The thing you do identically every morning — that's the machine's work

The coffee hasn't cooled yet and you've already opened five tabs, copying each cell exactly like yesterday, exactly like tomorrow.

You do it fast, almost on autopilot — hands so practiced your mind wanders elsewhere. And that "autopilot" is the signal. A job your hands can do while your mind is somewhere else doesn't need you sitting there. It's just taking the first ten minutes of your day — ten minutes that should have gone to the work that actually needs a mind.

The reason you still do it by hand is a reflex by now: "it's only ten minutes, programming this would take longer." Sounds reasonable. But it misses how those ten minutes compound.

01Ten minutes is never ten minutes

Repetitive work has a treacherous property: its cost lives not in one instance but in the count. Ten minutes each morning, times the number of days you'll keep doing this, is the real number. And the real number is far bigger than it feels:

The ring spreading out isn't ten minutes — it's a working week of your year, poured into one motion that never needed you. And that's only the part a clock can measure.

The part a clock can't measure is more expensive still. Mechanical repetition first thing in the morning erodes the one thing you most need for hard work: freshness. You spend your best slice of focus copying cells, and by the time a real decision needs it, it's gone. The true cost of doing repetition by hand isn't just time — it's the quality of everything else in your day.

02Three kinds of repetitive work, three different tools

This is where many newcomers veer wrong the other way: they hear "give it to AI" and throw every repetitive job at AI. But not every repetition suits AI. One question separates them: does this job have one fixed right answer, or is it different each time and needs reading-comprehension?

All "repetitive work," but the three kinds demand completely different things. The skill isn't handing off a lot — it's handing the right kind to the right tool.

Telling type one from type two is where the real token and money saving lives. A job with rigid rules — reformatting dates, summing by column, filtering on a clear condition — handed to AI is both more expensive (every run costs) and less reliable (it might misread a line) than a script that runs identically every time at near-zero cost. Conversely, work that needs comprehension — sorting feedback, summarizing long text, drafting context-aware replies — is beyond a script; that's the agent's turf.

03Picking the wrong tool costs more than doing nothing

Here's what to carry away. The trap isn't "not automating" — most people eventually realize the repetition should be handed off. The subtler trap is automating it the wrong way. Hand script-able work to AI and you pay AI fees for what a one-line command does for free, plus the new risk of it misreading. Hand judgment work to an agent and trust it blindly, and you've pushed a consequential decision onto something that bears no responsibility. Both cost more than leaving it alone.

So before you open a chat to hand off a repetitive job, pause one second and ask: does this have a fixed answer? If yes — a script. If it's different each time — an agent. If getting it wrong is consequential — keep it. That sorting reflex, after a few tries, becomes second nature, and it's the difference between someone who genuinely saves time and someone who's just busy in a new way.

Two deeper branches of this have someone who went ahead of you: when a one-line command beats the AI, and when the agent costs more than it's worth. That coffee cooling beside five tabs — it isn't worth a working week of your year. But it isn't worth paying AI fees for what a script does in a blink, either.