A paired venture screening study recorded lower analyst time for AI-assisted reports
Source check: Source checked. Not independent scientific confirmation.
Finding: In a paired study of 20 applications at one investment firm, mean analyst time was 44.7 minutes for manual screening and 18.9 minutes for screening with GPT-6 Luna-assisted reports. Assisted time included verifying sources and making corrections. Seventy percent of reports needed at least one correction.
Technology: GPT-6 Luna; document interpretation and evidence-assisted scoring
Family GPT-6: profile · models page
Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.
- AI contribution
- DealsFlow used GPT-6 Luna to extract and score supporting information from application documents and retrieved sources. Analysts then verified and corrected the resulting reports.
- Test and result
- Each application was independently processed by both workflows, with a different analyst in each and roles counterbalanced. The analysts also adjudicated the factual support of 240 cited claims.
- Limitation / still unknown
- Two analysts affiliated with the participating fund completed the unblinded comparison. Each application was assessed by a different analyst in each workflow, so counterbalancing leaves analyst-by-workflow effects unresolved. The assisted time excludes unattended machine time. Decision quality and returns were outside this comparison.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://arxiv.org/abs/2610.32380v1 · https://arxiv.org/html/2610.32380v1
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: linked separately.
Review date: 21 Nov 2026. Review scope: arXiv v1 timing and factual-support protocol and results, Appendix A model configuration, funding and ethics notes.
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.