Silence is not Clearance: Workflow-specific LLM Selection for Institutional AI — A Seven-model Study using the Private Intelligence Selection Function (PI-SF)
A workflow-specific study of language-model behaviour in a bounded institutional screening task. Seven candidate models were evaluated across seven synthetic cases, producing 49 canonical observations interpreted through the Private Intelligence Selection Function.
01 / Method
What the research does
The study freezes one screening workflow, seven synthetic finance cases and a qualitative Aligned / Partial / Misaligned classification. It examines governance-visible failure modes rather than producing a general model leaderboard.
02 / Limitations
What the evidence does not establish
The evidence is limited to one workflow and synthetic baseline. It is not global model qualification, proof of production readiness or proof of PI-OS efficacy.
03 / Institutional relevance
Why the research matters
Shows why plausible output is not sufficient evidence for institutional use and why model selection must be bound to a specific workflow, evidence boundary and acceptance standard.
04 / Full Paper
Read the complete paper
The complete first-party paper is available here without leaving the publication page. The reader loads only when requested.
© Little Black Journal LLP / Umran Nayani. All rights reserved. This paper is made available for reading and personal reference. No additional reuse licence is granted unless expressly stated.
05 / Citation & Provenance
Source record
- Suggested citation
- Umran Nayani. (2026). Silence is not Clearance: Workflow-specific LLM Selection for Institutional AI — A Seven-model Study using the Private Intelligence Selection Function (PI-SF). SSRN Electronic Journal. https://doi.org/10.2139/ssrn.7391379
- SSRN record
- SSRN 7391379
- Publication status
- Working paper published on SSRN. It is not described here as peer reviewed.
- First-party provenance
- Prepared from the verified public source record and the company’s bounded research archive. A first-party copy of the full paper is preserved on this site, with the SSRN record retained as independent publication provenance.
- Rights
- © Little Black Journal LLP / Umran Nayani. All rights reserved. This paper is made available for reading and personal reference. No additional reuse licence is granted unless expressly stated.
07 / Application
From research to institutional work
The paper is evidence of completed research, not a claim that every method has been deployed in every environment. Explore the relevant capability routes for current scope and commercial boundaries.