Nobody Uses It
The failure mode of automation was never technical. It is that six weeks later the spreadsheet is still open on someone's second monitor, and nobody wants to be the one to say so.
The automation worked. That was never the problem.
Six weeks after it shipped, one team is still keeping the spreadsheet. Another does not trust the output enough to act on it without checking. And the person it was built to help was never asked what would actually make their Tuesday easier. The system runs perfectly, on schedule, for nobody.
I have watched this happen for twenty five+ years across three industries, and it took me most of that time to stop diagnosing it as a technology problem. Every process has three versions: the one in the documentation, the one people actually follow, and the one shaped by who talks to whom and who gets blamed. Automating the first while ignoring the other two is how you build something technically correct and practically dead.
The people are not being irrational. Behind every manual step is someone whose judgment that step represented, and someone who will be asked to explain it if the machine gets it wrong at scale. “Nobody adopted it” is usually shorthand for “nobody was willing to own the failure mode.” That is not resistance to change. That is an accurate reading of the incentives.
What has changed this year is the ratio. Building the thing has become extraordinarily cheap — I can stand up a working automation in an afternoon now, for a few cents a run. Getting an organisation to actually use it costs exactly what it always did. So the bottleneck has moved, decisively, to the half of the work that does not look like engineering.
Which means the interesting question is no longer can this be automated. Almost everything can, now, badly. The question is whether anyone will still be running it a quarter later, and whether they are glad.
That is the boring metric. It is also the only one that has ever predicted anything.