[Jensson]

As you make something cheaper demand for it goes up. Easy to forget about it when you try to calculate the value of automation. I'd say in most cases automation pays off if you do it properly, not to do the intended task but to do all the unintended tasks, reusing the automation code for later, reducing errors from manual work and opens up doors for new product design paths in the future.

This assumes that you actually own the product, if you don't own the product and just have a job to do then why bother about saving time, so in this case you should still automate it because automation is more fun.

[hinkley]

We had a build process that took half an hour, so people only ran it a couple times a day. I cut it to fifteen minutes, saving half an hour a day per person.

Then I cut it to 7 minutes, saving an additional 30+ minutes, because everyone was now running the build 4-5 times a day.

What I really saved was wasting each other’s time and goodwill by reducing the frequency and severity of people checking in broken code that blocked everyone else. Because they were expected to test more often, and they made smaller commits because it wasn’t so painful.

[mjw1007]

Right, so we have a number of advantages of automation that aren't captured in a simplistic time calculation like the "Is It Worth the Time?" XKCD:

- if you can do something more quickly you might also choose to do it more often

- automation can reduce human errors

- once something is fully automated it can be used as a building block in a larger piece of automation

I can think of one more:

- automation can make it easier to hand over work to someone else, or increase the number of people who are able to do it

But also one disadvantage:

- automating something can lose a human "sanity check" (for example, you could generate reports full of missing data and maybe nobody will notice for a while)

Any more?

[Jensson]

> - automating something can lose a human "sanity check" (for example, you could generate reports full of missing data and maybe nobody will notice for a while)

Yeah, you shouldn't automate things that you don't know how to automate. Don't use ML for automation for example.