← Back to Blog

Essay 6 min read

The Task You Dread Is Telling You What to Automate

Everywhere you look right now, the message is the same. Automate. Build agents. You’re behind. So you open a chat, ready to finally do it, and you go blank. The cursor blinks. There’s so much it could do that you end up doing nothing with it at all.

The hardest part of using AI is knowing where to point it. And all the noise telling you to point it everywhere only makes the blank page worse.

I watched this happen with a coach in one of my cohorts. She keeps a separate space in her AI tool for each client, one box of notes per person. Every morning she wants a short briefing on the clients she’s seeing that day. So every morning she opens each client’s box, copies the notes, pastes them into a fresh space, and asks for the briefing there. As she put it, just thinking about it exhausts me. Is there a way for the AI to fetch that without me copy-pasting?

That question is the whole game. Not because the copy-paste is a big problem. Because of where it points.

You already know what to hand over

The blank-page version of this is “what could AI do for me?” It’s a terrible question. It’s too big and it has no edges, so it sends you browsing other people’s clever use cases looking for one that fits. You come back with a list of things that sound useful and a nagging sense that none of them are yours.

There’s a much smaller question that works better. What do I do by hand, every week, that I resent doing? I mean the rote step in the middle of your work that you’d skip if you could, the thing you’ve done forty times and will do forty more. Leave the work you love and the thinking only you can do out of it.

That resentment is information. It’s the most reliable sign you have of what to hand over, and few of us use it on purpose. The task you dread is a spec, already written, waiting for you to read it.

Frustration is a spec

Here’s what the coach’s exhaustion was telling her.

She was treating her morning as an AI problem and feeling stupid that it wasn’t solving itself. What she had was a missing part, and the bit she hated named it precisely.

Any piece of AI help has four jobs it can do. It can reason. It can hold your notes and your way of working. It can go and fetch and do things. And it can wake up on its own and run without you asking. Most setups only have the first two. They think, and they hold what you’ve given them, but they sit there waiting for you to do the legwork and the remembering.

The copy-paste she resented was the legwork, the going and fetching from each client’s box. That’s a part her setup didn’t have. Having to remember to ask for the briefing every morning was the remembering, another part it didn’t have. Her exhaustion was pointing at two specific missing pieces with a steady finger.

Once you name the missing part, you can write a sentence like “I need it to fetch across all my notes, and I need it to run on its own at seven each morning.” That’s a brief, and you can hand a brief to someone, or to something. You can’t hand over a sigh.

What this changes for you

You don’t need to know how any of it is wired to use this. You need to know which part is missing, and the part you’re missing is always sitting underneath the task you avoid.

So do this. Take one thing you do most weeks and dread. Pulling figures into the same report. Reformatting notes after every session. Chasing the same follow-up. Writing the same kind of message to a new person for the hundredth time. Pick one and look at what you’re doing with your hands and your attention.

Then ask which part of it you wish would just happen. Is it the fetching, the gathering of stuff from different places? Is it the doing, the same change made every time? Is it the remembering, the fact that it only happens because you held it in your head? Whichever one makes you exhale is the part to hand over. Name it in a plain sentence. I want it to gather X from Y. I want it to run by itself on a Monday. That sentence is your starting point. It’s more useful than any list of use cases because it came from your own week instead of someone else’s highlight reel.

Now you build it

That sentence is where you start building. You don’t need a course or a developer. You build it in the same chat window you already use, in three steps.

Brainstorm. Describe what you want in plain words and let the AI help you shape it. Talk through what it should pull in and what it should hand back, in a form you’d use. You’re done when you can say what it does in one sentence.

Test. Run it on a real example. Read what comes back. Tell it what’s right and what’s off, then go again. Three or four passes and the output starts to match what you pictured.

Schedule. If the work is regular, set it to run on its own. Seven each morning, every Monday, whatever the rhythm is. Now it wakes up without you.

The Brainstorm. Test. Schedule. loop shown left to right: Brainstorm, co-design it in conversation. Test, run it, fix it, repeat, across three or four passes. Schedule, automate it if it earns it. The Schedule step is drawn faded to show it is optional.

That last step is optional, and that matters. Most of the value is already there by the end of Test. If the part you dread is the fetching or the doing, you fix it there and you’re finished. You only reach for Schedule when the thing needs to run without you there. The dread told you which part was missing, and it also tells you when to stop.

What AI can do is take the parts of your work that never needed you, so you can spend yourself on the parts that do. You find those parts by paying attention to what you dread. The friction is the map. Follow it.

Colin

Stay Connected

The Intelligence Briefing

Every Friday I send one idea on AI and leadership. It’s the thinking I wish someone had shared with me when I started.

Sent every Friday to thousands of subscribers.