Automating My Own Job

A lot of payroll support is the same motion repeated: pull the record, run the checks, reformat the data, draft the same explanation, catch the same discrepancy. None of it is hard. All of it is slow — and slow is where mistakes hide.

The repetitive part wasn't just tedious; it was eating the time that should have gone to the cases that actually needed judgment. And a tired human doing rote checks at volume is exactly how a wrong number reaches a payslip. The bottleneck wasn't skill. It was attention spent in the wrong place.

So I started automating my own job — leaning on AI and light scripting to take the rote part off my plate: pulling the right data, running the reconciliations, drafting the first version of the explanation, flagging the discrepancy before it became a ticket. Rough at first, then reliable enough that I trusted it with the boring eighty percent and spent my own attention on the twenty that needed me.

This is where the habit started. The production AI skills I run on live payroll now — drafting contracts, matching census records — are the same instinct with sharper tools: get the machine to do exactly what a machine should, so the human is free for the part that needs a human. I just got tired, early, of doing by hand what didn't need a hand.