The form came back

Two weeks into the first lockdown I wrote a post from a kitchen table about working from home, and it ended like this:

Whenever it ends, I’d like the following to be a settled question rather than a favor: that a person who does better work from home some or most of the time should be able to, and should not have to justify it with a medical letter or a child or a boiler. We have now run the experiment across an entire industry simultaneously. Whatever the results are, “it can’t be done” is off the table, and I’d like somebody to write that down before everyone forgets and the form comes back.

Nobody wrote it down. The form has come back.

Not everywhere, and not at my place, where we’re hybrid and it works and I’m not writing this from grievance. But four days a week has become five at several large firms, three has become four at several more, and the direction is one-way.

The reasons haven’t changed, which is the tell

Here’s the thing I keep coming back to.

The reasons given for returning are collaboration, culture, mentoring, innovation and the value of unplanned conversation.

Those are the same five reasons that were given in 2019 for why remote working wasn’t possible. Word for word, in some cases from the same organizations.

Which is the interesting part, because in between those two statements we ran a four-year natural experiment across essentially every knowledge worker on earth. If the reasons were downstream of evidence, four years of evidence should have changed them, in either direction. Sharpened them, qualified them, produced a specific claim about which activities suffer and by how much.

They’re identical. Same list, same order, no numbers.

The reasons given in 2024 are the same five given in 2019, unchanged by four years of evidence, which tells you they were never downstream of evidence.

That doesn’t make the reasons wrong. It makes them convictions, and convictions are allowed, and they should be described as convictions rather than as findings.

What genuinely was lost

Now the part where I have to be fair, and it’s easy for me because I’ve been complaining about it in this blog for four years.

Teaching. I wrote in 2020 that code review had become nearly the only channel through which seniors and juniors have contact, and last summer I wrote that debugging is a method transmitted almost entirely by sitting next to somebody, and that the main way it gets taught quietly stopped happening in 2020. I believe both of those. They’re real losses and they land on the people with the least power to fix them.

The informal repair. A sharp exchange that dissolves by the kettle within the hour, versus one that sets for two weeks because the next contact is a scheduled call about a deliverable.

Ambient information. Knowing a project is in trouble because you can hear it.

And the beginning of things. Kickoffs, the first week of a new team, the messy generative bit where nobody knows what they’re building. Those do go better in a room and I’d fight to keep them there.

So: not an absolutist. There is a real case for presence, and it’s specific, and it’s about a handful of activities rather than about a number of days.

What a mandate actually produces

Which is why the shape of the policies is so odd.

The stated goal is collaboration. The instrument is a required attendance count. And the thing an attendance count produces, reliably, everywhere I’ve seen it, is people commuting for an hour in order to sit in an open-plan office wearing headphones on video calls with colleagues in other buildings.

That’s not collaboration. That’s the office as a location for remote work, at the employee’s expense, and everybody in it can see that’s what it is, which is corrosive in a way that’s hard to measure and easy to feel.

If the goal is teaching and generative work, the instrument would be different: whole-team days on the same day, arranged around the activities that need presence, with the deep individual work explicitly done elsewhere. Some places have done exactly that and it seems to work. It requires deciding what the office is for, which is harder than counting.

And the third one, which a few executives have now said publicly: a mandate is a way of reducing headcount without a redundancy process. If some proportion of people leave rather than comply, that’s a cost saving with no consultation and no payments. I find it hard to know how much of this is that, and it’s clearly not nothing.

Who pays

The same pattern as every time, and it’s the reason I’m writing this rather than shrugging.

The people most helped by four years of distributed work were the people the old arrangement served worst. Somebody with a condition that makes a commute expensive in a way it isn’t for me. Somebody doing the school run. Somebody who couldn’t afford to live near an office and had been quietly excluded from a whole tier of jobs. Somebody who moved 200 miles in 2021 on the strength of a policy that has now been changed by a person who wasn’t there when it was made.

Those are the people the reversal falls on, in the same order, and none of the announcements mention them.

Meanwhile a mandate is a small inconvenience for a senior person with a short commute and no caring responsibilities, which describes a lot of the people writing the announcements, and the difference between “small inconvenience” and “career-ending” is invisible from inside the first category.

Write it down while it’s obvious

I don’t think this is the end of it and I’m not being doom-laden. Labor markets cycle, this one turned against workers in about 2023, and it will turn back, and some of what’s being imposed now will be quietly relaxed when hiring is hard again. Which is its own indictment: it means the policy is a function of who has leverage rather than of what produces good work.

What I’m actually sad about is narrower. There was a window, around 2021, when something got proved and could have been settled, and the settling required somebody to write it down as a principle rather than as a concession. Nobody did, including me, and a concession is a thing that can be withdrawn.

Next time, and there’ll be a next time about something, write it down while it’s still obvious 🥲