AI keeps creating better jobs

Everyone is arguing about AI unemployment right now.

Anthropic is warning about job displacement. Policy people are trying to slow the whole thing down. The doomer take is everywhere.

But if you zoom out, we have seen this movie before.

Three times in the last 125 years, a technology showed up that looked like it would wipe out entire categories of work.

Then the opposite happened.

The task got automated.

The work expanded.

The human moved up the stack.

Radiology is the best current example

In 2016, the godfather of AI said we should stop training radiologists because AI would soon do the job better.

Fast forward to 2025.

Radiology jobs are at record highs. Residency spots just hit an all-time record. Average radiologist pay is up 48% since 2015, now around $520K a year.

And radiology is where AI has been deployed the hardest.

In 2024, 78% of all cleared AI-enabled medical devices were for radiology.

So if AI was supposed to crush one field first, this was the field.

Instead, it created more demand.

Why?

Because the job was never only “read the scan.” That was one task inside the job.

The real work is judgment. Communication. Explaining tradeoffs. Helping patients and doctors make better decisions.

AI took more of the repetitive work and made the human work more valuable.

The tractor did this first

From 1900 to 1960, tractors decimated farm employment.

In 1900, there were about 12 million agricultural workers. That was 41% of the workforce.

By 1970, agriculture was down to 3.5 million workers, or about 2% of the workforce.

On paper, that looks brutal.

But while agriculture lost 8 million jobs from 1910 to 1970, new industries gained 46 million jobs in the same period.

Retail. Professional services. Construction. Manufacturing.

Technology removed work people no longer needed to do by hand, then opened up more valuable work somewhere else.

That is the part most AI doomers miss.

The scary chart is usually the task being automated.

The opportunity is what humans get freed up to do next.

Spreadsheets did it again

The same thing happened with spreadsheets.

VisiCalc, Lotus 1-2-3, and Excel reduced the need for bookkeepers.

Around 400K bookkeeping jobs disappeared during the spreadsheet era.

But 1.3 million accountants and financial analysts were added.

That is a 3x return on displacement.

The work did not vanish. It moved up a level.

Instead of manually entering numbers all day, more people moved into analysis, planning, forecasting, and decision support.

This is the pattern I think operators should study.

If your job is one task, AI is a threat.

If your job is judgment around a set of tasks, AI is leverage.

That is why we are building Single Brain the way we are.

Stacking 20 disconnected AI tools just gives your team another integration problem.

A shared memory layer lets agents take the repetitive work, while your people spend more time on the calls, decisions, creative judgment, and customer problems that actually move the business.

ATMs did it too

In 1973, the New York Times predicted ATMs would cut bank teller jobs by 75%.

Instead, teller employment grew 81% from 1970 to 1988.

Why?

ATMs made branches cheaper to operate. Banks opened more branches. Tellers shifted from only processing transactions into more customer-facing work.

Again, the task changed before the job changed.

This is why I do not buy the “mass unemployment forever” argument.

Short-term displacement is real.

Some jobs will get hit. Some people will have to reinvent themselves. That part should not be minimized.

But history keeps showing the same pattern.

Automation compresses the boring work.

Demand expands around the higher-value work.

New roles appear that nobody had language for before.

One stat from the video says 60% of today’s jobs are occupations that did not exist in 1940.

Another says 85% of job growth over the last 85 years has been driven by creative destruction.

That is the game.

The move is to reinvent before you are forced to

History rhymes.

The tractor. The spreadsheet. The ATM. Radiology AI.

AI is the next version of the same pattern, just faster.

So the wrong move is pretending nothing changes.

The better move is asking: what part of my work is robotic, and what part becomes more valuable when the robot handles it?

If you are a marketer, that might mean less manual reporting and more strategy.

If you are in sales, less CRM cleanup and more relationship building.

If you are an operator, less chasing people for updates and more seeing the whole system clearly.

The people who win are not the people who protect every old task.

They are the people who keep moving up the stack.

Watch the full video here:
https://www.youtube.com/watch?v=ZhplFht3sJ0

To building work that compounds,

Eric Siu