On 11 March, the workplace analytics firm ActivTrak published the results of a study drawn from measured activity rather than a survey. Its Productivity Lab examined 443 million hours of work across 1,111 organisations and 163,638 employees, and compared how people worked before and after they adopted AI tools.
The finding that has drawn the headlines is that AI did not lighten the load. Across every category of work the study measured, time spent rose after adoption, by somewhere between 27 and 346 per cent.
Email rose 104 per cent. Chat and messaging rose 145 per cent. Not one category fell.
The tool that was meant to take work off the plate added to it. And it did so in the one part of the working day that is hardest to see: the coordination, the messaging, the switching between things.
What the number is really describing
The obvious reading is that AI has made everyone busier. The more useful reading is about what the study reveals when a genuine increase in capacity arrives inside a structure that has no way of absorbing it.
New capability entered the working day. Nothing left it. The same study found that focus efficiency fell to 60 per cent, a three year low, while collaboration rose 34 per cent and multitasking rose 12 per cent.
More was being produced, in more fragments, with less uninterrupted time to produce it in.
That is the shape of overload, and it is worth being precise about what overload is. It is not simply a large volume of work. It is a large volume of work with no agreed way to decide what takes priority, and no mechanism for deciding what comes off the plate when something new lands on it.
The explanation the market offers
A professional living inside that condition is handed a familiar diagnosis.
Get better at prioritising. Learn to manage your time. Adopt the system, the method, the app that promises to bring order to the day.
Every one of those answers locates the problem in the person. It assumes that somewhere in the flow of arriving work there is a correct order that the individual has simply failed to find. The misattribution is quiet and it is close to universal: the person concludes that they are disorganised, when what is actually missing is a decision that was never theirs to make.
An individual can rank their own tasks. They cannot rank the competing claims of three different people who are each assigning work without seeing what the others have already sent.
The conditions that produce it
The structural circumstances that generate overload are describable, and none of them is a fact about the person carrying it.
More work arriving than can be completed, with no agreed order of priority that everyone assigning it actually references. Multiple people assigning work without visibility of what is already committed. Work routed to whoever is most likely to absorb it, rather than to whoever has room.
A role that keeps expanding while nothing is ever taken off it. A departure or a vacancy quietly absorbed by the people who remain, rather than backfilled or descoped. No accepted way to decline or defer work without it counting against the person who does.
And beneath all of these, a layer of work that appears in no plan. The coordination, the rework, the covering for an absence, the messaging that the ActivTrak numbers just measured rising. It consumes real hours and it is counted nowhere, so the person doing it looks slower than the person who never has to.
Read the list again and notice what is absent from it. There is no sentence in it about the individual. Each condition is a property of how the work was arranged, and each was arranged by someone.
What tends to relieve it, and where it has to happen
The conditions associated with relief are structural, not personal, and this is the part the productivity literature tends to skip.
What eases overload is not a better personal system but an agreed priority order that everyone who assigns work is required to reference before adding to the plate. A visible account of where work has already concentrated, set against what each person is already holding. An accepted route for deferring or declining when capacity is full, one that carries no penalty for using it.
Making the invisible work visible belongs in the same category. Coordination, rework and cover counted as load rather than assumed to be free. A shared definition of what finished looks like, so that effort is bounded rather than expanding to fill whatever is available.
None of these is something the overloaded person can install alone. Each of them describes a change to the environment, and each depends on someone with the authority to change it.
What a leader sees, and what nobody measures
This is where the condition looks entirely different from the two ends of it, and the difference is the whole point.
The professional sees one instance. Their own week, their own inbox, their own sense of never reaching the bottom of it. What they cannot see is the distribution: whether the same pressure sits on everyone, or whether work has quietly gathered on the two or three people who never say no.
A leader can see that distribution, in principle. They are the only party who can ask whether task volume across the team is proportionate to capacity, whether a priority framework exists that everyone assigning work refers to, whether the coordination and cover that consume real time are counted when the next thing is handed out. They are the only one who can descope or backfill an absorbed gap rather than let it settle silently on the people who remain.
The trouble is that almost nothing an organisation measures points there. Engagement surveys record how people feel about their workload. Output metrics record what they produced.
Neither records whether the work was distributed in a way that any one person could reasonably carry. If the load itself is never counted, why would the person carrying it be the one expected to fix it?
So the condition compounds in a place the individual cannot see and the organisation is not looking, until someone capable quietly reaches the end of what they can hold.
Opinion: The Load Is Real. The Explanation Is Wrong.
Our position is that overload is a property of how work is arranged and distributed, not a measure of how well an individual copes. When more arrives than can be done, the useful question is not whether the person is prioritising correctly. It is who is permitted to decide what comes off the plate, and whether anyone is counting what is already on it.
The ActivTrak study is a useful marker precisely because it is not a survey of feelings. It is 443 million hours of observed behaviour, showing that a real increase in capability was absorbed by the structure and turned into more load rather than less.
The capacity to do the work went up. The capacity to decide what work mattered did not move.
That matters more now than it once did, and for a specific reason. As routine production is automated, the work that still belongs to people is judgement: the ability to hold a complex problem in mind long enough to see what is true about it. That runs on cognitive capacity, and cognitive capacity is the first thing overload consumes.
An organisation that lets work accumulate without any mechanism for deciding what to shed is not merely tiring its people. It is spending down the one resource it has that a machine cannot supply.
The provocation we would leave open is this. If a working environment can reliably bury capable people in more than they can do, and then offer them a course in personal prioritisation as the remedy, the question is not whether professionals should organise their time more cleverly. It is why the decision about what work matters is still treated as the private problem of the person who was never given the authority to make it.
ActivTrak Productivity Lab, "ActivTrak Productivity Lab Analyzes 443M Hours of Work Activity, Revealing AI Is Accelerating Work, Not Replacing It," 11 March 2026
ActivTrak, "2026 State of the Workplace," 11 March 2026
The contents of this article are for informational purposes only and do not constitute professional, legal, or financial advice.
.jpg)
