Journal · The AI Architect · 2026-08-16
A task is something you do once; a tool is something you run again. Do the thinking once, then just press go.
A task is something you do once; a tool is something you run again. Do the thinking once, then just press go.
The real tension isn’t laziness versus effort. It’s that most of us can’t tell, in the moment, whether we’re facing something new or something familiar wearing a disguise. So we redo the thinking every time, from scratch, because it feels safer than assuming we’ve seen this before. That instinct is reasonable. It’s also expensive. The cost shows up quietly, in hours spent re-deciding things you already decided last month.
Notice the repeat before it repeats a third time
The hardest part of turning a task into a tool is catching the pattern early enough to act on it. By the time something has happened five times, you’ve already paid the full price five times over. The trick is to get suspicious after the second time.
When you find yourself doing something and think “didn’t I just do this,” stop and ask a plain question: will this come up again in the next three months? If the honest answer is yes, you’re not looking at a task anymore. You’re looking at a tool that hasn’t been built yet.
Try this: for one week, keep a small list titled “Again.” Any time you catch yourself repeating a decision, a message, a spreadsheet setup, a way of onboarding someone, write it down. Don’t fix anything yet. Just notice. At the end of the week, look at the list. You’ll usually find two or three things that are obviously worth turning into something reusable, and you’ll wonder why you didn’t see it sooner. That’s normal. Repetition is invisible while you’re inside it.
Do the thinking once, but do it slower than feels natural
Here’s the part people get wrong when they try to build a tool out of a task: they rush the thinking because they’re eager to get to the “press go” part. But if the thinking is sloppy, you’ve just built a machine that reliably produces mediocre results. The tool doesn’t fix bad thinking. It scales it.
So when you sit down to convert something into a repeatable process, resist the urge to move fast. Ask what actually matters here. Ask what almost went wrong the last three times you did this by hand. Ask what you’d tell someone else if they had to do this without you in the room. That last question is especially useful, because it forces you to make explicit what’s usually just instinct.
Try this: the next time you build a checklist, template, or script for something you do often, write it as if you’re handing it to a competent stranger who has never seen your work before. Not a genius, not an expert — just someone reasonably careful. If your instructions only make sense to you, they’re not a tool yet. They’re a note to future-you, and future-you will forget more than you think.
Build the tool smaller than you want to
There’s a temptation, once you decide something deserves the “do it once” treatment, to make it comprehensive. Cover every edge case. Handle every exception. This is where good intentions turn into projects that never ship. A tool that handles ninety percent of your real cases and gets used every week beats a perfect tool that’s still half-finished in a drawer.
The fix is to build for the common case first, and leave a visible seam for the rare one. A template with a note that says “if this situation is unusual, stop and think it through manually” is honest and useful. It doesn’t pretend to be smarter than it is.
Try this: take one thing from your “Again” list and give yourself thirty minutes, not more, to build a first version. A checklist, a short script, a template document, a saved reply. Thirty minutes forces you to focus on the parts that matter most and skip the polish. Use it the next time the task comes up. Fix what’s actually wrong, based on real use, not imagined problems. This is how tools get better — through friction, not foresight.
None of this makes the initial thinking disappear. It just moves the effort to where it pays off more than once, instead of paying the same tax over and over without noticing. The idea is small enough to fit in a sentence, which is exactly why it’s easy to nod at and hard to actually practice.
The book goes further into how to spot which tasks are worth this treatment and which ones genuinely aren’t, because not everything should become a tool, and forcing it usually backfires. If this line landed for you, that’s the part worth sitting with next.
Go deeper. The full method is in The AI Architect. New here? Start with the free companion pack, or explore the series.