How the GCI Conviction Engine Works
Transparent methodology. 4 engines, 25 analysis agents, 10 report sections. Every claim source-rated. Every verdict structured.
Not a chatbot. A conviction engine built by GCC practitioners.
GCI does not generate free-form AI output. It runs a structured analytical pipeline designed by investment professionals with over a decade of real GCC deal experience. The conviction framework encodes judgment from the field, not prompts from the internet.
Five analytical stages, four independent engines.
Every report passes through the same five stages. Each stage runs across four independent frontier AI engines, with disagreement surfaced rather than hidden.
Identifies every hidden assumption in the deal thesis, so nothing the outcome depends on stays implicit.
Maps how market, regulatory, and operational variables interact, rather than scoring each one in isolation.
Chains evidence and dependencies into a single reasoning spine, so every conclusion traces back to its inputs.
Deliberately attacks the thesis with its strongest counter-arguments before any verdict is reached.
Produces the 10-section institutional-format report, with three financial scenarios, required diligence actions, and an evidence tier on every claim.
Every claim carries a named evidence tier.
No claims without attribution. Every numeric claim and factual assertion in a GCI report is tagged with one of the engine's evidence tiers, the same tags you can see on any published report on this site:
Independently confirmed by an official regulatory, legal, or primary source
Grounded in a specific statute, regulation, or licensing requirement, cited by name
Cited to a specific external source: established media or industry research
Stated by a counterparty or referenced source, and flagged as such
Model estimate built on stated assumptions, with the methodology disclosed
An explicit assumption, made visible so the reader can challenge it
Frequently asked questions
What are the 5 stages of the GCI Conviction Engine?
Stage 1 Assumption Extraction identifies every hidden assumption in the deal thesis. Stage 2 Cross-Variable Synthesis maps how market, regulatory, and operational variables interact. Stage 3 Linkage Mapping chains evidence and dependencies into a single reasoning spine. Stage 4 Contrarian Pressure Test deliberately attacks the thesis with its strongest counter-arguments. Stage 5 Evidence-Chain Report produces a 10-section institutional-format report where every claim carries an evidence tier.
Which AI engines does GCI use?
GCI cross-checks every Conviction Report across four independent frontier AI engines, with each engine's analysis weighed against the others and disagreement surfaced rather than hidden. The conviction engine is the judgment layer designed by the co-founders, Hemant Agrawal and Gaurav Agarwal; the engines are the research tools.
What do the 6 evidence tiers mean?
VERIFIED: the claim is confirmed by an official regulatory, legal, or primary source. REPORTED: the claim appears in mainstream media or an established industry report. STATED: the claim was stated by a counterparty or referenced public source. ESTIMATED: the claim is a model estimate based on stated assumptions. ASSUMED: the claim is an assumption made explicit to the reader. LEGAL: grounded in a named statute, regulation or licensing requirement. Every numeric claim and every factual assertion in a GCI report carries one of these six tags.
What are the 10 sections of a Conviction Report?
1 Executive Summary with the verdict on page one. 2 Investment Overview. 3 Market Context and Competitive Landscape. 4 Regulatory and Compliance Framework. 5 Financial Model with three scenarios (downside, base, upside). 6 Risk Matrix with HIGH, MEDIUM, LOW ratings. 7 Location Analysis. 8 M&A Comparables and Valuation Context. 9 Investor Suitability Assessment. 10 Verdict and Required Diligence Actions. Evidence-tiered citations and the disclaimer sit in the appendices. Every Conviction Report covers these ten areas; section order varies by mandate type.