We’re convinced AI is saving us time. But on hard, messy work, does anyone still remember how long it used to take? And should saving time even be the goal anymore?
I tell people AI saves me time. And every time I say it, I believe it.
But for much of what I do, I can’t be sure it’s true. AI worked its way into my days so gradually that I’ve lost track of the “before-and-after.” Outside of emails and repeatable data analysis, I honestly can’t remember how long my work used to take.
I’ve also stopped caring. Today’s work and yesterday’s work are apples and oranges. As AI capabilities grew (and my imagination grew with them), I changed the job itself.
My deliverables are bigger. My ideas are better researched and better vetted. The time I used to spend building slides now goes to simulating the tough questions I’ll face in the room. I might even spend more time than I did before, because those simulations send me back for extra rounds of revisions.
I may be saving time on the core job, but I keep inventing new, valuable work to fill those hours. The idea that AI saves me time is a vibe. Not a fact.
And yet PMOs rolling out AI get asked to report time savings constantly. How much faster will AI make your PMs? How many more projects can each PM take on? The answers drive budgeting, training, hiring, and allocation.
But if my experience is as common as I suspect, two things are increasingly true. First, real time saved is getting much harder to measure. Second, while we’re all busy reporting on productivity, we’re missing where AI may be more valuable than the time it saves.
I can’t measure how much time AI has saved me. Can you?
Even before AI, people were terrible at estimating how long their work takes. I’m no exception.
But even with tasks I actively monitored, I noticed something interesting. At first, I could measure the time savings. Over time, though, the task duration crept back up. Now that I had AI helping, I kept inventing new ways to do the task better.
Of course, “better” is subjective. Sometimes “better” meant a more robust set of deliverables, which gave the client more value. Sometimes it meant the task hurt less, so I started it earlier instead of putting it off. And sometimes “better” just meant more credible work: more research, more vetting, more red-teaming of ideas I’d gotten precious about.
I doubt I’m alone here. In 2025, METR, an independent research group, ran a trial with experienced software developers. Famously, developers using AI took 19% longer, but believed AI had made them 20% faster.
That finding isn’t the interesting part.
When METR tried to run the study again in early 2026, it couldn’t get a clean experiment. Developers got harder to recruit. A growing number said they wouldn’t want to do half their work without AI, and between 30% and 50% held some tasks back because they didn’t want to do those tasks without it.
What this tells me is that a before-and-after comparison stops mattering as soon as either of these is true:
- You’re so sure AI saves you time that working any other way feels dumb.
- You enjoy your work more with AI, so you’ll make whatever case you need to keep it.
Either way, the longer we use AI, the fewer people will agree to work without it, whether or not anyone can prove the time savings. Sooner or later, AI just becomes the new baseline. Nobody today says, “Before the internet, this report would have taken me two days!” At some point, the “before” stops being relevant.
Time saved only matters if project management is a division problem
A “before” nobody can recall is a real problem for PMOs asked to measure AI ROI in hours saved. Time you can’t remember spending is hard to measure, and harder still when people have a reason to report the number one way or the other.
But suppose you could measure it perfectly. What exactly would you be counting?
- Turning out a repeatable report in half the time it used to take?
- Sitting in more meetings, because you spend less time writing email or capturing notes?
- Running more projects at once, because the administrative load is lighter?
Savings like those aren’t worth much unless the only question on the table is “How many more projects can we do if we remove the administrative burden?” or “How many project managers can we let go and still run the same portfolio?”
Both questions assume project management is a concrete, finite set of tasks and routines, and that doing them in less time will raise throughput, lower headcount, or both.
That assumption turns project management into a division problem (divide the hours saved by the hours a project takes to run, and there’s your new capacity).
But seeing AI as “the same work, but faster” misses what it can really do for a PMO.
Speed is the least ambitious thing AI can give your PMO
Say a project manager uses AI to do everything on the list above. They spend less time on reports, no longer lose time cleaning up meeting notes, and triage and draft emails faster.
Now take away the pressure to bank those savings and do more of the same work faster. This PM gets to decide where their time does the most good. Here’s what a project manager in that position could do:
- Set up systems to monitor all three projects they’ve been assigned, where before at least one would have been neglected because it wasn’t the “squeaky wheel.”
- Send each key stakeholder tailored information, where before they’d only have had time for one generic status sent to everyone.
- Build substantive, red-teamed risk plans from AI analysis of large public and private data sets, where before that analysis would have taken weeks the steering committee wouldn’t approve.
- Turn around digestible decision briefs within 24 hours, where before analyzing the options and building executive-ready slides would have taken at least a week.
Look, AI may let a PM manage more, or let a PMO run leaner. Those are valid business cases, and I know some PMOs will take that route. But in my opinion, those PMOs are missing a core strategic competency: imagination. Whether anyone says so out loud or not, a PMO’s AI strategy is part of its company’s competitive strategy.
If you believe the point of AI is speed, you’ll soon be outpaced and outclassed by organizations that see its real power in delivering ambitious work. Those organizations hire and develop PMs who work from the second list, not the first one.
If You Only Do One Thing
Don’t wait for the question. Get ahead of “How much time have you saved?”
Over the next few months, share a steady stream of examples of what your team is doing with AI that they couldn’t do before, and put those examples in front of your steering committees. Keep the focus on the outcomes AI made possible (the risk caught early, the decision made in a day instead of a week) rather than the hours it saved on routine work.
And if someone still asks you for a number, have the courage to ask a question back:
“If AI gives each of our PMs back five hours a week, what do we want them doing with it?”