I read a line in Anthony Hunt's The AI-Powered Church that put words to something I had been circling for a year:
"AI removes the friction of repetitive tasks so you can be fully present in the moments that matter. Imagine getting hours of your week back, not to do more, but to do more of what matters most."
That last clause is the whole argument. Not to do more. To do more of what matters most.
Those sound similar. They are opposites.
Where the time goes
Here is the pattern I've watched play out.
A business automates something genuinely tedious. Three hours a week come back. Real hours, correctly measured.
Six months later, nobody has three extra hours. Not because the automation failed. Because the time got filled, immediately and invisibly, by work that expanded to occupy it.
Nobody decided this. There was no meeting where someone proposed spending the savings on more of the same. It just happened, the way water finds the low spot.
Saved time doesn't stay saved by default. It is claimed by whatever was already pressing. If you want it to go somewhere else, you have to say so and then defend it, which is harder than it sounds because the pressing thing always has a better argument in the moment.
Two different goals wearing the same clothes
Productivity asks: what else can I fit in this time?
Hours back asks: what would I do with this time if nothing had to fill it?
Almost every automation is sold on the first framing and almost every person I talk to actually wants the second. They do not want to process more invoices. They want to stop doing invoices at nine at night.
The technology is identical. The outcome is entirely determined by which question you were asking when you deployed it, and most people never explicitly asked either.
I named my own work around this deliberately. Giving people their hours back, not making them more productive. Partly because it's what clients actually want, and partly to keep myself honest, because the productivity framing is much easier to sell and I would drift toward it otherwise.
The measurement trap
Part of why this goes wrong is that one side of it's easy to count.
Output is countable. Invoices processed, tickets closed, posts published. It goes in a report and someone gets credit.
What the time was for is not countable. Dinner without a laptop. A conversation that wasn't rushed. Enough margin to notice a problem before it became one. None of that produces a number, so none of it survives contact with a system that manages by numbers.
This is not an argument against measurement. It is a warning that measurement has a bias, and the bias runs consistently against the thing most people were actually trying to buy.
If your only metric is throughput, you'll optimize into a place nobody wanted to go, and every individual step will have looked correct.
The harder question
There is a version of this that is uncomfortable and I would rather include it than write a tidy post.
If you got three hours back this week, do you know what you would do with them?
A lot of people do not. And the reason the time gets absorbed is not really that work is greedy. It is that unstructured time asks you what you actually value, and busyness is an effective way to avoid answering.
Filling the hours is easier than deciding what they're for. It also feels productive, which makes it a very comfortable place to hide.
I do not say that from a distance. I've caught myself using a genuinely well-automated week to take on more work rather than sit with anything, and calling it growth.
The practical version
If you're automating something, decide in advance where the time goes. Write it down. "This gives Sarah four hours a week and those hours go to client conversations, not to absorbing more accounts."
Then check in ninety days. If Sarah is doing more accounts, the automation didn't fail. The decision did, or nobody made one.
And if you're automating your own work, the same question, harder to dodge: what's this for?
Because if the answer is only that you will get more done, you have bought a faster treadmill. Which is a real product, and some people want it.
But it's not the thing most people are hoping for when they ask me whether AI can help. What they are usually asking, underneath, is whether they can have some of their life back.
That is available. It just doesn't happen on its own.