Last week, my VP stated in his Town Hall that adopting AI is actually
increasing our workload by as much as 20%. We have our normal job, plus
training AI to replace our boring workloads, then checking AI's results
as AI, by its nature, is non-deterministic. Supposedly, the payoff is
allowing Agentic AI to do the mindless stuff and let the humans do the
thinking stuff. What they haven't learned yet is the 6-week AI failure
point. It's not that AI is actually failing, it's the psychology that
after 6 weeks, humans stop monitoring AI "because it's working fine" and
the non-determinism grows as it is not being monitored and corrected.
(See note below.)
Since we're all salaried, management sees no direct increased cost. The
company has stated there will be no lay-offs due to AI. There won't be
any hiring, either. Now it's just a matter of how much extra they can
pile onto us while measuring increased productivity "due to AI adoption"
(AKA more work, same salary) until us old farts age out and retire and
the others get fed up and leave.
I feel like this incarnation of AI is like the first model year of most
cars, or the first release of any software product. The early adopters
suffer the most.
I hear what some of you are saying about disruptive technologies. My
great-grandfather was involved in one of the most the disruptive
technology of the WWI era - the automobile. He declined an invitation
from Walter Chrysler (yes, that Chrysler) to overhaul and revitalize an
ailing car company in 1919. He was a car-hater and thought rail was the
future. Don't be my great-grandfather. Don't be an AI-hater. After this
bubble bursts, the real value will emerge and cooler heads will turn it
into something useful.
Note: I understand that psychology. I developed a system to block
dangerous commands executed with privilege (AKA sudo or pbrun) based on
a consultant's observation. I had to set a weekly calendar reminder to
check it to see how well it operates because after a few weeks, it was
running smooth. It works as designed and I'm just that good. Okay, just
kidding - testing is a great ego deflator. Then we turned it into a
dashboard item that I check a couple times a week and laugh at ID10T
tricks. BOFH Lives!!! Now Audit requires we empirically test the
configuration quarterly and the process is formally evaluated annually.
See https://gtfobins.org/ for four hundred ways to become root or read
configs as non-root. And I get to test it every quarter. Every item.
Yes, I've semi-automated it, turning a 2 day mind-numbing evolution into
two hours of testing and test script adjustments.
Regards,
George Toft
On 7/2/2026 11:42 AM, Keith Smith via PLUG-discuss wrote:
This is what some of us predicted. I predict rocky roads ahead.
https://ktla.com/news/nationworld/ford-rehires-engineers-after-ai-backfires/
Keith
---------------------------------------------------
PLUG-discuss mailing list: [email protected]
To subscribe, unsubscribe, or to change your mail settings:
https://lists.phxlinux.org/mailman/listinfo/plug-discuss
---------------------------------------------------
PLUG-discuss mailing list: [email protected]
To subscribe, unsubscribe, or to change your mail settings:
https://lists.phxlinux.org/mailman/listinfo/plug-discuss