A writer sent home this month describes his new office. His desk is a kitchen counter, and cleaning it makes for good procrastination. His cafeteria is an emergency-stocked fridge, and raiding it makes for better. In two sentences the whole building folds down to surfaces. The desk, the lunchroom, the hallway, the meeting room: all of it is now one wiped slab of counter with a laptop on it.
He has company. By the middle of March, hundreds of thousands, maybe millions, have done the same. Amazon, Apple, Google, Twitter, and Airbnb have told at least some of their people to stay away. A survey of eight hundred human-resources executives, taken on a single day this month, finds that 88 percent of organizations have encouraged or required staff to work from home. Microsoft alone has asked about fifty thousand people near Seattle not to commute, and reports that its teams are still connecting, still collaborating, still getting the work done.
Here is the strange part. It is working. Or it looks like it is.
Look at what the tools are doing. In one week the daily user count on Microsoft Teams jumps from thirty-two million to forty-four million, and in a single day that week those users run more than nine hundred million meeting and calling minutes. A university in Bologna moves ninety percent of its courses online in four days. One network company measures video-conferencing traffic up about two hundred percent during working hours. A shared-workspace company sees demand up three hundred percent. Business VPN use rises a hundred and sixty-five percent worldwide. The machine that carries meetings, files, calls, and status updates did more than survive the move to the kitchen counter. It scaled.
So a level-headed executive, watching the dashboard in late March, is about to reach a reasonable conclusion. The meetings still happen. The documents still move. The work still ships. Maybe the office was mostly a very expensive room we rented to do things we could have done from home all along.
I think that conclusion is wrong, and I think it is wrong in a way that will take years to show up. But you cannot just assert that against a screen full of green numbers. You have to go back to a plainer question, the kind you ask when a measurement surprises you. What was the office actually making?
The two things the building produced
Start with what an office is for, from the ground up.
The obvious answer is tasks. People go to a building, and out of the building come finished things: a deck, a lease, a line of code, a closed sale. That is the part you can count. It is the part every productivity tool is built to move, and it is the part that just proved it can move from anywhere with a good connection.
But a building that has been running for years is quietly making a second thing, and it never appears on an invoice. It is making people. A junior hire walks in on a Monday not knowing what good looks like, and eighteen months later she knows. She can tell when a draft is not ready. She knows which client concern to take seriously and which to let pass. She knows when to speak in a meeting and when the real decision is being made in the four minutes before it starts. Nobody scheduled that. It happened while she was doing something else.
Two outputs, then. Tasks, which are visible and countable, and competence, which is slow and invisible. When the office emptied, the first output kept flowing and the second one got harder to see, harder to trigger, and easier to neglect. The trouble is that only one of them shows up in a week-one readout, and it is the wrong one.
Before we trust that, we should rule out the boring explanations, because a claim this large should have to survive them.
Ruling out the easy answers
The first easy answer: nothing important was lost, the office really was just the task container, and the container turned out to be optional.
Against that, listen to the people who study this for a living, not the readout. A management professor watching the shift calls it a research goldmine and then, in the same breath, warns that reading it straight will produce bad conclusions, because this is not a typical remote-work situation. Earlier studies looked at a handful of companies doing work that needed little supervision. This touches every industry at once, with no training, no policies, no chosen workspace, no childcare, none of the systems a real remote company builds over years. The green numbers are not measuring remote work. They are measuring an emergency that happens to run on the same software.
The second easy answer: fine, something was lost, but it was just chatter. Watercooler talk. Pleasant, not load-bearing.
This one is more tempting, so look at where the loss actually lands. When a business-school professor who has spent two decades studying dispersed teams writes up her guidance this month, she does not aim it at the whole company evenly. She points at a specific group. Newer employees, she writes, and people on critical projects, will require extra one-on-ones. Start your video meetings, she says, with the newest or lowest-status person, the one who usually speaks least. Spend the first six or seven minutes on the conversation that used to happen by itself, because people will not find replacements on their own. You have to coach them. The room used to supply, for free, something that now has to be scheduled, and the person who needed it most was the beginner.
The third easy answer: the tools will simply reproduce it, give them a month.
Some of it, yes. Watch which parts move cleanly. A meeting time, an agenda, a task owner, a file location, a written decision, a status update: all of that is explicit, and explicit things travel well through Zoom and Slack and a shared drive. That is exactly why the readout looks healthy.
Now watch which parts do not move. And to see why, you have to look at how a beginner actually learns, because it is not what most people think.
How competence actually moves
We imagine training as a transfer: the expert knows a thing, writes it down or says it, the novice receives it. Some knowledge does work that way. Most of the knowledge that matters does not.
Two learning researchers spent years watching apprentices, and set out, in their words, to rescue the idea of apprenticeship. What they found is that a newcomer does not start by being taught. She starts at the edge, doing small real parts of the work, watching the practice happen around her, and slowly moving inward until she is fully part of it. The learning is not a download. It is a position. You have to be in the room, near the work, at the edge of it, for a long time.
A management theorist gave the reason a name almost thirty years ago. Most of what an expert knows is tacit: personal, tied to context, hard to put into words. It moves through shared experience before anyone can write it down, if it can ever be written down at all. The file transfers. The tacit part does not. It has to be lived next to.
So here is the beginner's real curriculum, the one nobody prints. She learns taste when a senior person rewrites one of her sentences, kills a slide, and says a draft is not ready, and she absorbs the rule that was never stated. She learns trust by watching who checks the numbers, who bluffs, who knows when to stop talking. She learns timing by seeing when someone interrupts, when someone waits, and when someone walks a fight out of the room. She learns judgment from the gap between the official agenda and the real conversation that surrounds it.
None of that was ever a lesson. It was a glance at her screen, a sentence at her desk, a quick "not that way" before the mistake reached anyone important. In the office it was almost free, because it was ambient. The correction cost the senior person four seconds and cost the firm nothing anybody tracked. The building absorbed the bill.
The file transfers. The tacit part does not. It has to be lived next to.
Take the room away and every one of those four-second corrections has to become something. A message. A scheduled call. A comment thread. A meeting that goes on the calendar. And most of them, being small, simply do not get made. Nobody books a fifteen-minute video call to say "that heading is wrong." So the beginner keeps the wrong heading, and never learns why.
There is an asymmetry hiding in this, and it is the center of the whole thing. The senior worker already holds the vocabulary, the relationships, the taste, the judgment. She built all of it over years of exactly this ambient contact, and she can carry it to a kitchen counter and keep producing for a long time. The dashboard will love her. The beginner has almost none of it stored yet, and the counter cannot give it to him. So when access to the room goes from constant to scheduled, the cost does not fall on everyone equally. It falls first, and hardest, on the people with the least.
The one experiment that saw it
We are not entirely guessing. There is one clean study, from before any of this, and it is worth sitting with because it saw both outputs at once.
A Chinese travel agency with sixteen thousand employees let call-center staff volunteer to work from home, then randomly assigned them to home or office for nine months. The home workers got more productive: about thirteen percent, from taking fewer breaks and handling more calls in a quieter room. They were happier. They quit at half the rate. Every number a manager watches went the right way.
And their promotion rate fell, holding performance equal. The same people, doing measurably better work, advanced more slowly, because they were no longer in the room where advancement is learned and awarded. That is the finding to carry out of this essay. You can watch the work go up on a dashboard while the person's future goes down, and the two readings sit right next to each other, both true, one of them invisible unless you already know to look.
The strongest case against me
I have to give the other side its full weight, because it is strong and it is not going away.
The best version goes like this: office training was never good. It was just cheap and unexamined. A serious remote company can train a beginner better than osmosis ever did, because it is forced to make the tacit explicit. One software company has been fully remote since its first day, twelve hundred people across sixty-seven countries, and this week it published a handbook for everyone suddenly facing the same shift. Its first rule is document everything: build a single source of truth so no one has to walk to a cubicle to ask, keep onboarding in one place, record the meetings, write down the values. Where the office left timing to guesswork, good remote norms make it a rule: which channel for which message, when a reply is owed, what counts as an emergency. The office taught by accident and lost the lesson if you were out sick that day. A real remote system teaches on purpose.
And the office was often a worse teacher than we remember. When two large companies knocked down their cubicle walls to force people together, face-to-face interaction did not rise. It fell by about seventy percent. People put on headphones, stared at screens, and emailed the colleague fifteen feet away rather than have a hard conversation in front of the room. Proximity did not guarantee learning; sometimes it just guaranteed noise. The office also decided who got to learn at all. It rewarded the people who could afford the expensive downtown, tolerate the long commute, and fit the after-work social pattern, and it quietly filtered out everyone else. An apprenticeship you can only attend if you can afford to live near it is not a fair apprenticeship.
All of that is true. I concede every word of it. And it does not save the executive at the dashboard, for one reason.
The line
The case for deliberate remote training is a case about companies that built the thing. Most of the companies going home this month have not built it.
They have Zoom licenses and no handbook. Meeting links and no onboarding templates. A chat channel and no rule for who speaks first. They have, in other words, exactly the half of the office that carries tasks, and none of the half that carries people, because the second half was never a system they could see. It was a byproduct of the building. It came free with the rent, so no one ever wrote it down, and you cannot port a thing you never knew you had.
Task work survives the move home. The numbers climb. And a leader reads the climbing numbers as proof that the office was waste. The training cost did not vanish. It stopped being visible, which is a different thing, and the whole danger is in mistaking the second for the first. The building was never mainly hiding a cost. It was paying one, silently, on every hire it ever made.
What the counter can't teach
The advice going out to managers this week describes a new hire's first day as a shipping problem. Mail the laptop. Send the links to email and chat. If the paperwork is not digital, get it to her later. Put a welcome package in the post, a notebook and some pens and a company mug, where the office doorway used to be.
She will open the box at her own kitchen counter. She will log into the tools, join the calls, and do the tasks, and the tasks will get done. On the dashboard she will look completely fine.
What she will not do is lean toward the next desk at the moment a senior person kills a bad idea in a single sentence, and learn, without anyone deciding to teach her, what bad looks like. No one will schedule that call, because no one was ever billed for it. Six months from now she will be good at the work and not quite sure why the room ever mattered. Somewhere above her, a number will say the office was never necessary. The number will be counting the wrong thing, and it will be counting it very well.