CEOs can be profitably replaced by AI whereas the line workers not.

Keep pushing and sooner or later the shit will stick, right?....Right?

If/when the AI subsidies get pulled the cost/benefit is going to take a nose dive. Things could get really interesting... [1]https://thehill.com/opinion/te... [thehill.com]

[1] https://thehill.com/opinion/technology/5581314-ai-investment-government-restraint/

What the CEOs are really saying is that AI is making THEIR work more efficient. And since CEOs are just like normal people, AI must be making everyone's work more efficient too. After all, no one has every accused a CEO of being myopic.

Two observations:

1. AI amplifies existing competence. Current AI systems are not autonomous problem-solvers -- they are accelerants. Users who already understand their domain can spot errors, ask better follow-up questions, and integrate outputs efficiently. For less experienced or less analytical workers, AI often creates rework rather than savings, which is consistent with the “AI tax” described in the surveys.

It’s not strictly about being “smart,” but also about task structure and feedback loops. Some highly capable workers are in environments where AI cannot be safely or efficiently applied (compliance-heavy workflows, fragmented tooling, high-stakes accuracy requirements).

2. Impact varies strongly by role, as knowledge work is not homogeneous. Roles involving synthesis, drafting, ideation, coding, analysis, or decision support benefit far more than roles dominated by coordination, approvals, interpersonal judgment, or rigid process constraints. Executive workflows are especially well suited to AI assistance, which explains the perception gap between leadership and individual contributors.

Also, adoption maturity matters. Many organizations have introduced AI without training, workflow redesign, or incentives, which predictably limits upside regardless of worker capability. Would it be exactly surprising to see this outcome? Not to me.

So let’s say you work at a company, and they have an OKR “use AI to improve productivity”. If you aren’t familiar with OKRs, they are something that your manager tells you to do, and the more of them you do the more likely you are to get a raise, or bonus, or at least keep your job. Sometimes they are flat out assigned to you, sometimes you get to pick off a list, sometimes you and your manager come up with them together. You manager likely has an OKR handed down to them to get their employees to use more AI, so this isn’t a pure blue sky thought experiment. It is a realistic situation.

So you end up with this OKR. You can ignore it and at the end of performance cycle you can either fess up “I didn’t even try”, or you can say “I gave it a shot and AI didn’t help”, or you can flat out lie “AI helped some”, or “AI helped a lot”. Which do you do? Which do you do if you were also pretty lazy on the other OKRs and really have almost nothing good to report? If you aren’t lazy yourself you already know any OKR with no real way to double check is something you report having achieved because you were too lazy to do any of ‘em.

If you aren’t lazy you have a bunch of OKRs that you actually managed to do (get 90% of TPS reports in on time, screen all your bugs at least twice a week, whatever). You also likely have a few you didn’t, and you have this one here. Maybe you didn’t bother with AI which gives you the chouces up above in “lazy”, or maybe you gave it a shot and it didn’t help, so you can report a failure, but you worked so hard on the other OKRs, do you really want to jeopardize your bonus because you have this AI OKR?

If you are super honest maybe you will report the AI OKR as a bust. Maybe your company actually has a “if you aren’t failing at least 20% of your OKRs you didn’t set high enough expectations” policy, and sure you can pass or fail some number of the unverifiable OKRs as needed to hit that magic success rate...but that is rare, it is far more common for a company to treat OKRs as “more is better!”.

Plus even if you are fundamentally an honest person, I’m sure you used AI once or twice to summarize someone else’s long emails into something shorter and maybe inaccurate, but surly that saved time, right? At least as long as they weren’t too inaccurate! Or maybe you used it to fluff up a short email/report into something longer, even if you then spent just as long double checking that it isn’t now inaccurate as you would have fluffing it up yourself (plus now everyone else ends up with longer emails they use AI to summarize...). Or you write code for a living, and you AI’ed up some code, and that saved you like 10 hours of coding, I mean it cost 45 hours of extra debugging, but you saved 10 hours somewhere , so you can report meeting your OKR without a lie!

If you are asked by someone outside your management chain, and outside your company the honest answer is somewhere between “I didn’t try”, to “it didn't help”, to “it saved me time in one place, but maybe cost me more in another place”, and occasionally “yeah it was helpful somewhere "

So workers are reporting “yeah, AI makes me more productive” up the management chain because that makes rewards flow back down the management chain. Which makes CEO’s think “this shit works!”, I mean it is exacerbates the problem of upper management job being the kind of thing AI can do anyway, of taking in a ton of data and making choices without understanding what the fuck is really going on anyway, so CEO’s already see AI “working” and they are inclined to believe it, especially when their whole management chain reports it as working...

It reiterates my view of AI as the great equalizer. You're doing generic art? Be afraid of AI. Your're middle management but your job is more about selling yourself than actually doing competent work, and if you give generic or stupid instructions, people will work around that? Be afraid of AI. Really, it's not true for everyone (roto-artists come to mind), but in most professions, if an AI can do your job, you're already a waste of everyone's time.

85% reported time savings of one to seven hours weekly, much of it was offset by correcting errors and reworking AI-generated content

Assuming there was no AI, how long would it have taken these people to create whatever it is they used AI for? For example, if creating the spreadsheet would take 3 hours, but using AI it only took 15 minutes plus an additional hour to correct errors, that is still a net saving of almost 2 hours.

For reworking content, how long would it have taken them to come up with the content themselves? If you're not someone who can create a document on the fly, how long would it take you compared to having AI do it then reworking it?

Not saying AI is the best thing since Betty White*, but if you're doing comparisons you need to take into account everything involved. Simply saying, "I spend two hours each week correcting what AI does" does not indicate if that is good or bad. Compare it to how long it would take you to do the same thing versus using AI and correcting it.

* Betty White was born in 1922. Sliced bread didn't happen until 1928.

My director is extremely excited about how much time AI saves him writing up his weekly summaries to pass up the chain. He can write a single sentence, pass it to the AI to be written up into a big fancy report, the next guy up the chain uses another AI to decipher the big fancy report into a single sentence summary that may, or may not, be vaguely like the original sentence used to prompt the AI that wrote the big fancy report, and both of them are super pumped that they saved so much time. My director saved writing time, his next-up-the-chain saved the time of having to read what my director wrote. They've just used a bit more compute to avoid typing and reading things that were, at best, busywork created to make it seem as if they were involved in the process of what happens beneath them. AI can really help with busywork reporting.

I'm guessing neither of them will realize they're automating away the entire reason their jobs exist.

CEOs save time using Ai because their jobs are easy and they are kinda stupid, so they normally take a lot longer to do easy tasks than the average worker. Even the current garbage Ai that exists can do their job better than they can. Real workers can't be so easily replaced.

Early adopters of immature tech rarely have a smooth transition.

There will be a lot of confusion and wasted time as workers try to figure out how to use the tools and fix the mistakes the tools make.

This gets worse because the tools are rapidly evolving.

Changing a complex system is hard, slow and expensive.

> So now the Emperor walked under his high canopy in the midst of the procession, through the streets of his capital; and all the people standing by, and those at the windows, cried out, "Oh! How beautiful are our Emperor's new clothes! What a magnificent train there is to the mantle; and how gracefully the scarf hangs!" in short, no one would allow that he could not see these much-admired clothes; because, in doing so, he would have declared himself either a simpleton or unfit for his office. Certainly, none of the Emperor's various suits, had ever made so great an impression, as these invisible ones.

Remember-- CEOs are not accountants. They're managers with the EXPRESS obligation to continually increase investment. They must always be in the process of encouraging people to buy their stock.

The major CEOs will never say, "I was too stupid. I was taken in by the promise of large language models thinking it was actual, Star-Trek-level artificial intelligence. We increased our software expenses by 20%, reduced our workforce headcount by 20%, and it hasn't paid off. In fact, we're significantly less financi