We have been here before (sort of)

Every few decades a machine turns up that is going to end a job, and every time, somebody writes the obituary early. It is worth looking at a few of those obituaries, because the history is more useful than either “it’ll be fine” or “it’s different this time”, which are the only two things anyone ever says about it.

Three stories, then the part people leave out.

The cash machine and the bank teller

Cash machines arrived in the 1970s and spread for decades. In the US their number quadrupled to about 400,000 between 1995 and 2010. Bank tellers, whose job was largely handing out cash, should have been finished.

They weren’t. The economist James Bessen found that US teller jobs actually rose, from about 500,000 in 1980 to about 550,000 in 2010. The mechanism is the interesting bit. Cash machines made a branch cheaper to run, so each branch needed fewer tellers, down by more than a third between 1988 and 2004. But because branches were cheaper, banks opened more of them: urban branches rose by more than 40%. And the tellers who were left did less counting and more talking. Loans, accounts, “have you thought about a credit card?”. The job became a sales job with a till attached.

On this site, bank tellers score 42. The counting and cash-handling tasks are still the most exposed. That has been true since 1975.

The spreadsheet and the bookkeeper

VisiCalc arrived in 1979 and did in seconds what a bookkeeping clerk did in a week: recalculate a whole ledger when one number changed. NPR’s Planet Money ran the numbers: since about 1980, some 400,000 bookkeeping and accounting clerk jobs in the US went away. Over the same period about 600,000 accountant jobs were added.

So the spreadsheet didn’t shrink accounting. It made accounting cheap, and when something gets cheap people buy a lot more of it. Businesses that used to do their numbers once a year started asking “what if?” every afternoon. Somebody had to ask the questions, and that somebody was an accountant.

The split still shows up in the data. Accounting and bookkeeping clerks score 36 here. Accountants score 49. Same field, very different amounts of the job that a machine can reach.

Desktop publishing and the typesetter

In the early 1980s, getting words onto a printed page took a small army of specialists: typesetters, paste-up artists, camera operators, plate makers. At their peak there were around 4,000 typesetting firms across North America. Then the Mac, PageMaker and the laser printer turned up, and graphic designers began setting their own type.

Most of those typesetting firms are gone. But “designer” became one of the most common creative jobs going, and the designer’s job quietly absorbed four or five of the old specialist jobs into one person with a mouse. The work didn’t vanish. It moved onto a different desk and got a new name. (Fun footnote: “desktop publisher” became a job title of its own, and the US government now expects it to shrink by 14% over the next decade. The machine that ate typesetting is now being eaten.)

The part people leave out

Here’s where the cheerful version of this story usually stops. It shouldn’t.

The teller story had a second act. After 2010, mobile banking did what the cash machine couldn’t: it kept people out of the branch entirely. Branches started closing, and the US Bureau of Labor Statistics now projects teller jobs to fall 13% between 2025 and 2035. The same job can survive one machine and lose to the next.

“The field grew” is not the same as “the people were fine”. The 400,000 bookkeeping clerks who lost their jobs were not, in the main, the same people as the 600,000 new accountants. Some retrained. Many didn’t. A job category recovering is not the same as a person recovering, and the gap between the two is where the hardship lives.

It was slow. Each of these took decades. People had time to retire out of the old job and be trained into the new one. Whether today’s tools arrive on that timescale is a genuine open question, and nobody honestly knows the answer.

So what do these stories actually teach?

Three things, and none of them is “relax”.

1. Machines take tasks, and jobs reshuffle around what’s left. That’s why every score on this site is built from tasks, not job titles. The teller kept the conversation and lost the counting.
2. Cheaper often means more. When a machine makes something cheap, demand can grow enough to create more work than it destroyed. It doesn’t always. But it happens often enough that “fewer tasks means fewer jobs” is not a law of nature.
3. The winners were the people nearest the new work. Tellers who could sell. Bookkeepers who could advise. Typesetters who learned the Mac. If you want to know which way to lean, look up your job and see which of your tasks are the ones the machines like. Then lean the other way.

We have been here before. Sort of. The honest version is that we’ve been here several times, it went better than predicted for the jobs and worse than the averages suggest for some of the people, and it took long enough that most of them could adjust. That last part is the bit worth watching.