OCCAM’S FORENSIC JURY · Methodology · InsightfulAgents.AI LLC

Follow the Cash. Not the Narrative.

A Trademarked Four-Lens Forensic Analysis Methodology

Most “AI stock analysis” collapses to one model’s opinion. The OCCAM’S FORENSIC JURY runs every public company through four independent AI models — each with a different forensic mandate, each operating on the same source data, each producing its own verdict before the others see it. Disagreement is amplified, not suppressed.

01The Discipline

Wall Street process. Audit trail Wall Street doesn’t publish.

Cash-flow primacy. Disclosure-language analysis. Narrative versus numbers. External cross-references. Those four steps are the analytical sequence professional equity research has used for decades. The discipline is the constant. What changes is the visibility of the work.

Most public-facing AI stock analysis collapses those four steps into a single model’s confident summary. The discipline disappears into the synthesis. The Forensic Jury reverses that order. Each step runs as an independent lens with a different forensic mandate. Each step produces its own verdict before seeing the others. The synthesis is the last step, not the only step. If three lenses agree and one dissents, the dissent is the most informative output of the run.

The result is a process you can audit. Not a recommendation. Not a black box. Every assumption is sourced. Every dissent is preserved. Every override is documented. The same analytical discipline that runs inside research desks, run as a methodology with a public audit trail.

02The Four Independent Analysts

Occam-the-Auditor Primary Lens

The Books

Reads through the cash flow truth. Mandate: GAAP-vs-cash divergence, working-capital strain, capex commentary, dividend coverage math, balance-sheet integrity, refinancing risk. The Auditor ignores tone and rhetoric and refuses to give management the benefit of the doubt without numerical evidence.

Occam-the-Architect Primary Lens

The Words

Forensic linguistics on disclosures. Mandate: hedging-language detection, tone shifts between prepared remarks and Q&A, precision drift across quarters, the gap between confidence claimed and confidence supported. Executives are trained to sound confident — which is why pattern deviations matter.

Occam-the-Storyteller Secondary Lens

The Absences

The narrative arc and what is left out of it. Mandate: notable omissions (topics that should have been discussed but weren’t), narrative gaps, cherry-picked metrics, year-over-year comparisons that suppress unfavorable periods. The Storyteller treats silence as data.

Occam-the-Sentinel Secondary Lens

The Outside World

Real-time cross-references. Mandate: claim-vs-filing reconciliation, industry context (are competitors saying the same thing?), market-data sanity checks, regulatory and news context that closed-data models cannot see. The Sentinel cites sources.

03How the Verdicts Merge

Weighted-Mean Merge with FLAGGED Override

Each lens produces an independent verdict on a five-point scale (STRONG_SELL, SELL, HOLD, BUY, STRONG_BUY) plus a confidence rating. Verdicts are scored, weighted, and merged via weighted-mean with a deterministic rounding rule. The Auditor and the Architect carry more weight than the Storyteller and the Sentinel — cash flow integrity and disclosure language carry more decision-weight than narrative architecture or external context.

One override sits above the math: any single lens returning FLAGGED forces the merged verdict to FLAGGED at HIGH confidence regardless of the others. The Jury is built to amplify dissent, not suppress it.

Merged confidence reflects the spread between lenses. When one lens dissents at MEDIUM or HIGH confidence, or any lens votes at LOW confidence, the merged verdict cannot claim HIGH confidence. The methodology is structurally honest about its own uncertainty.

04What the Jury Is Not

Not voting. The lenses are weighted, not equal. Verdicts are scored, not counted.

Not consensus optimization. Disagreement is preserved and surfaced — explicit conflict notes appear in every dossier where the lenses diverge.

Not majority rule. A single FLAGGED verdict overrides the others. Forensic dissent is treated as more informative than forensic agreement.

Not opinion aggregation. The lenses are not asked “what do you think?” — they are asked to apply specific forensic methodologies to the same source data.

OCCAM’S FORENSIC JURY™ is a trademarked methodology of InsightfulAgents.AI LLC. The four-lens architecture, the persona designations (Auditor, Architect, Storyteller, Sentinel), the weighting structure, the FLAGGED override, and the merge rule are part of the trademarked methodology. The specific persona prompts and instruction-tuning behind each lens are trade secrets and are not published. White-label customers receive verdicts, audit trails, and rendered dossiers — never the underlying prompts.