“Generative AI has its uses, but it hardly surprising that most business have struggled to find return on their investment, given the massive reliability issues that continue to plaguage it.”
Our reading, the sentence that was checked:
Most businesses have struggled to get return on generative AI investment because of its ongoing massive reliability issues.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Turns out Generative AI was a scam, paragraph 13.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-turns-out-generative-ai-was-a-scam-7fdefa-a2
We recorded 8 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-28. The source may have been edited after 2026-02-23. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Causal claim
- Deadline, as the speaker wrote it
- None stated. This site does not add one.
- Deadline state when assessed
- No horizon
- Deadline state on 2026-10-29
- No horizon (the release’s reference date)
From the speaker
- most business have struggled to find return on their investment
- given the massive reliability issues
- continue to plaguage it
Our interpretation
- Read 'most' as more than half of businesses investing in generative AI, including pilots.
- Read 'struggled' as difficulty realizing or demonstrating benefits sufficient to justify investment, not necessarily permanent losses.
- Read reliability failures as a material cause of that difficulty, rather than necessarily its sole cause.
- Use calendar 2025, the latest completed annual reporting period before publication, for the ongoing business experience described.
- This is a population-level causal claim, not a financial figure for an identified company; metric_scope is null.
Counts as supported if: Representative evidence shows that more than half of generative-AI investors struggled to realize or demonstrate returns, and links reliability failures materially to that difficulty.
Counts as contradicted if: Comparable representative evidence shows that most investors obtained returns without such difficulty, or credible causal evidence rules out reliability failures as a material explanation.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Does the evidence support the cause?
Outcome: Unresolved
The record was checked, and the evidence available by the cutoff does not settle it.
Scored on a plain reading. The speaker named no measure for this claim, so it was judged on the reading an informed reader would take, stated below under our interpretation. Both reviews had to find that reading fair.
- Evidence as of
- Deadline state when assessed
- No horizon
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
NANDA's January–June 2025 study reported little measurable return for most sampled organizations, but emphasized implementation approach, learning and workflow fit rather than model quality [e1]. Google Cloud's September 4, 2025 announcement reported that 74% of surveyed executives achieved generative-AI ROI within the first year; its sample covered enterprises with deployed systems [e2]. Deloitte's October 22, 2025 research described slow, hard-to-measure returns across AI more broadly [e3]. These different populations and return measures do not establish whether most generative-AI investors struggled during 2025, or whether reliability failures materially caused that difficulty [e1, e2, e3].
Limitations
- NANDA's interviews, conference survey and public-initiative review do not establish prevalence among all generative-AI investors [e1].
- Google's deployed-adopter sample excludes unsuccessful non-deployers, and reported ROI does not establish that returns came without difficulty [e2].
- Deloitte covers AI generally; none of the cited studies isolates reliability's causal contribution to the claimed majority difficulty [e1, e2, e3].
Evidence (3 items)
Context only: Project NANDA researchers, hosted by MLQ,
determined by approach.
Why it matters here: The January–June study reports little measurable return for most sampled organizations, but emphasizes learning, workflow fit and implementation approach. Its interviews and conference survey do not establish population prevalence or isolate reliability's causal effect.
- Publisher
- Project NANDA researchers, hosted by MLQ
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: Google Cloud,
within the first year.
Why it matters here: This is material counterevidence to a blanket failure-to-obtain-returns reading. The vendor-sponsored survey covers enterprises with deployed generative AI, excludes unsuccessful non-deployers, and does not establish whether obtaining returns involved struggle.
- Publisher
- Google Cloud
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Context only: Deloitte,
hard to measure.
Why it matters here: The survey documents delayed and difficult-to-measure returns, with organizational change, integration and attribution among the explanations. It covers AI broadly and does not establish the claimed generative-AI reliability effect.
- Publisher
- Deloitte
- Kind of source
- Independent measurement
- Describes events of
- Public from
What would settle it
A representative study of 2025 generative-AI investors, including failed pilots and deployed systems, with comparable measures of realized returns and difficulty obtaining them, plus evidence separating reliability's causal contribution from integration, organizational and other factors.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Turns out Generative AI was a scam
Generative AI has its uses, but it hardly surprising that most business have struggled to find return on their investment, given the massive reliability issues that continue to plaguage it.
- Where
- paragraph 13
- Published
- 2026-02-23
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-turns-out-generative-ai-was-a-scam-7fdefa-a2: Current. Outcome: Unresolved. Evidence as of 2026-10-28. Public date not recorded.
Recorded events
Assessment publishedBy this site
First assessment published after both reviews and the final-text check passed.
About assessment marcus-turns-out-generative-ai-was-a-scam-7fdefa-a2.