Someone pointed me to this paper from 1983 (!): Ironies of Automation by Lisanne Bainbridge (PDF).
It talks about factory control rooms but, reading it today, it could be about generative AI.
What it basically says is: engineers tend to automate everything they can or know how to automate. Then humans get saddled with whatever’s left: usually keeping an eye on the system and stepping in when things go south.
But here’s where the “solution” becomes the problem: skills fade if you don’t use them. An experienced operator who spends months watching a machine do what used to be their work, slowly turns into an inexperienced operator.
Then, one day, when the machine does fail, probably in some weird way, they’re supposed to jump in and use skills they’ve now lost to take over and fix things.
The obvious analogy today is writing code, but it applies to other fields as well, in tech and elsewhere.
In 2009, the crew of an Air France plane lost control when the autopilot switched off due to ice. They’d never been trained to fly by hand at that altitude. Always trusting the machine.
A study of Polish doctors in 2025 found that, after they added AI to help spot issues in colonoscopies, they got much worse at spotting them on their own.
The examples are everywhere.
Let the chatbots write your emails, reports, and your code – you’ll feel super productive for a while. Then you’ll get worse at writing, analysis, and code. All the skills you need to notice when the LLM is being confidently wrong.
Automation can be useful, just make sure you don’t automate your brain away.
Colin