How employment changes when firms adopt generative AI
ramp.com
ramp.com
[1]: https://libertystreeteconomics.newyorkfed.org/2026/06/the-po...
Similar to the current post there’s just not enough data to be conclusive.
I don't see this narrative being more prominent than the "AI won't be displacing jobs because things worked out in the past" counternarrative.
For example, only recently has The Economist published articles that go more towards that narrative (AFAIK) [1].
> there’s just not enough data to be conclusive.
With this, I can agree. We can still extrapolate based on what we currently know.
[1] https://www.economist.com/leaders/2026/05/14/prepare-for-an-...
Cool you spent 2 days and $200 building a react UI for your spreadsheet.
Is it easier to implement a design now? Then let’s redesign every quarter.
Is it easier to refactor the whole codebase? Then let’s rewrite in whatever new hotness.
This is not different from web frameworks 10-20 years ago. Is it easier to make a website using Rails? If so, why did I see an explosion in devs-per-project after I started using Rails? Because of the BS.
I'm sure there is a big caveat in there somewhere.
The tech everything's-hypergrowth era is, for now, over. Most of the low-hanging fruit that we've collected for the last 20 years of bull market (putting tech into every business) is gone, and companies hired in advance of having the next business to overlay on their current ones. For many that didn't materialize.
I am confused. Wasn't the initial claim that AI kills all jobs?
I feel the claim right now is not really correct. One needs to do a thorough analysis of the whole job market across different countries, say, over 5 years. Or at the least 3 years but very complete and unbiased either way.
The counter claim that LLMs were special for some reason has never been supported. And a lot of its proponents have other axes to grind, like UBI, Socialism, Communism etc. It would suit their worldview for this time to be different, and so thats the message they push.
I think the bigger conclusion is that LLMs are, for now, having a minimal effect on the labor market. And I think this makes sense. In spite of individual claims of 10x productivity boost or whatever, their effect on the bottom line of companies seems to remain quite unclear, at best. On the contrary the token-maxing catastrophe seems to have resulted in companies becoming far more price conscious towards LLM usage.
In preindustrial societies nearly everyone was working to support themselves and their families, these days a huge percentage of the population is in education, retirement, prison, disability, etc.
The only way jobs have kept up with automation is when you ignore population growth, but more people naturally increases the required workforce. You inherently need more police, food, etc when you have more people.
What evidence?
Wasn’t employment in engineering down before all this could’ve gotten enough widespread traction to actually displace jobs?
> There's never been an automation revolution that worked otherwise.
Requires all automation to have zero effect on net jobs long term. Meanwhile the percentage of people working is down so that’s clearly not true.
Ignoring for a second that net jobs and employment rates aren't really the same thing.
>In preindustrial societies nearly everyone was working to support themselves and their families, these days a huge percentage of the population is in education which may be investing in their future but isn’t directly producing anything. Retirement as a percentage of the population similarly exploded etc.
The first thing anyone does in employment statistics is remove non participants. Bringing them back in is weird. If you don't need or want a job its kind of a non sequitur to be lumped in with the employment seeking population. AI doomers aren't suggesting that its going to gainfully retire the population.
>preindustrial societies
Pre industrial societies can be loosely grouped into "People farming to make 3-5 times the food they need" and "city dwellers". Now that a single person can farm for 100s of people, we do have hundreds more city jobs going. We dont have a huge number of out of work farmers sitting around doing nothing. Likewise, Banks employ more people after introducing ATM's than before. Likewise cloud didnt leave IT people lining up at the dole office, but it just moved them from cleaning up on prem messes to cleaning up cloud messes and onprem messes.
that's a common trope, and its both true and false
true: more bank employees after ATM's
false: less bank employees after smartphone banking
28 million in 2026 vs 18.8 million in 2001 is a ~50% jump…
So no the numbers are down significantly even with outsourcing.
LNP: Taxes less.
Labor: Taxes less (as a proportion of GDP)
(Seeing as GDP includes government spending, it seems like you can make the rate of tax go down by increasing government spending by more than taxes increase)
If a claim is about X, introducing another variable doesn't invalidate the claim about X.
Ignoring that when the global population is likely to peak soon due to falling birth rates is a massive mistake. Jobs in a world of falling population and increasing automation is largely uncharted territory.
Its just for the government to rebalance the economy if those jobs are desirable to reshore.
Not as a percentage of the population. In 1967 the US population in 1967 was 200 million today it’s 350 million. Globally the growth was even faster.
Your farming productivity numbers are also very exaggerated. The actual percentage of the population was very high because of food spillage, and inconsistent productivity due to weather etc. They often needed ~90% of the adult population to be farmers or people starved, and famine was still common. Right before industrialization that number started to drop which is largely what caused industrialization. You got a positive feedback loop of increased productivity when a vastly larger percentage of the population could do something other than produce food.
Didnt I address that explicitly?
You also found it odd I didn’t “remove non participants” but preindustrial societies didn’t necessarily have formal jobs for people. You’ll note I didn’t exclude modern stay at home parents as non workers because labor is the metric not if you have formal employment.
Net Jobs can increase while their percentage of all jobs drops. These things aren't necessarily linked or related. There isnt really a built in expectation that banking for 10 million people needs 10 times less staff than banking for 100 million people.
If automation doesn’t cost jobs, then growing the population shouldn’t change the overall employment rate meaningfully.
No, they are mostly right wing capitalists. LLM is not a left wing project. UBI is something they talked about, but super hard to believe it is anything but a way to shut up critics.
> The claim is and always has been that automation increases net jobs.
That was definitely not the common claim last 4 years.
> There's never been an automation revolution that worked otherwise.
That is not what historical record shows. In the short term, larger unemployment are to be expected. Over long term, it evens out.
It's not the first industrial revolution, we've been in a good dozen of them now and absolutely none have resulted in large unemployment.
The only large unemployment we've experienced until now were linked to economic crisis which had nothing to do with industrial revolutions
The proponents of the claim not the proponents of LLMs.
If that is not soul crushing I would not know what is.
"Look at all the traditional signals of thought, quality, or attention that it inserted, which being totally counterfeit have actually become strictly counterproductive--at least to anybody who has to actually use it!"
- explained back to me the original request I had filed with details
- assumed the request was about something it was not
- didn’t reply to my question
- had no actionable element whatsoever
- repeated all of the above twice with different words.
This is a senior person running a marketing agency. The absurdity here is that I have a high cost per hour that falls back onto our shared client. Who I will now have to brief about how I spent about 2 hours due to this comms back and forth and deciphering LLM-novels on a problem that could have been solved in a 30min slot if the agency only took 5 minutes to read, understand and use their brain. The added time was for me to basically go back to my comms, understand if I had communicated poorly (I asked a non-tech third party: they got it in 3 minutes), and reassess everything twice. Result: two hours to do something that amounted to changing a link behavior.