Ask what you should automate first and the internet hands you a list of tools. That is a shopping list, not an answer.
The honest answer has nothing to do with software. I have built a handful of small internal tools for a waste and logistics business, the unglamorous kind. A calculator the crew open on site to work out which disposal route is cheaper before they drive to the wrong one. A planner that lays out the day’s jobs across two vans. Paperwork that generates itself. None of it is clever. All of it survived, which is rarer than it sounds.
Here is the test that decides which job goes first, and it is three questions long.
Does it repeat with one right answer? Not just repeats. Repeats and has a correct outcome you could point at. Choosing a tip site has a right answer. Deciding how to word a difficult email does not, and no amount of enthusiasm will make it one.
Can you check the output afterwards? If you cannot tell a good result from a bad one once it lands, you have not automated a job. You have handed your judgement to something that will never tell you it got it wrong.
Does it fail loudly? This is the one everybody skips, and it is the one that bites. Automation that breaks noisily is an inconvenience. Automation that breaks quietly and keeps filing confident reports is a liability you will not find for months, because nothing looks more finished than a clean result from a broken tool.
Score your candidates on those three and the biggest time sink usually loses. It is big precisely because it is messy and full of judgement, which is exactly what machines are worst at and what you are paid for.
The first thing to automate is not your worst afternoon. It is the boring repeated job with one checkable answer and a failure you cannot possibly miss, and it is almost always smaller than the thing you were about to pick.
If this is the exact wall you are staring at, tell me what you are seeing: mail@sandeepkv.com, or message me on LinkedIn.