ActivTrak's Productivity Lab pulled 443 million hours of work activity from 1,111 organizations and 163,638 employees and asked a simple question: what changes for the average worker once their company brings AI into the workflow. Not what leadership hopes changes. What the activity logs actually show.
The answer isn't the one most rollout plans are built on. App usage didn't get lighter after AI showed up, it got heavier, in some cases up 346 percent. Email climbed 104 percent. Chat climbed 145 percent. And the one thing you'd expect AI to protect, long stretches of uninterrupted focus, got shorter instead, down to an average of 13 minutes a session.
Most leaders assume AI adoption is a subtraction problem: fewer manual tasks, fewer hours, more room to think. The data says it's actually an addition problem. AI didn't remove work from the day. It added a new layer of coordination, review, and back and forth on top of the work that was already there.
Here's the root cause hiding in the numbers: AI made it faster to start something, but it didn't make it faster to finish thinking about it. Every AI-drafted email still needs a human read. Every AI-summarized chat thread still needs a human decision. So instead of removing steps, AI multiplied the number of small interactions a person has to touch in a day, which is exactly why email rose 104 percent and chat rose 145 percent while the actual workday only shrank 2 percent.
AI didn't hand your team fewer things to do. It handed them more things to check.
Call it the density trap: the workday doesn't get shorter, it gets denser, more app switching, more fragmented attention, backed by the study's own numbers, daily focused time down 23 minutes and the average focus session down to just 13 minutes.
The clearest read on where the extra load landed comes straight from the study's own comparisons:
The real tell isn't that usage went up, of course it did, that's the point of the tools. It's where it went up. Chat rose faster than email, which means the growth isn't in people getting more done independently, it's in more back and forth, more real-time interruption. And weekend work rising more than 40 percent while the actual workday barely moved is the clearest signal in the whole study: the extra work isn't happening on the clock, it's bleeding into time nobody scheduled for it.
The daily focus session, the stretch of time an employee can work on one thing without switching, dropped to 13 minutes. That's the real cost line most rollout plans never put in the budget.
Tier 1: Draw the Line Between AI Hours and Focus Hours
The Concept: The study's 13-minute average focus session isn't a discipline problem, it's a design problem. When AI tools keep chat and email open and answering, there's no natural stopping point built into the day anymore, so focus gets chipped away in small pieces instead of protected in blocks.
The Application: Put fixed, calendar-blocked focus windows on the team's schedule, not just as an individual habit but as a team norm, where chat and email notifications are expected to go quiet. Treat it the same way you'd treat a meeting nobody is allowed to book over.
Tier 2: Make Someone Own the Coordination Layer
The Concept: The 104 percent rise in email and 145 percent rise in chat didn't happen because people suddenly had more to say. It happened because AI created more outputs that needed a human to route, review, or reply to. Without someone accountable for that layer, it just piles onto whoever happens to be online.
The Application: Name one person per team (rotating is fine) responsible for triaging AI-generated outputs each day, drafts, summaries, flagged threads, before they hit everyone else's inbox. It's a small role, but it's the difference between AI output landing as noise or landing as decisions already made.
Tier 3: Redesign for Fewer Handoffs, Not Faster Ones
The Concept: Most AI tools today speed up one step in a longer chain, which is exactly why total activity rises even as the workday doesn't shrink. Speeding up a single step in a broken chain just moves the bottleneck, it doesn't remove it.
The Application: Before adding another AI tool to a workflow, map how many human handoffs the finished output still needs. This is where a tool like Cowork earns its place, not because it's faster at one step, but because it's built to carry a task through to a finished deliverable instead of handing you one more draft to route to someone else.
Measuring AI success by adoption rate, not by outcome. It feels like a win when usage numbers climb, more messages sent, more tools opened. But this study shows usage climbing is not the same as work getting lighter. The fix: track focused-work time and weekend work alongside adoption, not instead of it.
Assuming faster app usage means more output. A 145 percent jump in chat volume looks like productivity on a dashboard. It can just as easily mean more people are stuck coordinating instead of producing. The fix: ask teams directly whether AI tools reduced the number of people they had to loop in on a task, not just whether they used the tool.
Letting AI collapse the boundary between work hours and personal time. Weekend work climbing more than 40 percent while the workday barely shrank means the time AI was supposed to free up didn't disappear, it moved to Saturday. The fix: treat weekend activity data as a warning sign worth reviewing, not a quiet sign of dedication.
The uncomfortable finding in this data isn't that AI doesn't work. It's that AI works exactly as advertised, and the advertised version was never actually about giving people their time back. It was about doing more, faster, inside the same day.
That's the real management question sitting underneath every AI rollout: are you deploying these tools to reduce your team's load, or just to raise what the load looks like. The data says most companies, without meaning to, chose the second option.
AI adoption was never really a tools problem. It's a people and management problem wearing a tools costume. Micah curates the AI News Brief every day with exactly this kind of story, the ones that don't make it into the vendor press release. Subscribe at your.biggestgoal.ai.
ActivTrak's Productivity Lab pulled 443 million hours of work activity from 1,111 organizations and 163,638 employees and asked a simple question: what changes for the average worker once their company brings AI into the workflow. Not what leadership hopes changes. What the activity logs actually show.
The answer isn't the one most rollout plans are built on. App usage didn't get lighter after AI showed up, it got heavier, in some cases up 346 percent. Email climbed 104 percent. Chat climbed 145 percent. And the one thing you'd expect AI to protect, long stretches of uninterrupted focus, got shorter instead, down to an average of 13 minutes a session.
Most leaders assume AI adoption is a subtraction problem: fewer manual tasks, fewer hours, more room to think. The data says it's actually an addition problem. AI didn't remove work from the day. It added a new layer of coordination, review, and back and forth on top of the work that was already there.
Here's the root cause hiding in the numbers: AI made it faster to start something, but it didn't make it faster to finish thinking about it. Every AI-drafted email still needs a human read. Every AI-summarized chat thread still needs a human decision. So instead of removing steps, AI multiplied the number of small interactions a person has to touch in a day, which is exactly why email rose 104 percent and chat rose 145 percent while the actual workday only shrank 2 percent.
AI didn't hand your team fewer things to do. It handed them more things to check.
Call it the density trap: the workday doesn't get shorter, it gets denser, more app switching, more fragmented attention, backed by the study's own numbers, daily focused time down 23 minutes and the average focus session down to just 13 minutes.
The clearest read on where the extra load landed comes straight from the study's own comparisons:
The real tell isn't that usage went up, of course it did, that's the point of the tools. It's where it went up. Chat rose faster than email, which means the growth isn't in people getting more done independently, it's in more back and forth, more real-time interruption. And weekend work rising more than 40 percent while the actual workday barely moved is the clearest signal in the whole study: the extra work isn't happening on the clock, it's bleeding into time nobody scheduled for it.
The daily focus session, the stretch of time an employee can work on one thing without switching, dropped to 13 minutes. That's the real cost line most rollout plans never put in the budget.
Tier 1: Draw the Line Between AI Hours and Focus Hours
The Concept: The study's 13-minute average focus session isn't a discipline problem, it's a design problem. When AI tools keep chat and email open and answering, there's no natural stopping point built into the day anymore, so focus gets chipped away in small pieces instead of protected in blocks.
The Application: Put fixed, calendar-blocked focus windows on the team's schedule, not just as an individual habit but as a team norm, where chat and email notifications are expected to go quiet. Treat it the same way you'd treat a meeting nobody is allowed to book over.
Tier 2: Make Someone Own the Coordination Layer
The Concept: The 104 percent rise in email and 145 percent rise in chat didn't happen because people suddenly had more to say. It happened because AI created more outputs that needed a human to route, review, or reply to. Without someone accountable for that layer, it just piles onto whoever happens to be online.
The Application: Name one person per team (rotating is fine) responsible for triaging AI-generated outputs each day, drafts, summaries, flagged threads, before they hit everyone else's inbox. It's a small role, but it's the difference between AI output landing as noise or landing as decisions already made.
Tier 3: Redesign for Fewer Handoffs, Not Faster Ones
The Concept: Most AI tools today speed up one step in a longer chain, which is exactly why total activity rises even as the workday doesn't shrink. Speeding up a single step in a broken chain just moves the bottleneck, it doesn't remove it.
The Application: Before adding another AI tool to a workflow, map how many human handoffs the finished output still needs. This is where a tool like Cowork earns its place, not because it's faster at one step, but because it's built to carry a task through to a finished deliverable instead of handing you one more draft to route to someone else.
Measuring AI success by adoption rate, not by outcome. It feels like a win when usage numbers climb, more messages sent, more tools opened. But this study shows usage climbing is not the same as work getting lighter. The fix: track focused-work time and weekend work alongside adoption, not instead of it.
Assuming faster app usage means more output. A 145 percent jump in chat volume looks like productivity on a dashboard. It can just as easily mean more people are stuck coordinating instead of producing. The fix: ask teams directly whether AI tools reduced the number of people they had to loop in on a task, not just whether they used the tool.
Letting AI collapse the boundary between work hours and personal time. Weekend work climbing more than 40 percent while the workday barely shrank means the time AI was supposed to free up didn't disappear, it moved to Saturday. The fix: treat weekend activity data as a warning sign worth reviewing, not a quiet sign of dedication.
The uncomfortable finding in this data isn't that AI doesn't work. It's that AI works exactly as advertised, and the advertised version was never actually about giving people their time back. It was about doing more, faster, inside the same day.
That's the real management question sitting underneath every AI rollout: are you deploying these tools to reduce your team's load, or just to raise what the load looks like. The data says most companies, without meaning to, chose the second option.
AI adoption was never really a tools problem. It's a people and management problem wearing a tools costume. Micah curates the AI News Brief every day with exactly this kind of story, the ones that don't make it into the vendor press release. Subscribe at your.biggestgoal.ai.