Four in the afternoon. In your inbox is a list of fifty people who all need the same message — a reminder to submit a form, a heads-up about a date, a confirmation of an order. The content is nearly identical. Only the name changes, a number, a due date.
You copy the template, paste, fix the name, fix the number, send. Then paste again, fix again. By the twelfth one you paste the right body but leave the previous person's name in. Nobody dies, but you just sent Sarah an email that opens "Hi, Mark."
01"Identical" and "different per person" are two jobs — don't blend them
Look closely at those fifty emails and they aren't fifty jobs. They're one skeleton repeated fifty times, plus a few blanks each filled differently.
Typing each by hand means re-copying the skeleton fifty times — exactly the part that never changes. And if you dump the whole list on AI and say "write me fifty emails," you swing to the other extreme: you let it invent both the skeleton and the way to fill it, for fifty people, in one shot. It hands back fifty lookalikes — and the parts that actually matter (the number, the deadline) are the ones most likely to drift, because it's guessing rather than filling.
✕ Treat it as 50 separate jobs
✓ Split skeleton from variables
The difference isn't "faster." It's that you pour all your attention into exactly one skeleton, instead of spreading it thin across fifty and praying you stay sharp every time.
02Ask first: does this repeat even need AI
This is the step most people skip. Once you've split it into "a skeleton plus a table of blanks," you notice the filling-in part needs no intelligence at all. Take the name in column A, drop it in the greeting; take the number in column B, drop it in the amount — that's mechanical, exactly-one-right-way work. And exactly-one-right-way work is done faster, and never "creatively" wrong, by a plain mail-merge.
Save the AI for the part that genuinely needs thought: how the skeleton should read so it isn't stiff, or the handful of people whose situation needs a different sentence entirely. Give it that, and it shines. Give it the job of pasting fifty names, and you pay for the privilege of a new risk — that it quietly reworded a sentence you never checked.
Write one complete sample email, marking what gets swapped: "Hi [NAME], a reminder that [FORM] is due [DATE]." This is the part you scrutinize — once.
One table: a row per person, a column per blank in the skeleton. The real data lives here, separated from the prose.
The exactly-one-right-way filling: let a mail-merge or a small script do it. Only pull AI in if you need it to reword for flow or handle the off-template cases.
Fifty correct merges are supposed to look alike. You only need to glance at the rows whose data looks strange — missing name, impossible date — instead of reading every email.
These four steps split one pile of "repetitive work" into: one job that needs a brain (the skeleton), one that needs clean data (the table), and one that's mechanical (the merge). Each goes to the right worker — and the "worker" here might just be a built-in feature.
So next time you're staring at fifty near-identical somethings, don't start by typing the first one. Ask first: what repeats verbatim, what differs per person? Split them, and you'll find the repeat usually doesn't want AI, it wants a script, while the part that does want AI deserves a clear instruction, not "write me fifty". The thing that wore you out was never fifty emails. It was that you hand-copied one skeleton, fifty times.