What the Q1 Productivity Numbers Don't Tell You About Where Your Team's Time Actually Went

U.S. nonfarm business productivity grew 0.3% last quarter. Unit labor costs rose 1.8%. Real, inflation-adjusted wages grew a grand total of 0.1% for the year ending in March. If you spent the last twelve months rolling out AI tools across your org and expected those numbers to finally move, they didn't.

I added three AI tools to my own workflow this year — one that transcribes and summarizes meetings, one that drafts status updates from raw notes, one that handles the calendar Tetris I used to do by hand. Each one genuinely saves me time on the task it was built for. And my Thursday afternoon block — the ninety minutes I've kept since January for "catching up on statuses" — is still exactly ninety minutes long. The tools didn't shrink it. They just changed what fills it.

The direct answer: U.S. productivity rose just 0.3% in Q1 2026, and real wages grew only 0.1% over the year, despite widespread AI adoption across knowledge work. The likely explanation isn't that AI failed to save time — it's that the hours it freed up got reabsorbed by the same coordination structure that existed before: more status-chasing, more reconciling numbers across tools, more verifying that an AI's output was actually right. The saved time didn't disappear from anyone's calendar. It just moved to a different line item.

The Numbers That Don't Add Up

Here's what makes the Q1 data genuinely strange. This wasn't a quarter where AI adoption stalled — by every internal measure I've seen at portfolio companies and peer orgs, tool adoption climbed all year. Drafting tools, meeting summarizers, auto-generated reports — all of it got rolled out faster in the last twelve months than in the three years before combined.

And yet output per hour worked barely moved, and take-home pay, once you account for inflation, moved even less. A McKinsey Global Institute analysis this year found that knowledge workers still spend close to half their week on email and hunting down information or the right person to unblock a task — the exact category of work every one of these AI tools was supposed to shrink. If the tools are working at the task level and the aggregate number won't budge, the problem isn't the tool. It's what's wrapped around it.

Where the Saved Hours Actually Went

I watched this happen in real time with our own weekly ops report. The drafting step — pulling last week's numbers into a narrative — used to take about 45 minutes. With a drafting assistant, it now takes closer to 10. That's a real, measurable win. But the report itself still takes almost exactly as long to get out the door as it did a year ago, because the other three steps never went anywhere: someone still checks the AI's numbers against the source system, someone still chases the one team that didn't update their tracker, and someone still reconciles a figure that doesn't match between two tools that were both supposedly "up to date."

Task-level AI makes one step of a four-step process faster. If the other three steps are untouched, the total time barely changes — it just gets redistributed. The report goes out at the same hour it always did. The 35 minutes that vanished from drafting reappear somewhere else on the calendar, usually somewhere less visible.

The Draft-Verify-Chase-Reconcile Loop: Three Places Coordination Overhead Hides in Plain Sight

Once you name the pattern, you start seeing it everywhere: draft, verify, chase, reconcile. AI has only ever touched the first step. The other three are where the coordination tax actually lives, and they show up in three specific places.

The verification tax. Every AI-generated number, summary, or status update now needs a human to confirm it's actually right before anyone acts on it — and that confirmation step rarely existed as a named task before, because a human wrote the first draft and implicitly trusted their own math. Now someone has to open the source system, cross-check the figure, and only then sign off. I've watched a "5-minute AI summary" turn into a 20-minute review cycle once you count the verification pass honestly.

The status-chasing AI can't remove. A drafting tool can turn your notes into a polished update. It cannot make the VP of Sales actually send you the deal-count number you asked for on Tuesday. Most of the coordination tax was never about writing the update — it was about extracting the input from a person who has twelve other things due the same day. If your AI adoption strategy only touched the writing step, the chasing step is exactly as painful as it was last January.

Reconciliation across tools that don't agree. The more systems a company runs, the more often two of them disagree about the same number — a deal stage in the CRM that doesn't match the forecast deck, a headcount figure that's stale in one dashboard and current in another. AI tools that summarize one system in isolation don't fix this; they just produce a faster, more confident-sounding summary of a number that might already be wrong. This is exactly the blind spot Rhythms' Radar was built to close — it flags a data conflict or a stalled initiative on day three instead of day thirty, so nobody discovers the discrepancy for the first time in the meeting where it actually matters.

How to Count It on Your Own Team

Most leaders experience coordination overhead as background noise — present, mildly annoying, never quantified. The fix isn't a new dashboard. It's two weeks of honest counting. Track every status-related message, deck, or update request that moves through your team without directly producing a decision or a deliverable. Not every Slack message — just the ones whose entire purpose is "where do things stand."

When I ran this for my own team, the number surprised people who'd have told you, confidently, that things were running fine. It usually does. The counting exercise works precisely because it turns something ambient into something with a number attached — and a number is the only thing that gets a line item in next quarter's planning.

You're not double-checking a human's transcription of the data. You're looking at the data. That's the difference a review built on live, source-grounded numbers makes — if you're already running some version of a weekly or monthly operating review, that's where Reviews removes the verification tax at its root instead of just making the draft faster.

What Changes When the Structure Changes, Not Just the Task

The uncomfortable read on the Q1 numbers is that most organizations bought AI tools the way you'd buy a faster typewriter. It genuinely makes one step quicker. It does nothing to the three other steps in the draft-verify-chase-reconcile loop, and those three steps were always where the real time went. Our own Playbooks work exists because of exactly this pattern — recurring cadences like check-ins and planning cycles that pull their inputs automatically from the tools where work already happens, so the coordination doesn't get rebuilt by hand every single week regardless of how fast any one draft gets written.

I don't think the Q1 report means AI isn't working. I think it means most of us bolted a faster engine onto the same chassis and were surprised the car didn't go faster. The structure around the task — who chases whom, who verifies what, which tool is the source of truth — is still the thing actually setting your team's speed limit. Worth checking, before the Q2 numbers come out and everyone acts surprised again.

Try it free at rhythms.ai.

Frequently Asked Questions

Why hasn't AI adoption shown up in productivity statistics yet?

Because most AI adoption so far has been layered onto the existing coordination structure rather than replacing it. A tool saves time drafting a summary, but the same status-chasing, verification, and reconciliation work still happens around that draft — so the saved hours get reabsorbed into those steps instead of disappearing from anyone's week.

What is "work about work" and how big a problem is it, really?

It's the coordination labor that surrounds actual decision-making: status updates, deck prep, chasing responses, reconciling numbers across systems. A McKinsey Global Institute analysis this year put knowledge workers at close to half their week on email and information-hunting alone — a figure that dwarfs almost any productivity gain a single AI tool can deliver on its own.

How do I find out where my team's coordination overhead actually lives?

Track for two weeks how many status-related messages, decks, or update requests move through your team without directly producing a decision or a deliverable. Most leaders are surprised by the volume once they count it explicitly instead of experiencing it as background noise.

Does adding more AI tools reduce coordination overhead on its own?

Not by itself. A tool that drafts a status update still requires someone to gather the inputs, verify the output against the source system, and reconcile it against whatever the other tools say. The coordination structure around the task stays untouched even when the task itself gets faster.