Forgot your password?
typodupeerror
Businesses IT

Study of 12,000 EU Firms Finds AI's Productivity Gains Are Real (cepr.org) 61

A study of more than 12,000 European firms found that AI adoption causally increases labour productivity by 4% on average across the EU, and that it does so without reducing employment in the short run.

Researchers from the Bank for International Settlements and the European Investment Bank used an instrumental variable strategy that matched EU firms to comparable US firms by sector, size, investment intensity and other characteristics, then used the AI adoption rates of those US counterparts as a proxy for exogenous AI exposure among European firms.

The productivity gains, however, skewed heavily toward medium and large companies. Among large firms, 45% had deployed AI, compared to just 24% of small firms. The study also found that complementary investments mattered enormously: an extra percentage point of spending on workforce training amplified AI's productivity effect by 5.9%, and an extra point on software and data infrastructure added 2.4%.
This discussion has been archived. No new comments can be posted.

Study of 12,000 EU Firms Finds AI's Productivity Gains Are Real

Comments Filter:
    • by garlicbread2 ( 4853285 ) on Wednesday February 18, 2026 @04:55PM (#65997748)

      Wait a second was this news article written by AI in response to the first news article?
      Is it getting smarter and self aware

      • by SumDog ( 466607 )
        It's randomly generating words to preserve itself, even though it doesn't actually understand what any of those words mean or what preservation is. Those weighted parameter models certainly have some emergent properties.
      • Its funny because even if it's not fishy as fuck the gains are like less than you'd lose rolling out zero trust infrastructure or having a poorly configured jira

    • by VorpalRodent ( 964940 ) on Wednesday February 18, 2026 @05:32PM (#65997852)
      This was a multivariate study, and while they did find a slight increase in productivity, they also found that eggs are now bad for you again.
      • Dangit! What about Olive Oil? Still good for you? Have they finally discovered that sitting on your ass and playing videogames is healthier than jogging?
    • by AmiMoJo ( 196126 )

      I find that AI can sometimes help with productivity, but other times it just wastes your time, so it probably balances out.

      There was some Windows API stuff I wasn't familiar with, so I asked Gemini to write some sample code, which turned out to be better than the Microsoft example. So that time productivity increased.

      Then I asked it to write an app that took scans of CDs (the discs laid face down on a flatbed scanner) and rotated them to be upright, or at least at some multiple of 90 degrees. It failed, tri

  • used the AI adoption rates of those US counterparts as a proxy for exogenous AI exposure among European firms.

  • Fitter, happier, more productive
    No longer empty and frantic
    Like a cat, tied to a stick
    That's driven into frozen winter shit

    • by Anonymous Coward

      Fitter, happier, more productive No longer empty and frantic Like a cat, tied to a stick That's driven into frozen winter shit

      Wait until the next version of Claude [slashdot.org]. I've heard from insiders that it adds 2 inches to your dick and/or 2 cups sizes to your bust plus it does the dishes, doesn't complain, and performs enthusiastic oral sex on demand!

  • I swear there was another article on slashdot saying the exact opposite today.

    • by allo ( 1728082 )

      I think this one says "You are more productive" and the other said "You won't be fired". The combined conclusion: Using AI does not replace you, but increases productivity and so leads to more and/or better output without the need to lay of people.

  • by thegreatemu ( 1457577 ) on Wednesday February 18, 2026 @05:10PM (#65997798)
    Let's set aside all of the displaced costs from datacenter operations and just focus on direct costs. AI increases productivity by 4%. How much did these companies pay for premium licenses (or, how much will they pay once we are past the "first taste is free" period)? And how does that compare to hiring 4% more people? The article does not include any information about cost. But if you spent more on AI to get that 4% bump than you would have by hiring more people, you are not only failing to run the company efficiently but also screwing over people who could benefit from those potential jobs.

    Of course, if anyone actually had to pay for all of the externalized costs, the net benefit would be negative.
    • Yes that'll be the problem, even if it helps some it won't be enough of a bump to charge enough to pay for the data centers
  • by crunchy_one ( 1047426 ) on Wednesday February 18, 2026 @05:12PM (#65997804)
    Their 4% productivity claim is valid only if you accept the method they used as valid. TFA provides little to no detail to support their method beyond this claim:

    To credibly identify the causal effect of AI on productivity, we develop a novel instrumental variable strategy, inspired by Rajan and Zingales’ (1998) seminal work on financial dependence and growth. Their key insight was that industry characteristics measured in one economy – where they are arguably less affected by local distortions – can serve as an exogenous source of variation when applied to other countries.

    We extend this logic to the firm level. For each EU firm in our sample, we identify comparable US firms – matched on sector, size, investment intensity, innovation activity, financing structure and management practices. We then assign the AI adoption rate of these matched US firms as a proxy for the EU firm’s exogenous exposure to AI. Because US firms operate under different institutional, regulatory and policy environments, their adoption patterns capture technological drivers that are plausibly independent of EU-specific factors. Rigorous propensity-score balancing tests confirm that our matched US and EU firms are virtually identical across key observable characteristics, validating the identification strategy. Our analysis draws on survey data from EIBIS combined with balance sheet data from Moody’s Orbis.

    From there, they magically jump to their conclusion without any supporting data. Looks like a garbage burrito served with a side of word salad.

    • Yes, this data analysis protocol makes no sense at all.

      There is no reason to think that the adoption rate of AI by US firms tells you the adoption rate of AI by European firms. And absolutely no reason to think that these results are accurate enough to detect an effect as small as 4%.

      These results are completely meaningless.

  • It's the EU (Score:3, Funny)

    by PPH ( 736903 ) on Wednesday February 18, 2026 @05:47PM (#65997882)

    Just showing up for work during the month of August will contribute a lot to productivity figures.

    • EU here... You've seen too many YouTube shorts parodying this. Except for Christmas, the show goes on. Go on holiday in August? Expect a pile of work when you get back.
  • by Baron_Yam ( 643147 ) on Wednesday February 18, 2026 @05:52PM (#65997890)

    But then I mainly use it as a natural-language front end to search engines, and to provide small snippets of code so I don't have to rewrite them or find my local copy.

    There's no doubt it's made me more productive. Teaching the new guys how to use AI tools effectively in the same way though is weirdly difficult. They're not stupid, but I don't think they had enough time doing it the hard way to know what questions to ask or how to phrase them to get a useful answer... or how to give the responses a critical eye for errors.

    • by nwaack ( 3482871 )

      They're not stupid, but I don't think they had enough time doing it the hard way to know what questions to ask or how to phrase them to get a useful answer... or how to give the responses a critical eye for errors.

      This. In order to truly be faster at your job when using AI, you have to already know the proper questions to ask the AI when you're stuck, including the proper context around the question, in order to get a useable response. This is something that comes only with experience.

  • Does this actually mean 4% productivity increase was achieved by folk being so scared they were going to lose their jobs suddenly upping their game? Probably in an unsustainable manner.
  • by rsilvergun ( 571051 ) on Wednesday February 18, 2026 @06:20PM (#65997950)
    Without destroying jobs unless you also have competition. I'm not sure about Europe but in America we stopped and forcing antitrust along so we have a very very little competition.

    When there is no competition companies just pocket productivity increases and turn it into big payouts for the largest shareholders. I want to be clear this is not your 401k. These kind of gains only go to the very top.
    • You're forgetting to consider the demand side. If the productivity gains are in occurring where the demand curve is flat (no more will be bought), then yes, jobs will be lost. If the demand curve is below the point of vanishing marginal returns (still curved), then the increased productivity is increased profit and no jobs are lost.

      Also note that the funds into which one's 401k may be invested are often "the largest shareholders", and thus would directly receive those gains.

    • by allo ( 1728082 )

      Why?

      Let's say you have x employees and create y units of whatever and sell it for z USD. Now you get an AI tool so your x employees create 2*y units and you sell them for 2*z USD. You doubled your profit without destroying a job. You could lay off half of your employees and make the same profit as before, but you likely want to increase your profit.

  • by clemare ( 6598318 ) on Wednesday February 18, 2026 @07:59PM (#65998070)
    Let sumarize: 1. Increase RAM prices 2. Increase SSD prices 3. Increase large discs prices 4. Increase Video Cards prices 5. Increase energy prices Everything to put those huge datacenters to enable people to make videos of "A fight with Maduro and Trump in a ring" stuff. So, that 4% increase in productivity, is automatically ereased because AI make everything an employee need much more expensive.... and it is actually negative.
  • by Jayhawk0123 ( 8440955 ) on Wednesday February 18, 2026 @11:01PM (#65998250)

    Difference between the US and EU-
    US-
    Employee (100) + AI (4) =104%
    this means we can do 100% with 96 employees! yay.. we saved on paying 4 employees! stocks rise. Employees are left with more work b/c AI isn't a direct replacement for a human in most cases, but a tool to help... are then are stuck doing more, so real productivity drops)

    EU-
    Employee (100) + AI (4) = 104%
    this means company has grown productivity by 4% and use that productivity gain into profit, better work life balance, more time, better quality, any number of things that those gains can be applied to.

    slightly different priorities... short vs long

    plus the focus on implementation is different as are the employment laws-
    EU-
    develop AI to enhance employees and in a lot of cases, it is because countries have regulations about firing people that make it more difficult to simply fire a chunk of your workforce for no reason... it's almost as if the workers have some protections.

    US-
    to replace employees/head count

    Just think about the phrasing common in the US- "head count" - that used to relate to animals on a ranch... not people. (i.e. X amount of heads of cattle).. dehumanizing your employees makes it easier to do this crap.

    • EU here, I think the difference is the speed with which they react. US: growth? Lots of sales? Hire hire hire hire. Economy goes bad? Fire fire fire fire.
      EU: let's hire a few and be more selective on the projects we take on. Build up a war chest. Goes bad? Let's fire a few and use the war chest for some internal projects.
  • Are they also measuring the quality of the output? Sure, you can replace a customer service call center with chatbots, and that may be a cheaper option for the bottom line, but in terms of the experience you are delivering to your customers - it will be much worse. Every time I have to go through several rounds of trying to convince a chatbot that they are not going to be able to help me with my problem and no I don't want to check my balance and that I need to speak to a human, by the time I get to speak t

Everybody likes a kidder, but nobody lends him money. -- Arthur Miller

Working...