“AI generates a lot of output (which fits many people’s informal notion of productivity), but it hasn’t yielded much in the way of RoI for many companies (per studies from MIT, McKinsey, Bain, and many others), and hasn’t materially changed GDP.”
Our reading:
AI produces lots of output but has not yielded much return on investment for many companies.
Assessment: Supported
What happened
Was it true for the period described?
Outcome: Supported
The evidence matches what they said.
Judged on a plain reading. They named no measure, so we used the reading a typical informed reader would take. It is spelled out under “How we judged it”.
Evidence up to . Checked .
Why
The evidence supports limited realized financial benefits for many companies during June 8, 2025–June 7, 2026 [e1, e2]. PwC surveyed CEOs from September 30 to November 10, 2025; 56% reported no significant financial benefit from AI to date [e1]. Deloitte surveyed executives at organizations with working AI implementations from August 15 to September 5, 2025; 15% of generative AI users and 10% of agentic AI users reported significant, measurable ROI [e2]. Dun & Bradstreet's May 4, 2026 release reported at least some measurable ROI for 60% of businesses, including broad or strong returns for 24%; these positive results remain compatible with limited benefits for many others [e4]. Support concerns meaningful financial payoff, without establishing a common audited net ROI ratio [e1, e2, e4].
Caveats
- The surveys use different populations and self-reported measures; they do not establish comparable net returns after all costs [e1, e2, e4].
- Not reporting significant ROI does not establish zero returns or eventual investment failure [e2].
- Bain's April 2026 cost-savings figures do not directly measure investment returns [e3].
- The evidence does not directly quantify aggregate AI output volume or establish the separate GDP assertion [e1–e4].
Evidence (4 items)
Supports the claim: PwC,
56%
Why it matters: PwC surveyed 4,454 CEOs between September 30 and November 10, 2025. The large group reporting no significant financial benefit supports the ordinary financial-payoff reading, although the sample is not restricted to mature AI deployments.
- Publisher
- PwC
- Kind of source
- Independent measurement
- About events on
- Published
Supports the claim: Deloitte,
measurable ROI
Why it matters: The survey covered 1,854 executives at organizations with working AI implementations, collected August 15–September 5, 2025. It reports significant current ROI for 10% of agentic AI users and 15% of generative AI users, supporting limited meaningful returns for many adopters without establishing zero returns.
- Where
- Section 'Generative AI vs agentic AI: Different tools, different timelines', agentic AI paragraph
- Publisher
- Deloitte
- Kind of source
- Independent measurement
- About events on
- Published
Context only: Insurance Journal / Bloomberg,
reductions of 10% or less.
Why it matters: Bain's April survey covered executives at 951 large companies and found savings often below expectations. Cost reductions are not an investment-return ratio, and savings below 10% can still be valuable, so this is contextual evidence rather than decisive proof of low ROI.
- Publisher
- Insurance Journal / Bloomberg
- Kind of source
- News report
- About events on
- Published
Context only: Dun & Bradstreet,
measurable ROI
Why it matters: This survey announcement reports early-2026 results from a quarterly study of 10,000 businesses across 32 countries. It counters a blanket claim that AI produces no returns, but positive returns for some businesses remain compatible with limited returns for many others.
- Publisher
- Dun & Bradstreet
- Kind of source
- Independent measurement
- About events on
- Published
How we judged it
- Kind of claim
- Observation
- Deadline
- None given. We don’t add one.
Conditions
- for many companies (their words)
Where we had to interpret
- Assess the supplied normalized proposition about enterprise returns. The GDP assertion is outside this claim.
- Read 'many companies' as a substantial group of companies, without requiring all companies or a particular majority threshold.
- Read 'much in the way of RoI' as meaningful realized financial benefit from AI investment. Marcus does not specify a formal return ratio or hurdle rate.
- Treat the output language as a qualitative contrast between producing work and realizing financial benefits, without inventing a volume threshold.
- For this ongoing observation, use the latest annual observation window ending at publication: June 8, 2025 through June 7, 2026. Survey results describe conditions when respondents were surveyed, not when reports were published.
- This is a qualitative claim about the prevalence of limited returns across companies, not a financial amount or accounting metric for a particular company; metric_scope is null.
Right if: Credible enterprise studies within the target period identify a substantial group of AI users reporting little or no meaningful realized financial benefit, despite practical use of AI.
Wrong if: Comparable studies establish that limited realized financial benefits are confined to an exceptional or negligible group of AI-using companies. Evidence that some companies succeed does not alone contradict 'many companies'.
We wrote this test after the outcome was known.
We found 15 claims in this post. See them all.
Full record
Source and dates
Marcus on AI, Slop, productivity, and why the AI-fueled world is going nowhere mighty fast, paragraph 1.
- Said
- We read the source
- live page, full text, public
- Evidence up to
- Checked
We read this wording on 2026-10-28. The post may have been edited after 2026-06-07.
Where it was said (1)
Said once in the posts we read.
: Marcus on AI, Slop, productivity, and why the AI-fueled world is going nowhere mighty fast
AI generates a lot of output (which fits many people’s informal notion of productivity), but it hasn’t yielded much in the way of RoI for many companies (per studies from MIT, McKinsey, Bain, and many others), and hasn’t materially changed GDP.
- Where
- paragraph 1
- Published
- 2026-06-07
- We read it
- , live page
- Access
- full text, public
History
Versions
- Current: Supported. Evidence up to 2026-10-28.
What happened to this record
PublishedBy us
Published after both AI reviews and the final check passed.