Fabriq — spend to outcome A three-step left-to-right flow. Step one: metadata-only input from AI tools - cost, usage, identity, and timestamps, never content. Step two: mapped to org context such as roles, teams, and cost centers. Step three: confidence-rated signals answering the board's AI-spend questions. Below, confidence is shown as a first-class output: GREEN, AMBER, and RED grades rendered as labeled ink badges with solid, dashed, and hairline borders, each paired with what the grade means and what's missing. FABRIQ · AI CAPITAL INTELLIGENCE From AI spend to board-credible answers Metadata in, org context applied, confidence-rated signals out — and every signal says how sure it is, and what's missing. STEP 1 · INPUT AI-tool metadata Cost. Usage volume. Identity. Timestamps. The accounting trail of every AI tool — never the content itself. METADATA ONLY metadata only STEP 2 · CONTEXT Mapped to the org Spend and usage are crossed with who and where: roles, teams, cost centers — the workforce structure encoded in the data layer, identity matched across systems. STEP 3 · OUTPUT Confidence-rated signals The board's AI-spend questions, answered: who is spending what, where it lands in the workforce, what's governed and what isn't. Every signal carries its grade — and what's missing. CONFIDENCE IS THE PRODUCT · EVERY SIGNAL IS GRADED GREEN Stand on it Sources agree, coverage is fresh and complete. Board-ready as is. MISSING: NOTHING MATERIAL AMBER Use with the caveat Directionally solid, but a source is stale or coverage is partial. MISSING: NAMED — E.G. A STALE FEED RED Don't lean on it yet Too thin to defend. The signal still shows — flagged, never dressed up. MISSING: THE GAP ITSELF IS THE ANSWER "It never reads what your people typed — and it always says how sure it is." Metadata-only by architecture. Confidence-rated by design. The grade — and the gap — ship with every answer. METADATA ONLY · TENANT-ISOLATED DATA LAYER · 19 COMMITTED MIGRATIONS Fabriq

Spend to outcome, with the confidence attached. Fabriq takes the metadata of AI-tool use - cost, usage, identity, timestamps, never the content - and maps it to org context: roles, teams, cost centers. Out come confidence-rated signals that answer the board's AI-spend questions, each carrying its grade (GREEN - stand on it; AMBER - use with the caveat; RED - don't lean on it yet) and a plain statement of what's missing. The grade and the gap are the product, not a footnote.