I cancelled an automation tool this week. Paid subscription, been running for months, doing exactly what it was built to do.
And it was making my life worse.
Not because the tool was bad. Because I had it wired to automate work that didn’t need doing in the first place. It was faithfully pushing content to places nobody was watching, on a schedule I picked back when I had a different plan, and I was paying every month for the privilege.
That’s the part nobody tells you about automation. A machine doing the wrong work is worse than a man doing the wrong workbecause at least the man gets tired and starts asking why.
A machine doing the wrong work is worse than a man doing the wrong work, because at least the man gets tired and starts asking why.
Everybody sells it that way. Do more, do it quicker, ten times the output. And that’s how people end up producing ten times as much of something nobody asked for.
Automation is about taking a decision you already made and making it stop needing you. That’s it. That’s the whole thing.
Which means there’s a step before the tool, and almost everybody skips it: you have to have already decided the thing is worth doing. If you ain’t sure, automating it just means you’ll be unsure faster and at scale.
One: does this actually need to happen? Not “is it nice.” Not “does everybody do it.” Does the business get worse if it stops? Half the tasks people automate should have been deleted. Deleting is free. Automating costs money every month, forever.
Two: is it the same every time? Automation eats repetition. It chokes on judgment. If the task changes based on context, mood, timing, or who’s on the other end, you don’t have an automation problem: you have a system-you-haven’t-defined-yet problem.
Three: what happens when it breaks? Because it will. And a broken automation is quiet. It don’t call you. It just stops, or worse, keeps going wrong on schedule. If nobody would notice for two weeks, you didn’t build a system, you built a liability.
Let me be specific, because vague lessons don’t teach nothing.
I had a tool set up to take my long-form video and push clips out automatically. Sounds smart. It was smart, on the day I set it up. But I built those rules around a plan I later changed. Different platforms mattered, different formats mattered, and the automation kept running the old plan with total confidence.
It never occurred to me to check, because the whole point of automating something is that you stop thinking about it. That’s the feature and that’s the trap, and they’re the same thing.
Months went by. Bill went out every month. Work went out every week. Neither one was doing anything for me.
When I finally sat down and looked, the answer wasn’t “fix the settings.” It was this shouldn’t be running at all. So I killed the loops and cancelled it, and my week got better immediately, not because I got faster, but because I stopped paying for motion.
I keep coming back to this because it keeps being the answer.
Motion is busy. Movement is going somewhere. Automation is the most powerful motion-generating machine ever built. It’ll produce activity forever without ever asking whether the activity matters.
Motion is busy. Movement is going somewhere.
So the discipline ain’t technical. It’s the same discipline as everything elsethe audit that stops you repeating yourself, pointed at your tools instead of your habits. What’s actually running right now? Who checked it last? What would break if it stopped, and would anybody notice?
Most people can’t answer that about their own stack. I couldn’t.
Here’s the real math, and it ain’t about hours.
Time and energy is two different currencies, and people keep paying in one and budgeting in the other. A task that takes twenty minutes but sits in your head all day didn’t cost you twenty minutes. It cost you the day.
That’s the work worth automating. Not the longest task. The one that follows you around. The thing you think about in the shower. The thing you check your phone for. Automate the tasks that occupy you, not just the ones that take a while, because occupied is the expensive part.
Every content operation runs the same loop whether you named it or not:
Make it → put it where people are → see what happened → do more of what worked.
Four steps. Most people automate step two and never build step three, which is exactly backwardsdistribution without measurement is just louder guessing. You’ll fill every platform and still not know which one is worth your Tuesday.
So build the loop by hand first. Ugly, manual, embarrassing. Run it three or four times. You’ll learn two things: which steps are genuinely identical every time (automate those), and which ones secretly required you to decide something (never automate those, and stop pretending you can).
The tool comes last. The tool is the easy part. There’s a hundred of them and they all work fine. That was never your problem.
Here’s the thing I didn’t expect when I started actually building systems instead of buying them.
Most of what I wanted automated wasn’t a task. It was a decision I kept re-making from scratch. How long should this be. What tone. Which links. What do I never say. Every single time, starting over, burning the same energy on the same question.
The fix wasn’t software. It was writing the decision down once, plainly, so it stops being a decision. That’s a standard. And a standard is the cheapest automation there is: it costs nothing monthly and it works even when the internet’s down.
If you want the version of this I use for structuring what I hand a machine, the six layers I stack before I ask AI for anything is the same principle, aimed at a different listener. Vague in, vague out. That’s true of a person, a tool, and a subscription.
People use these words like they’re interchangeable and it costs them.
You automate a task. You delegate a decision. A task has one right answer every time. That’s why a machine can hold it. A decision has context, and the moment you hand context to something that can’t ask a follow-up question, you’ve built a very fast way to be wrong.
You automate a task. You delegate a decision.
The test is simple: could you write this down completely enough that a stranger could do it correctly on their first day, with no questions? If yes, automate it. If they’d have to come ask you something, anything, that’s a decision wearing a task’s clothes.
And here’s what most people do with those: they automate them anyway, then spend more time fixing the output than the task took by hand. That’s not a system. That’s a hobby with a monthly fee.
If you only ever set up one, make it the one that runs when you’re not there.
Not the task that takes longest. The one that has to happen on a schedule you can’t always keep. The thing that breaks if you’re sick, or working late, or the kids got a game. Automation’s real gift ain’t speed. It’s that it doesn’t have a bad week.
For most people building something on the side, that’s one of three things: the thing that goes out on a regular day, the follow-up that has to happen within a certain window, or the record that has to get written down while it’s fresh. Those three fail the same way, not because they’re hard, but because life happens on top of them.
Everything else can wait. Genuinely. The stuff you do while you’re already at the desk was never the bottleneck.
Last one, and it’s the one that bit me.
Every automation you set up needs one line written somewhere a human will find it: what it does, when you built it, and what it assumes. That last part is the one that saves you.
Because the assumptions are what expire. Not the code. The code runs fine forever. It’s the reason behind it that rots. Six months later the plan changed and the machine didn’t get the memo, and it can’t tell you it’s confused because it isn’t. It’s doing exactly what you said, perfectly, in a world that moved.
That’s the same discipline I use for everything else I want to survive mewhat pressure actually reveals about how you built applies to systems as much as to people. The load finds the weak spot. It don’t create it.
And if you’re going to be the one holding all this, keep the resources in one place instead of scattered across tabs and foldersthe vault where all the resources actually live exists for the same reason. Scattered systems are the ones that quietly stop working.
Do this one this month. It takes twenty minutes and it usually pays for itself immediately.
Pull up every recurring charge on the business card. For each one, answer three things in writing: what it does, when I last used it, and what breaks if I cancel it today.
If you can’t answer all three, that’s your answer.
Do not skip this because it feels small. Subscriptions are the quietest money you spend. They don’t ask, they don’t remind you, and they renew whether the thing is working or not. The tool that’s automating nothing is still automating the payment.
And while you in there, check what each one is actually doing, not what it was set up to do. Those drift apart, and the drift is invisible until you look.
Don’t start with a tool. Start with a week.
Write down every repeatable thing you did for seven days. The actual boring log, not the version where you look productive. At the end, sort it into three piles: delete, decide once, automate.
The delete pile is usually the biggest and it’s free money. The decide-once pile becomes your standards, and that’s where most of your time comes back. Whatever’s left after those two, the genuinely identical, genuinely necessary, genuinely repeatable, that’s the automation pile, and it’s a lot shorter than you think.
Then buy one tool for one job. Watch it for a month. Check it does what you think it does.
Dig for what’s actually eating your week. Build the system that carries it. Grow past the version of you that had to hold all of it in your head.
And check on the machine sometimes. It ain’t as smart as you hope, and it never gets tired enough to ask questions.
