Most of the case for AI gets made in hours. A tool will save your staff an hour a day on documentation, or cut a report from an afternoon to twenty minutes. Some of those numbers hold up and some don’t, and I’d check them. But even when the time savings are real, I think there’s a second question that matters more, and it rarely gets asked: what happens to the hour?
I said this in a post last spring and I still think it’s the most useful sentence I have on the subject. Saving the time and using the time are two different skills.
Work fills the time it’s given
In 1955 a British naval historian named C. Northcote Parkinson wrote an essay for The Economist that opened with the line “Work expands so as to fill the time available for its completion.” He was writing about British government offices, where the number of officials kept growing even as the work they were there to do got smaller. It has been called Parkinson’s law ever since.
Applied to AI, it means the saved hour doesn’t stay open. It gets filled, and usually within the week.
I know what fills mine, because I’ve watched it happen. I go back and adjust something that was already finished. I ask AI to make it better three more times. I start a new thing before the first one is done.
In an agency, the hour fills the same way, with different things. A form picks up two more fields because now there’s time to fill them out. A report gets longer. A meeting that was thirty minutes goes back to an hour. Or the hour gets assigned, and the staff member who saved time on Tuesday has one more thing on their plate by Friday.
Saving the time and using the time are two different skills.
Why no one notices
None of this looks like a problem while it’s happening. Everyone is still busy. Things are getting done, and some of them are getting done better. If you asked staff whether the tool was helping, most would say yes.
That’s what makes it hard to catch. When I wrote about this for people running their own businesses, the question I asked was whether you can tell the difference between productive and productive-looking. An hour spent polishing a finished report feels the same from the inside as an hour spent on the thing that was stuck. A busy week tells you people were working. It can’t tell you whether the saved time went anywhere.
Here is what I expect to happen in a lot of agencies. Three months in, someone asks what changed. The people you serve aren’t getting seen any sooner. The project that was stuck last year is still stuck. No one is leaving at five who wasn’t before. The agency paid for the tool, staff learned it, the time came back, and the week absorbed it.
I wrote in my first note that the time has to be re-placed on purpose, and that in an organization part of that decision belongs to leadership. Parkinson’s law is the reason. If no one decides where the hour goes, the work that’s already there will take it.
What staff are watching for
There’s a second cost, and it shows up in trust before it shows up anywhere else. Staff are already asking what AI means for them, whether or not they say it out loud. I wrote in my last note that if an agency can’t say where the time goes, people will assume the answer is more work.
Parkinson’s law is how that assumption comes true without anyone deciding it. No leader has to announce that saved time will turn into more tasks. It happens one reasonable request at a time. “Since that report is faster now, could you also take this on?” Each request makes sense when it’s made. Added up over a few months, the person who learned the tool fastest is carrying the most, and everyone around them has noticed.
That’s the part I’d worry about most. The staff who would have been your champions learn that getting faster gets them more work, and they stop telling you when something saves them time.
Decide before the time shows up
The plan for the saved time has to come before the tool, because once the hour has been absorbed it’s very hard to find again. I’d ask leadership to answer three questions for each piece of work they hand to AI.
- 1
Which hours?
Name the task and roughly how much time it takes now. “Documentation” is too big. “The weekly summary each supervisor writes, about two hours” is something you can check later.
- 2
Where do they go?
Pick one thing. More time with the people you serve, one stalled project, supervision that keeps getting cut short, or staff going home on time. That last one is a real answer, and it may be the one that keeps people.
- 3
What does done look like?
Decide what you’d expect to see in sixty days if it worked, and who is going to look.
What that looks like for one task
Take the weekly summary as an example. Say each supervisor spends about two hours on it, and with AI drafting from their notes it takes forty-five minutes, including the time to read it and fix what’s wrong. That’s a little over an hour back per supervisor, every week.
Before anyone starts, leadership says where that hour goes. Maybe it goes to supervision, because the one-on-ones have been getting cut to twenty minutes for a year. So the plan is written down in one sentence: the time supervisors save on the weekly summary goes back into full-length one-on-ones.
Then someone checks. In sixty days, are the one-on-ones running their full time? If they are, it worked, and you can see it. If they aren’t, you’ve learned something worth knowing. Either the time savings weren’t real once you counted the fixing, or the hour went somewhere else and you can go find out where.
The numbers in that example are made up to show the steps. What matters is that the hour had a destination before it existed, and that one named person was going to look.
Start with your own hour
You don’t need an agency-wide plan to try this. Pick one task of your own that AI has already made faster, and answer the three questions for yourself this week. It’s harder than it sounds. I’ve done it and still caught myself spending the hour on something I had already finished.
Doing it yourself first also changes the conversation with staff. You can tell them what you saved, where you meant it to go and where it went instead. That’s a more honest place to start than a memo about efficiency.
If the honest answer to the second question is “we need them to take on more,” say that out loud and decide it on purpose. Staff will figure it out either way, and they’ll trust the next rollout a lot less if they had to figure it out themselves.
The tool can give the time back. Someone still has to decide what the time is for, and in an agency that call belongs to leadership.