Some economists are warning there’s no sign of AI-related job displacement appearing in the labor data. Altman claimed it’s just a matter of time until it does.
Productivity gains from AI are appearing in many of the same fields where entry-level employment is starting to decline. Employment among software developers aged 22–25 has plummeted nearly 20% since 2024, even as their older colleagues’ headcount grows. The pattern repeats in other jobs with higher levels of AI exposure, like customer service. Meanwhile, firm surveys indicate executives expect this trend to accelerate, with planned headcount reductions outpacing recent cuts. Translation: The disruption is targeted and just beginning.
There’s already an increased productivity KPI at the corpo I work at. They expect faster completion of deliverables by some percentage as AI adoption grows (imposed).
At least in software development they can expect whatever the fuck they want, but when people are overworked due to trying to fulfill unreasonable expectations they make stupid mistakes so there are more bugs that slip through to much later phases were they’re harder to correct and the consequences of them are way more costly to undo and have no time for “spring cleaning” kind of work such as Refactoring code, so the codebase much more quickly becomes heavy, unwieldy and more bug prone - so harder to change (for example, do add new features or fix problems) and more likely to fail when it is change - leading to the need for a full rewrite (which costs $$$) much sooner.
In other words, increase the load on people and what you’re doing is causing a bit more work results delivered now in exchange for A TON of otherwise unnecessary work that will have to be done later (from extra bug fixing and even fixing the consequences of bugs, to full system re-writes because the old code base is an unmanageable mess) as well as a massive fall in productivity because the code has turned into a hot mess.
And this is without even taking in the account the consequences of using AI to generate code: unless we’re talking about such a minuscule project that it can be generated all in one go (i.e. non-professional stuff or tiny helper scripts), it’s not going to be a single consistent design with consistent coding practices, thus from the start already being a hot mess, way harder (read time consuming, read costly) to bugfix and change.
IMHO, there’s going to be some serious fireworks in just about all companies that went full-on with vibe coding, with the recent sequence of truly idiotic problems with Github and Windows 11 probably being a visible display of the beginning of that.
(Mind you, the funny bit is that all those senior techies that are already being hired to fix this shit and will need to be hired in larger numbers, are way more likely to tell any manager with unreasonable KPIs to go fuck themselves and even leave if pressed too hard to work themselves to the bone to try to achieve that which they know with absolute certainty is impossible, since it already is and it’s going to be even more a “sellers’ market” for the more experienced types so leaving is no problem).
Someone should add more KPIs into the mix. Surely we can play business politics better than those shit brained AI peddling monkeys.
What’s your rework percentage?
What’s your failure rate?
What’s your code churn rate?
What’s your spec-drift rate?
What’s your mean time to recovery?
What’s your defect-escape rate?
What’s your code cohesion score?
What’s your problem-discovery time?
What’s your cost per change over time?
What’s your risk appetite for losing all your devs to an inadequate replacement?
We’ve been finding ways to track dev performance for a long time. Surely we can call upon the correct set of KPI overlords to demonstrate, without actually saying, that AI is hot garbage for the scope it’s given.
Meanwhile:
https://hai.stanford.edu/news/inside-the-ai-index-12-takeaways-from-the-2026-report
This looks more like senior staff being forced to pick up more work with the same pay.
There’s already an increased productivity KPI at the corpo I work at. They expect faster completion of deliverables by some percentage as AI adoption grows (imposed).
At least in software development they can expect whatever the fuck they want, but when people are overworked due to trying to fulfill unreasonable expectations they make stupid mistakes so there are more bugs that slip through to much later phases were they’re harder to correct and the consequences of them are way more costly to undo and have no time for “spring cleaning” kind of work such as Refactoring code, so the codebase much more quickly becomes heavy, unwieldy and more bug prone - so harder to change (for example, do add new features or fix problems) and more likely to fail when it is change - leading to the need for a full rewrite (which costs $$$) much sooner.
In other words, increase the load on people and what you’re doing is causing a bit more work results delivered now in exchange for A TON of otherwise unnecessary work that will have to be done later (from extra bug fixing and even fixing the consequences of bugs, to full system re-writes because the old code base is an unmanageable mess) as well as a massive fall in productivity because the code has turned into a hot mess.
And this is without even taking in the account the consequences of using AI to generate code: unless we’re talking about such a minuscule project that it can be generated all in one go (i.e. non-professional stuff or tiny helper scripts), it’s not going to be a single consistent design with consistent coding practices, thus from the start already being a hot mess, way harder (read time consuming, read costly) to bugfix and change.
IMHO, there’s going to be some serious fireworks in just about all companies that went full-on with vibe coding, with the recent sequence of truly idiotic problems with Github and Windows 11 probably being a visible display of the beginning of that.
(Mind you, the funny bit is that all those senior techies that are already being hired to fix this shit and will need to be hired in larger numbers, are way more likely to tell any manager with unreasonable KPIs to go fuck themselves and even leave if pressed too hard to work themselves to the bone to try to achieve that which they know with absolute certainty is impossible, since it already is and it’s going to be even more a “sellers’ market” for the more experienced types so leaving is no problem).
Someone should add more KPIs into the mix. Surely we can play business politics better than those shit brained AI peddling monkeys.
We’ve been finding ways to track dev performance for a long time. Surely we can call upon the correct set of KPI overlords to demonstrate, without actually saying, that AI is hot garbage for the scope it’s given.
deleted by creator
I do wonder where all the non-juniors are spawning in, i guess the it will be time to whine about a shortage then.
They assume AI is only to get much better and will be able to replace higher expertise levels every year until everybody developer is replaced.
You assume long term planning ? Haw haw. Line must go up this quarter!
You know how it goes. Fire people and assume AI is better next quarter. If it doesn’t work out, it’s also a problem for next quarter.