Want AI to Take Over a Repetitive Task? Explain It Like You Would to Another Person

June 26, 2026

You sat down with high hopes, asked the tool to handle a real task from your week, and watched it hand back something that missed the mark.

So you closed the tab and thought, “Well, AI isn’t there yet.”

I hear some version of that a lot. Someone tries to hand off a task they do all the time, the result falls short, and they walk away sure the tool wasn’t up to it.

Here’s the part worth slowing down on. Most of the time, the tool wasn’t the problem. The instructions were.

And the fix has almost nothing to do with AI.

Why it couldn’t do it

The work you do every day, you understand by feel. You’ve done it so many times that the steps have gone quiet. You don’t think them through anymore, you just do them.

That’s a good thing for getting work done. It’s a hard thing when you try to hand the work to someone else, because the steps you skip in your head are the exact steps another person, or a tool, would need spelled out.

There’s a classic exercise that shows this. Write out step by step how to make a peanut butter and jelly sandwich, then hand your directions to someone who follows them literally. They smear peanut butter on an unopened bag of bread, because you never said to open it. You assumed it. We assume a hundred small things about our own work without noticing.

So when you ask AI to do a task you’ve never actually written down, you’re not really asking it to do the task. You’re asking it to guess the parts you left out. That’s where it falls short.

Forget it’s AI for a minute

Here’s the most useful shift I can hand you. When you want a tool to take over a repetitive task, stop thinking about the tool. Think about how you would explain the task to another person.

Not a tech question. A teaching question. If someone were going to take this off your plate, what would they need from you to do it well? The what, the why, and the how. Enough that they could run with it without you hovering over their shoulder.

That first time through, you’re not really doing the task. You’re training. You’re writing down how this works so it can be carried forward with excellence, faster, every time after. The effort you put in once is what makes every future round quick. And notice that none of that thinking is about AI. It’s about getting clear on your own work.

What this looked like in real life

Let me give you a real example from this week, because it makes the idea concrete.

Someone I work with sits through a recorded public meeting and afterward writes a memo for the public. The goal is that anyone who missed the meeting could read the memo and understand what happened on each agenda item, the key points and the discussion around them.

Right now that takes real time. They get the agenda, they get a set of notes someone took, and they build the memo. Then a second person who was in the meeting reviews it to make sure nothing important got left out. At the start, that was two to three hours of combined work for one memo.

So we set AI aside and just asked: if we were going to teach a new person to do this, what would they need to know? We walked through it out loud.

What’s the structure of the memo, and why is it built that way? What source material are we working from? Who’s the audience? What is the reviewer actually looking for when they read it, and what does a finished, “yes, that’s right” memo look like to them?

Most of those answers had never been written down.

They lived in someone’s head. Done by instinct, over and over, but never spelled out where another person could pick it up.

One nice moment came up in that conversation.

The worry was, do we have to capture every single thing said in the meeting? We landed on a better answer. Since the meeting is recorded, the memo could include the timestamp where each agenda item was discussed.

The summary covers the key points, and if a reader wants the full back and forth, they jump straight to that spot in the video. That detail lives in the transcript, so the tool can pull it while it builds the memo.

That solved the real problem. The memo doesn’t have to hold everything. It holds what matters and points to the rest.

The question that actually matters

Notice what we were doing in that conversation. We were not asking “how can AI do this?” That question is backwards, and it’s where a lot of people get stuck.

The question was simpler. If we were going to ask anyone to do this, a person or a tool, what do we need to think about?

• What do we need to have defined?
• What answers do we need ready?
• What would someone need in order to repeat this with confidence?

Once those pieces are clear, the structure, the source, the audience, the review criteria, the format, they can go into a project inside your AI tool and get used again and again. The repetitive task gets handled, your time goes to other things, and you have the comfort of knowing someone else could step in and do it too.

Breaking the work down into those pieces is the part most people find tough, because so much of it has been living quietly in your head. That’s the work. But it’s also where almost all the value is.

Where to start

So here’s the shift. When AI “couldn’t do it,” that usually isn’t a sign the tool fell short. It’s a sign the task was never defined, and you can fix that without thinking about AI at all.

This is really what executing is. It’s taking something you already know how to do, something you could describe if someone sat you down and asked, and writing it out clearly enough that it can be repeated. Define it, and you can scale it.

You don’t have to map your whole week to feel this. Pick one task. A repetitive one you know cold. Sit down and describe it the way you’d explain it to another person: what it is, why it matters, how you want it done. That one exercise will get you further than any prompt.

Even the most skeptical folks, the ones who don’t love trusting AI with judgment calls, tend to see the value here.

This isn’t asking the tool to think for you. It’s setting up an assembly line for the work you already understand.

That’s a fine place to start.

Related Resources

You Don’t Need a Plan to Start With A.I. You Need One Thing.
If this left you thinking “fine, but which task do I even start with,” that piece helps you find the one worth picking first, so you’re not trying to fix your whole week at once.

A Simple Way To Move Past Your A.I. Anxiety and Be Highly Productive
If picturing AI as someone you train is what made this feel possible, this is where that idea takes root, and it meets you on the days the whole subject still feels intimidating.

Where Does AI Fit in Your Business? Start With the Problem, Not the Tool.
If the “stop thinking about the tool” part is what stuck with you, this carries the same instinct beyond a single task and helps you see where AI is worth your time across the business.