Methodology · v2.10

ARAI Readiness

AI investment / readiness composite for {legal_name}. EXTRACTION RULES: return type="composite_signal" with one entry per dimension, each 0-100 + source_url. Required sub-dimensions (use these exact names): (a) ai_patent_intensity — 0-100 mapped from USPTO G06N (machine-learning) patent volume in the last 5y (0 → 20, 1-10 → 40, 11-100 → 65, 101-1000 → 85, 1000+ → 95); (b) ai_investment_tier — 0-100 mapped from public commitments / R&D dedicated to AI (e.g. a named AI platform / framework launch, multiple announced AI-related acquisitions, multi-billion R&D allocation per 10-K) — 95 for mega-cap with named AI platform + 5+ AI acquisitions, 75 for active deployment + 1-3 acquisitions, 55 for early stage, 30 for no public commitment; (c) ai_leadership_rank — 0-100 from Stanford AI Index / McKinsey AI Index / OECD AI Observatory rankings (top-5 → 100, top-25 → 85, top-100 → 70, listed → 50, unlisted → 25); (d) ai_product_deployment — 0-100 from named consumer products / APIs shipped with AI (a flagship AI platform spanning device + cloud + assistant = 95; major AI features in flagship products = 80; limited AI features = 55; none = 25). Each signals[].value MUST be a real assessment based on cited evidence (NOT a raw patent count, NOT a 0-or-1 boolean). DO NOT emit a single 0-100 scalar — the composite is the whole point. Server applies composite_signal_score. If {legal_name} has NO public AI signal at all (no G06N patents, no AI Index ranking, no named AI platform / acquisition / R&D commitment) return type="absent" reason="no_company_specific_data" — do NOT fabricate or estimate an AI-readiness score; the absence is a clean not_applicable, not a fabricated bottom.