I asked a friend last month what he actually did all day. He runs a 40-person product team at a SaaS company. He thought about it for longer than the question deserved.
“I move things,” he said. “I move things between people.”
He laughed when he said it. Neither of us thought it was funny.
This is the version of professional life almost nobody puts on a LinkedIn profile, but it is the version time-tracking data keeps surfacing. Most of what we call work is not the work. It is the layer of activity that exists because the actual work has to pass through human hands. And that layer — the meetings, the status updates, the handoffs, the approvals — is on a different trajectory than the rest.
The org chart was a solution to a coordination problem. The coordination problem is disappearing.
Where the 60% Goes
Asana ran a study across more than 10,000 knowledge workers and found that 60% of the average workday is spent on “work about work” — chasing updates, sitting in meetings that should have been emails, switching between tools to reconcile what is supposed to be a single conversation [1]. That leaves 27% for skilled work and 13% for strategic planning. On a 40-hour week, that is roughly 11 hours doing the job you were hired for and 24 making sure the rest of the organization knows what you produced.
Microsoft’s Work Trend Index, built from a different dataset — anonymized telemetry from millions of Microsoft 365 users plus 31,000 survey responses across 31 markets — landed in the same place. The average employee spends 57% of their time communicating in meetings, email, and chat, and 43% creating things. Workers are interrupted every two minutes during core hours: 275 times a day [2]. Two independent studies, two methodologies, two anchor numbers that almost touch.
Asana’s annual breakdown makes the time cost concrete. The average knowledge worker spends 103 hours a year in unnecessary meetings, 209 hours on duplicative work, and 352 hours talking about work [4]. That is 664 hours — sixteen full work weeks, every year, spent not doing the job.
When I show people these numbers, the response is rarely surprise. It is recognition.
Why the Coordination Existed in the First Place
The numbers read like a productivity scandal — proof that modern offices are broken. But the coordination layer is not an accident. It was the design.
In 1937, Ronald Coase wrote a short paper asking why firms exist at all. If markets were efficient, every transaction would happen between independent contractors. They do not, he argued, because coordinating through markets is expensive. Search costs, negotiation costs, enforcement costs — every external transaction carries an overhead. Firms exist because doing things internally is cheaper than doing them externally [5].
The org chart is what that internal coordination looks like. Managers route information. Project managers schedule handoffs. Approval chains substitute for contract enforcement. The middle of every organization is a Coasean machine.
A 2025 paper by economists at NBER took Coase’s framework into the age of agents and called the result the “Coasean Singularity” — the point at which AI agents collapse all three transaction cost categories toward zero at once. Search costs vanish when an agent scans millions of options in seconds. Negotiation costs vanish when terms optimize in real time. Enforcement costs vanish when contracts execute automatically [8]. If that sounds abstract, watch what happens when an agent goes from a customer’s question to a refund to a ledger entry without anyone routing it.
The headlines focus on automation — AI doing tasks faster. Coase’s framework points at something stranger. When the cost of coordinating drops toward zero, the structures we built to manage coordination lose their economic reason to exist. The org chart is not getting automated. It is getting outcompeted by its own absence.
It Is Already Happening
That is theory. Here is a real org chart that disappeared.
Klarna will do. The Swedish fintech ran with 5,527 employees through the end of 2022. By late 2023 it was at 3,422 — a 40% reduction, driven by a hiring freeze, attrition, and the deployment of AI agents across customer service and internal operations. The headline framing was “AI replaces workers.” Look at which workers, though. The roles that survived were the irreducibly judgment-heavy ones: engineers, product leads, compliance specialists. The roles that dissolved were the escalation managers, team leads, and service handlers — the layer that existed to route customer questions between humans who could answer them [7]. Klarna did not automate coordination. It collapsed the bottleneck that made coordination necessary.
Shopify ran the same experiment as policy. Tobi Lutke issued a memo: before any team hires a human, it has to “prove why certain jobs can’t be done using AI” [7]. The brilliance of the policy is that it forces every coordination role to justify itself. A status-tracking role cannot survive that question; an information-routing role cannot either. The roles that pass are the ones where human judgment is doing real work — strategy, design, irreplaceable taste.
Fortune reported a quieter case that lands the same point at team scale. An AI-first healthcare company moved a 10-person engineering team down to 3, according to the company’s own restructuring report [10]. The 7 who left were not the worst engineers. They were the ones whose actual job — the day-to-day reality of how they spent their hours — was coordinating between the other engineers.
The team lead spent most of her week in status meetings and stakeholder updates. The project manager was a coordination node by definition. Even the QA leads spent more time orchestrating coverage than running tests. When AI agents handled code generation, test orchestration, and deployment routing end-to-end, those seven roles did not lose to a smarter robot. They lost their reason to exist.
Gartner forecasts that 20% of organizations will use AI to flatten their structure by 2026, eliminating more than half of middle management positions in those firms. Middle managers were already 29% of US layoffs in 2024 [6]. The coordination layer goes first because the coordination problem goes first.
But Is Coordination the Actual Work?
The pushback that lands hardest is not “this isn’t happening.” It is “you are misreading what is happening.” It comes in three versions.
The first: coordination is not waste. It is how teams build shared understanding and trust. Delete it and you do not get a leaner team — you get a fragmented one. There is real research behind this. Google’s Project Aristotle and Amy Edmondson’s work on psychological safety both show that what looks like overhead on a calendar can be the connective tissue of a working team. I take that seriously.
But I would split the category in two. Sense-making coordination — negotiating ambiguity, aligning on intent, building trust — is a different animal from relay coordination — status updates, handoff routing, approval chains. The 60% figure is dominated by relay; Asana’s largest buckets are “unnecessary meetings” and “talking about work.” Compress the relay layer and sense-making does not just survive. It gets oxygen.
The second version is sharper: the 57-60% figure conflates real work with overhead. When a product manager spends three hours in a stakeholder conversation, that IS the work. Calling it coordination is a category error.
Partially true. The corrective comes from Rob Cross and colleagues at HBR, who studied 300+ organizations and found that 3 to 5 percent of employees generate one-third of all value-added collaboration [3]. The rest of the collaborative time is the floor of the building, not the activity inside it. The Microsoft data adds the texture — half of meetings unplanned and ad hoc, workers checking email before 6 a.m. Not the profile of a workforce engaged in deep collaborative work. The profile of one drowning in coordination debt.
The third pushback leaves me uncertain. AI will not eliminate coordination — it will replace one form with another. We swap status meetings for agent audits, handoff routing for tool conflict resolution, and end up coordinating between AI systems instead of between humans. There is evidence for this: a 2025 California Management Review piece documents organizations where uncoordinated departmental AI deployments produced “duplicated effort, conflicting processes, and a general decrease in coordination” [9].
I concede the entropy risk as real. Where I push back is the analogy. Email was supposed to reduce meetings; it created more. Slack was supposed to reduce email; it created always-on. The assumption is that AI follows the same arc. But email and Slack made coordination cheaper, and cheaper coordination produced more of it. AI agents are doing something different: they remove the execution gap that created the need for coordination in the first place. Cheaper coordination begets more coordination. Eliminated need begets less.
How to Audit Your Own Coordination
A week spent examining how your time is used can help you assess which parts of your role are exposed.
Step one: time inventory. For five working days, tag every 30-minute block as one of four things: relay coordination (status, routing, approvals), sense-making coordination (alignment, ambiguity, trust), skilled execution (work that requires your judgment), or strategic direction (deciding what to build). Use calendar data and tool telemetry as a cross-check — self-perception runs about 30% optimistic.
Step two: dependency map. Take your ten biggest relay activities. For each, trace the information chain from origin to action. Most chains run 2-4 human relay nodes deep. The project manager updates the director who summarizes for the VP who briefs the executive. One meeting on your calendar, four nodes in the system.
Step three: AI collapse test. Ask of each chain: if an agent could go from origin to action — generating the report, making the decision, executing the task — what survives? Score three: fully collapsible (status aggregation, routine reporting). Score two: partly collapsible (code review where AI handles correctness but architecture needs you). Score one: not collapsible (negotiating priorities between stakeholders with real disagreements).
Step four: role exposure. Aggregate the scores per person. Roles where more than 60% of time scores a three are coordination-dominant — those change shape when the execution layer changes. Roles where more than 60% scores a one are judgment-dominant — those expand.
Step five: structural redesign. Take the org chart. Remove every box and every line whose function is score-three coordination. What is left is what your organization will look like by the end of the decade.
This is not a layoff exercise. It is a structural one. The point is to see your work for what it is before the market does.
What Remains When the Layer Is Gone
The possibility is that compressing the relay layer leaves more time for judgment.
If 60% of your week is relay coordination, then the 27% you do between meetings — the thinking, the building, the deciding — was already your real job. You just had no time for it. When the relay layer compresses, what expands is the work you wanted to do in the first place.
There is one objection I want to concede out loud. Managers are not just coordination nodes. They mentor. They advocate. They develop the next generation. They shield their teams from political weather. All of that is real, and it is the part of management that does not get automated.
But the mentorship is already being starved. Emergenetics found that only 37% of new supervisors get any training upon promotion, and 74% report no ongoing training afterward [11]. The development pipeline is not failing because managers do not care. It is failing because 60% of their week is consumed by relay work that crowds out everything else. Remove the relay, and the human work managers were supposed to be doing gets to happen. The org chart does not vanish. It gets rewritten around judgment instead of routing.
My friend, the one who said he moves things between people — he is one of the sharpest product leaders I know. The taste, the strategic instinct, the ability to read a market and an org at the same time. That is the work. The moving was the bottleneck. And the bottleneck is getting cheaper to dissolve every quarter.
The uncomfortable part of “your job was never the job” is recognizing the gap between what you were hired for and what your calendar shows. The freeing part is realizing the calendar is what changes.
References
- [1] Asana. Anatomy of Work Index: How Work About Work Gets in the Way of Real Work. Molly Talbert, 2025.
- [2] Microsoft WorkLab. Breaking Down the Infinite Workday — Microsoft Work Trend Index Special Report. 2025.
- [3] Cross, R., Rebele, R., & Grant, A. Collaborative Overload. Harvard Business Review, January-February 2016.
- [4] Asana, Anatomy of Work Index, annual time breakdown section.
- [5] Shahidi, P., Rusak, G., Manning, B., Fradkin, A., & Horton, J. The Coasean Singularity? Demand, Supply, and Market Design with AI Agents. NBER / University of Chicago Press, 2025.
- [6] Fortune. AI Is Already Changing the Corporate Org Chart. 2025.
- [7] The AI Economy. Balancing Act: Klarna, AI, and the Future of Work. 2025; Fortune, 2025.
- [8] Shahidi et al., The Coasean Singularity, NBER, 2025; Warin, T. From Coase to AI Agents. California Management Review, 2025.
- [9] Warin, T. From Coase to AI Agents. California Management Review, 2025.
- [10] Fortune. AI Is Already Changing the Corporate Org Chart. 2025.
- [11] Emergenetics. The Middle Management Squeeze 2026: Why Supervisors Are Often Overlooked and How to Empower Them. 2026.