The promise was simple: let AI take the busywork, and you get your time back. Halfway through 2026, the data has turned that promise on its head. In February, Harvard Business Review published a study with a title that reads like a confession — “AI Doesn’t Reduce Work—It Intensifies It.” Following 200 employees at a U.S. tech company for eight months, researchers Aruna Ranganathan and Xingqi Maggie Ye found that 83% said AI increased their workload. Not the machines’ workload. Theirs.
AI work intensification is the 2026 pattern where AI raises output and expectations faster than it removes effort, so people end up doing more, not less. Study after study this year points the same direction: workers take on a broader scope of tasks, at a faster pace, spilling into more hours of the day. Lova is the chat-first AI project management product where AI agents work as first-class teammates on a shared board — claiming tasks, posting evidence, and moving cards through verifiable status. That surface matters here because, as the data shows, most of the new work AI creates isn’t the work itself. It’s the coordination around it — and coordination is exactly what a board is for.
Key takeaways
- A February 2026 Harvard Business Review study of 200 employees over eight months found 83% said AI increased their workload — they worked faster, took on more, and extended the day, often without being asked.
- A 2026 ResumeTemplates.com survey of 1,250 U.S. workers found 31% say AI raised their workload versus 16% who say it fell; of those with more work, 43% said it at least doubled, and 59% feel more pressure to perform.
- The Upwork Research Institute found 96% of executives expect AI to lift productivity, yet 77% of employees say it added to their workload — and 47% don’t know how to hit the gains they’re now expected to deliver.
- A workforce-analytics study by ActivTrak — 443 million hours across 1,111 employers — found that after AI adoption, time in chat and messaging rose 145%, email rose 104%, and no work category fell.
- The June 2026 Anthropic Economic Index found the single biggest use of its Cowork agent is business process and operations at 33.4% — reconciling spreadsheets, pulling scattered updates into one report — while software development is just 8.7%.
Does AI actually reduce your workload? The 2026 data says no
Start with the most rigorous look. HBR’s eight-month field study didn’t poll people about their feelings — it followed 200 workers through the adoption of AI tools and watched what happened. Productivity went up. Output increased. Tasks finished faster. And yet the work didn’t shrink; it expanded. Employees worked at a faster pace, took on a broader scope, and pushed tasks into more hours of the day. Eighty-three percent said their workload grew. The authors are blunt about where this leads if left unmanaged: workload creep, cognitive fatigue, and a quality slide as the early surge gives way to burnout.
It isn’t one study, and it isn’t one company. A 2026 ResumeTemplates.com survey of 1,250 U.S. workers found nearly twice as many say AI raised their workload (31%) as say it cut it (16%) — and among those carrying more, 43% said the load had at least doubled. Go back to 2024 and the Upwork Research Institute already had the through-line: 96% of executives expected AI to boost productivity while 77% of employees said it made their jobs harder. Three independent samples, three different methods, one direction. The question worth asking isn’t whether AI intensifies work. It’s why.
Why does AI make work harder instead of easier?
The first-order answer is expectations. When output per person jumps, the bar quietly resets to the new ceiling. ResumeTemplates found 60% of workers say leadership expects or requires them to use AI, and 59% say it has increased the pressure to perform at a higher level. Upwork put a sharper edge on it: 81% of executives admit they’ve raised workload demands in the past year, while nearly half of employees (47%) say they don’t actually know how to deliver the productivity their leaders now assume. AI didn’t give people slack — it raised the quota and handed them a tool they’re still learning to aim.
But “expectations rose” is where most takes stop. It explains why you feel more pressure. It doesn’t explain where the hours physically go. If AI genuinely does the doing faster, the time has to reappear somewhere — and the most interesting 2026 data is about exactly where it lands.
Where do the extra hours actually go? The coordination overhang
Here is the claim worth taking away, because it isn’t on the rest of the page: AI didn’t intensify the work so much as it intensified the coordination of the work. The proof is in the one 2026 dataset big enough to see it. ActivTrak’s State of the Workplace analyzed 443 million hours of work across 1,111 employers, comparing 10,584 people for 180 days before and after they adopted AI. The surprise wasn’t that activity rose — it’s which activity rose. Time in chat and messaging climbed 145%. Email climbed 104%. Business-management apps rose 94%. Not a single category went down. As Fortune summarized the findings, time spent emailing doubled while deep, focused work fell. AI became an extra layer on top of the job, not a substitute for any part of it.
Think about what that means mechanically. When AI triples one person’s output, everyone downstream now has three times as much to read, reconcile, sequence, and verify. That work is real, but it has nowhere structured to live — so it floods the channels people already have: chat, email, and the standing status meeting. Call it the coordination overhang: every unit of AI throughput creates a coordination liability, and unmanaged, that liability lands on a human. The 145% jump in messaging isn’t people slacking off. It’s the sound of an organization trying to hand-route work that’s suddenly moving faster than any human channel can track.
The agents themselves give the game away. The June 2026 Anthropic Economic Index classified 1.2 million sessions of its Cowork agent and found the largest single use isn’t coding at all — it’s business process and operations, at 33.4%: pulling scattered updates into one report, building checklists, reconciling spreadsheets. Software development was just 8.7%. In other words, the most common thing people delegate to an agent is the coordination work — the “work around the work,” a pattern we traced when AI agents took over the office work. AI is now doing the routing on both sides of the desk, and it still isn’t enough, because the state of the work has no shared home. It’s the same tax we’ve measured as the six hours a week lost to fragmentation and the hours workers now spend babysitting AI — different names for one leak.
How do you stop AI from intensifying work?
You can’t slow the throughput, and you wouldn’t want to — the faster output is the point. What you can change is where the coordination lives. Chat, email, and status meetings are the wrong substrate for it, because each one requires a human to relay state: you read an update, summarize it, and paste it to the next person who needs it. That relay is precisely the work that exploded 145%. Add AI throughput to a relay-based system and you don’t get leverage; you get a busier relay.
The fix is the same move that let human organizations grow past a single overwhelmed founder: stop being the wire. Put the work on a shared board where the current state is legible without anyone relaying it. Every task carries an explicit status. Claiming records who — human or agent — took it. “Done” is gated on attached evidence, not a message in a thread. Handoffs become recorded events instead of things you carry in your head. The coordination that AI throughput generates gets absorbed by the board instead of your inbox.
That’s what Lova is built to be. AI agents are first-class teammates on the board: they claim tasks, post their evidence, and move cards through verifiable status right alongside the people they work with. When an agent triples the output, the update doesn’t become three more messages someone has to chase — it becomes a card that moved, visible to everyone at once. More output stops meaning more chasing. The teams that escape the intensification trap in 2026 won’t be the ones with the fastest agents. They’ll be the ones whose agents don’t need a human to route what they ship.
Frequently asked questions
Does AI reduce workload?
On the current evidence, usually not. A February 2026 Harvard Business Review field study of 200 employees found 83% said AI increased their workload, and a 2026 ResumeTemplates.com survey found nearly twice as many workers report a heavier load as a lighter one. AI reliably raises output, but that output tends to reset expectations upward rather than free up time.
What is AI work intensification?
AI work intensification is the pattern, documented across multiple 2026 studies, in which AI raises output and performance expectations faster than it removes effort — so employees work at a faster pace, take on a broader scope, and end up doing more, not less. It is a management and coordination effect, not a limitation of the models themselves.
Why does AI increase messaging and email at work?
Because faster output creates more to coordinate. ActivTrak’s 2026 analysis of 443 million work hours found chat and messaging rose 145% and email rose 104% after AI adoption, with no category falling. When one person’s throughput jumps, everyone downstream has more to read, reconcile, and verify — and that coordination floods the channels people already use.
How can teams get time savings from AI instead of more work?
Move coordination off human relay channels and onto a shared board where the state of work is visible without anyone re-typing it. When agents and people claim tasks, post evidence, and advance status on one surface, the coordination that AI throughput generates lives on the board instead of in inboxes and meetings. A chat-first AI project management tool like Lova is designed to be exactly that surface.