefits, like allowing you to adjust to the unexpected.

Edit: Someone already said much the same thing. Oh well.

When I visited the Philippines, one thing that hit me was how many extra people they have standing around doing jobs that just don't exist back home. A man dedicated to a single door in a mall. Someone who calls a lift for me in the hotel. 4 workers in an empty store instead of just 1 here at home.

The low cost of labour must have an impact on creative solutions. You have little need for automation and AI when you can just throw more bodies and minds at it for a pittance.