Stanford DEL — Retracted
Claim reviewed: Stanford's Digital Economy Lab Enterprise AI Playbook documents bimodal ROI: 12% of enterprise AI deployments clear 300%+ ROI with 40-60% operational cost reductions; 88% sit at or below break-even.
First scheduled review of this record, pulled forward from the 180-day cadence after a batch-2 editorial re-review on 10 Jun 2026 flagged the figure (AM-029 verification).
The claim text does not appear in the source. The record asserts: “Stanford’s Digital Economy Lab Enterprise AI Playbook documents bimodal ROI: 12% of enterprise AI deployments clear 300%+ ROI with 40-60% operational cost reductions; 88% sit at or below break-even.” The full text of the cited PDF was extracted and searched on 10 Jun 2026. It contains no bimodal ROI distribution, no 12% cohort, no 300%+ ROI band, and no 88% at-or-below-break-even body. Every “88” in the document measures something else: organisational AI adoption (“88% of organizations use AI in at least one function”), one single-case coding-productivity gain, a data-unlock share, a page number, and a breach-cost figure ($4.88M). The terms “bimodal” and “break-even” (as an ROI cohort) do not occur at all.
The source structurally cannot produce the claimed finding. The playbook studies 51 successful deployments by design (41 organisations, 9 industries). A success-only sample cannot yield a deployment-failure or ROI-realisation distribution. This is not a revision or a methodology dispute; the archived sentence was never in the archived document, including in the Wayback snapshot logged with the record on 19 Apr 2026.
Where the figure actually comes from. The verified figure carrying the same numerals is IDC research with Lenovo, reported by CIO.com on 25 Mar 2025: 88% of observed AI proof-of-concepts do not make the cut to widescale deployment, and roughly 4 of every 33 POCs graduate to production (about 12%). That is a pilot-to-production graduation rate — a different metric from an ROI distribution. The archived claim is best explained as that IDC finding fused with the Stanford DEL’s name and an invented ROI methodology; the full corpus forensics are at docs/editorial/stanford-1288-exposure-map-2026-06-10.md.
Verdict: Retracted. Under the archive grammar, “weakened” describes a claim eroded by counter-evidence; this claim failed existence verification against its own source. Downstream editorial claims that rested on this record carry their own dated corrections (AM-029 Not holding; AM-024, AM-031, AM-040, AM-042, AM-129, AM-132, AM-201 Partial, all 10 Jun 2026). The record and this memo stay public per the no-silent-removal rule.