Behind the Book

The Nine Emails That Almost Ruined a Payday

If you ask me what the single most important thing I do in a given week is, running a business with employees and contractors, I’d tell you it’s payroll. Not marketing. Not sales. Payroll.

People will forgive a lot of things from a business owner. What they will not forgive, and what they should not forgive, is you messing with their paycheck. I’ve worked for companies that missed payroll before, and I know exactly what that does to morale. So I’ve always built policies specifically designed to make sure I never have a payroll mistake.

The Part That Was Just… Busywork

The heavy lifting — direct deposit, tax forms, compliance — was already handled by an online payroll platform. That wasn’t the problem. The part still sitting entirely on my shoulders was the reporting: every pay period, I’d manually enter what everyone’s pay should be, then email out the reports myself, timed to land at the right moment.

That’s a completely reasonable thing to want oversight on. It’s also, if I’m honest, a task that had turned into a lot of repetitive, low-value busywork — the same kind of formatting and forwarding, on a schedule, for years.

So I Automated It. Mostly.

I had AI help me build a script that pulls the information, formats it correctly, schedules the email, and sends it — while still giving me a clean report so I can verify everything before it’s final. I didn’t hand over the decision-making. I handed over the data entry and delivery, which were never where my actual value was being added.

I’d love to tell you it worked perfectly the first time.

It did not.

The Nine Emails

While I was setting up the new email notification piece, the system got built wrong somewhere in the scheduling logic. It sent everyone on payroll the same email nine separate times in a single day before I caught it and got it fixed. Nine identical pay report emails, back to back, landing in the inboxes of the exact people I was trying to make this smoother for.

Fortunately, everybody was understanding about it — I think most of them got a decent laugh out of it, honestly. But it’s a good, humbling reminder: you double-check what AI builds for you, especially anything touching money or people’s inboxes, before you trust it to run unsupervised. I’d checked the report contents carefully. I hadn’t stress-tested the sending logic nearly as carefully, and that’s exactly where it broke.

Twenty Minutes a Week

Once it was actually fixed, the results were almost anticlimactic in the best way: I’d guess it saves me somewhere around twenty to thirty minutes a week. Not a dramatic transformation. But twenty or thirty minutes a week, on the one task in your business you can least afford to get wrong, adds up to something real — plus one less recurring thing sitting in the back of my mind every pay period.

Not everything AI does for you has to be a headline. Sometimes it’s just twenty minutes back, every week, on the task that matters most.


This is adapted from Chapter 14 of Small Business, Smart Tools — a real-world case study in using AI to run a business, teach a class, and live a life.

Small Business, Smart Tools book cover
SMALL BUSINESS, SMART TOOLS
Get the full story.
This is one story out of twenty-one. Get the book for the rest — real, unfiltered, and occasionally embarrassing.
Get the Book →