The impact of AI already goes further than just delaying hiring - at least in fields adjacent to engineering, such as technical writing. Anecdotally:
For the past 10 years, one of my best friends has been the senior copy editor for [Fortune 500 company's] sprawling website, managing more than a dozen writers. It's a great job, full time, mostly remote, with fantastic benefits (including unlimited PTO, a concept that I can't even fathom as a freelancer). The website comprises thousands of pages of product descriptions, use cases, and impenetrable technical jargon aimed at selling "solutions" to whatever Fortune 500 executives make those kinds of mammoth IT decisions.
Recently, he was telling me how AI was impacting his job. He said he and his writers started using GPT a couple years ago to speed things up.
"But now I have to use it. I wouldn't be able to work without it," he said, "because in the last year they laid off all but two of the writers. The workload's the same, but they put it all on me and the two who are left. Mostly just to clean up GPT's output."
I said, "I don't know who ever read that crap anyway. The companies you're selling to probably use GPT to summarize those pages for them, too." He agreed and said it was mostly now about getting AIs to write things for other AIs to read, and this required paying fewer and fewer employees.
So while AI may be a nice productivity booster, it's not like there's unlimited demand for more productivity. Companies only need so much work done. If your employees are made 4x more productive by a new tool, you can lay off 75% of them. And forget about hiring, because the tools are just getting better.
Coders like me don't want to believe this is coming for us, but I think it is. I'm lucky to have carved out a niche for myself where I actually own a lot of proprietary code and manage a lot of data-keeping that companies rely on, which effectively constitutes technical debt for them and which would be extremely onerous to transition away from even if they could get an AI to reverse engineer my software perfectly (which I think is still at least a few years off). But humans are going to be an ever-shrinking slice of the information workforce going forward, and staying ahead of those layoffs is not just a matter of knowing a lot about the latest AI tech or having a better resume. I think the smart play at this point is to prepare for more layoffs, consider what it would take to be the last person doing your entire team's job, and then wedge yourself into that position. Make sure you have the only knowledge of how the pipeline works, so it would be too expensive to get rid of you.
> So while AI may be a nice productivity booster, it's not like there's unlimited demand for more productivity. Companies only need so much work done. If your employees are made 4x more productive by a new tool, you can lay off 75% of them. And forget about hiring, because the tools are just getting better.
Many companies are also way overstaffed, IME (thinking non-software/"tech" F500s here)
Having worked as a consultant with various F500 companies over the last few years, there's loads of people that do very little work, and much of the work is low value--myself included; I make no claims I'm above any of this.
I've encountered countless project managers that do nothing other than move Jira tickets around.
Me: "Hey I'm blocked, can you get me in contact with $TEAM that owns this stuff"
PM: "Uh no, ask $PERSON"
How many of this person does any company need?
Even developers--I've worked with loads that take a week to set up some Angular project or cloud resources, and the even darker part of all of that is the whole project is destined to fail, cause the sales org sold em on some "modernization" thing that'll never get off the ground, that they don't have the staff to maintain, and they don't have the organizational will or discipline to integrate.
I've been on countless projects like this, there's piles of excess people doing low value (or no) work at all, saved only from unemployment by the sheer complexity of byzantine, bureaucratic organizations.
Honestly though, I think this is a structural and training issue, not a matter of who or what is answering the phone. Someone somewhere recently chained together a bunch of AI models to see what would happen if they emulated various departments in a large company, and they immediately learned how to pass the buck to each other and obfuscate the fact that nothing really got done. I don't think effiency, customer service or interdepartmental cooperation are really the problems the C-suite are trying to solve for by adopting AI anywhere they can. Or even productivity writ large, for that matter. Viewed through the lens of short term gains, AI looks like a way to cut costs and maintain at least the same level of bloat. In its current form, I'm suspicious as an investor of any company that treats it as a panacea for their structural and hierarchical maladies.
>Someone somewhere recently chained together a bunch of AI models to see what would happen if they emulated various departments in a large company, and they immediately learned how to pass the buck to each other and obfuscate the fact that nothing really got done.
I would love to read more about that; do you have a link?
This has been my experience as well. So much slop like this well before AI came out. Great for rest and vest/grifting but not good in general.
Wherever the vest disconnected from having a viable or profitable or useful product to invest in is where this whole shitshow of nontechnical people ruling over engineering serfs started. Roughly it went from search to ad networks to social graphs to crypto to NFTs to AI to vibe coding (with a physical side in jacking blue collar jobs in delivery, taxis and hotels), all driven by each 4-year crop of unimaginative business school frat boys piling on the last, as if they had been hazed into the delusion that they were smarter than the engineers they would hire and dispose of to get around the pesky laws. Now we're in the third generation of investors infused with the idea that one can grift indefinitely that way, the way that say, anyone who bought a house in California between 1946 and 2007 was guaranteed to make a fortune sitting on it. What we have run out of is people with non-derivative ideas and the actual skills to implement the same. To the extent that AI is just massive IP theft that might let the boys club continue to build the garbage they're currently choosing to build, by recirculating what has been done before, it's clearly a dead end. But they won't go down without a fight. What's so strange to me is, as a nerd boy born in 1980 in California, I definitely did not expect in the late 90s that the yuppies of 2025 would be even less original and more rapacious than their parents. That's what's been most disappointing about the first quarter of this century.
My father was a commercial artist in the 80’s-2000’s. He was also very early on the transition of analog desktop publishing to digital graphic design. This is exactly how the ad agency guys talked about things: at first, it was great. We could double our output and revenue, but soon we were caught in a rat race to the bottom as the tools improved and the skill floor fell away.