New study by the National Bureau of Economic Research: A survey of 6000 CFOs, CEOS throughout US, Europe, UK and Australia comes to the conclusion that businesses predict that "AI" will improve productivity by a whopping 1.4%.

[nakob@anarchism.space]

@tante Does this include the 100% productivity gains that electricity companies will have due to AI?

[nilz@norden.social]

@stockach @tante

I still hear that "we must learn how to use it right" , that AI keeps getting better and better but the organisations learning curve is too slow. Go visit a prompting seminar!

Ha ha

[amorpheus@kind.social]

@tante The current purpose of AI goes beyond economics.

[q@goeppingen.social]

@tante not sure who created the study but it's wrong.
I know incoming raged response of AI sycophancy with insults by backwood neckbeards

[ftranschel@norden.social]

@tante I'd wager that even the 1.4 percent go away once you account for people simply behaving differently with AI available - and it obviously will turn negative once the skill dependencies are fully established for "power users".

[tante@tldr.nettime.org]

@q you can check out the results and comments. What do you think the researchers did wrong? Did the CEOs lie?

[ftranschel@norden.social]

@Amorpheus @tante That may very well be, but it is much harder to measure.

This result alone is the nail on the coffin of scaling, because in order to offset the investment frenzy, not even ~10% net gains would be enough. And there is not even a *hint* that this'd be true for edge cases such as coding where arguably one could *assume* some real gains to be found.

[ftranschel@norden.social]

@nakob @tante Productivity != profitability.

[amorpheus@kind.social]

@ftranschel @tante Yet knowing that ai is not profitable enough to legitimize the effort, they all continue their agenda. I wonder why.

Are they stupid or do they see revenue where the market does not?

Now this is just a thought, not a conviction...

If the current ai hype proceeds, it will become an unvaluable tool for worldwide surveilance and oppression.

[vladimir_lu@hachyderm.io]

@nilz @stockach @tante Is a “prompting seminar” like a visit to an “untherapist” (where you shout loudly and randomly at passers by that you are perfectly fine)

[ftranschel@norden.social]

@Amorpheus @tante

Conspiracies aside: In a bubble market, it *is* possible to transfer wealth even if there is *none* in the long run.

[collimated_thought@defcon.social]

@tante And yet "Despite $30–40 billion in enterprise investment into GenAI, this report uncovers a surprising result in that 95% of organizations are getting zero return." From an MIT study on the outcome of AI projects. https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf

[rainer_rehak@mastodon.bits-und-baeume.org]

@tante I guess this already small increase mostly comes from people working longer working hours now because of the bosses' high expectations when introducing "#AI" tools.

Also see this related news: https://techcrunch.com/2026/02/09/the-first-signs-of-burnout-are-coming-from-the-people-who-embrace-ai-the-most/

[tante@tldr.nettime.org]

@Rainer_Rehak Might be. Right now about half is just "firing people" (which then gets the rest to do what you described) and the hope for very marginal output increases.

[maxheadroom@hub.uckermark.social]

@tante @q this. It's a survey that asked for their opinions (partially). The CEOs and CFOs and other C-suite execs in these surveys surely have no clue what's going on in the daily life of their employees who have to actually work with all the AI stuff. The C-level guys don't even read their own email but have staff summarising them in PowerPoint's. So I'd say the data basis is very thin here...

[tante@tldr.nettime.org]

@maxheadroom @q they do know (at least for the last years) the numbers: How much OpEx and CapEx are there in contrast to revenue. So They can pretty accurately say what happened in the past without knowing who works how

[maxheadroom@hub.uckermark.social]

@tante @q I still question their capacity to relate this to AI one way or the other. Nevertheless, the low value of expected increase is still telling. Given the cost and likely increase in opex by all the users.