MW As bubble debate rages, a new study finds why companies are getting stuck by the AI productivity paradox

By Steve Goldstein

Companies do benefit from attracting AI workers, but only from lean organizations, study finds

Companies can boost their productivity from AI -- but only if implemented correctly, a study finds.

There feels something toppy about the rants from Palantir's Alex Karp and OpenAI's Sam Altman against short sellers - in the latter's case, entirely imaginary, as the ChatGPT maker is still a private company - given the meteoric stock-market gains so far from the AI trade.

On the other hand, the possibly revolutionary impact from artificial intelligence argues that maybe it is still early days, no matter how abrasively these chieftains behave.

There's a new study that has been published, on how AI actually helps - or doesn't - companies. It doesn't so much contradict a popular study from MIT, finding 95% of generative AI pilots at companies are failing, but states the AI implementations that do work see companies having their productivity boosted two to three times more than traditional IT.

It was hosted on arXiv, a popular site for research papers that have yet to be peer reviewed. Drawing on data from Revelio Labs, which itself culls data through LinkedIn, the study finds companies benefit by attracting workers with AI talent.

But, and this is the key point, not always - "hiring from flatter and more lean-startup-method-intensive firms generates significant productivity gains, whereas hiring from firms lacking these traits yields little benefit," the study says.

"These findings reveal that AI spillovers differ fundamentally from traditional IT spillovers: while IT spillovers primarily arise from scale and process standardization, AI spillovers critically depend on the experimental and integrative environments in which AI knowledge is produced," the research paper adds.

The lean startup method is an "approach emphasizing rapid experimentation, iterative development, and data-driven decision making." Just adding AI won't help a company until it engages in the data infrastructure, employee reskilling, and process reengineering needed to take advantage.

Granted, the paper doesn't touch on, much less settle, the AI bubble debate. Even real productivity enhancements have a price, and it could be that the hyperscalers and AI companies have overestimated the potential revenue that companies will be willing to spend on the technology.

But the paper does touch on how companies can get over what's been called the AI productivity paradox, that productivity initially falls before rising as AI is implemented.

The market

U.S. stock futures (ES00) (NQ00) pointed to a mixed start, following the worst day in a month for the S&P 500.

   Key asset performance                                                Last       5d      1m      YTD      1y 
   S&P 500                                                              6771.55    -1.73%  0.26%   15.13%   14.21% 
   Nasdaq Composite                                                     23,348.64  -2.55%  1.32%   20.91%   22.99% 
   10-year Treasury                                                     4.086      1.00    -3.50   -49.00   -34.90 
   Gold                                                                 3974.1     0.82%   -2.13%  50.57%   48.96% 
   Oil                                                                  60.99      1.04%   -2.10%  -15.14%  -15.17% 
   Data: MarketWatch. Treasury yields change expressed in basis points 

The buzz

The Supreme Court will hear arguments on whether President Donald Trump imposed tariffs legally under the International Emergency Economic Powers Act.

Democrats convincingly won key races including the governorships in Virginia and New Jersey and the New York City mayoral race in campaigns that all drew on concerns over affordability and President Donald Trump's stewardship.

Advanced Micro Devices $(AMD)$, which has an investor day coming up, reported sales both for the current quarter and the last quarter ahead of guidance.

Super Micro Computer $(SMCI)$, another company tied to AI, saw its shares retreat on margin concerns.

Pinterest struggled as the online photo sharing provider guided toward soft holiday-season revenue.

ARM $(ARM)$, Qualcomm $(QCOM)$, AppLovin (APP) and Robinhood (HOOD) report after the close of trade.

Best of the web

Deutsche Bank explores hedges for data center exposure as AI lending booms.

Some 26 billionaires were against him, but here were the two billionaires supporting Zohran Mamdani.

Carl Icahn, nearing 90, is still trying to right his empire.

The chart

The U.S. government is not releasing scheduled job-openings data this week, but numbers from recruitment firm Indeed - watched by Fed Chair Jerome Powell among others - point to slowing openings, note analysts at Goldman Sachs.

Top tickers

Here were the most active stock-market tickers as of 6 a.m. Eastern.

   Ticker  Security name 
   NVDA    Nvidia 
   TSLA    Tesla 
   AMD     Advanced Micro Devices 
   PLTR    Palantir Technologies 
   GME     GameStop 
   TSM     Taiwan Semiconductor Manufacturing Co 
   META    Meta Platforms 
   BYND    Beyond Meat 
   AAPL    Apple 
   NVO     Novo Nordisk 

Random reads

Think Starbucks is pricey? Here's a $1,000 cup of coffee.

Seems bright - a black hole flared with the light of 10 trillion suns.

A mascot for a soccer club known ironically as the Brewers has been relieved of his duties after stumbling around during a match.

-Steve Goldstein

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

November 05, 2025 06:46 ET (11:46 GMT)

Copyright (c) 2025 Dow Jones & Company, Inc.