AI Decision Governance Demo

This demo shows how Krapheno governs AI agent decisions in real time. Each decision proposed by an AI tool is evaluated against declared policy rules and assigned a verdict before execution.

What this demo shows

• AI agent proposes a campaign action (budget change, audience expansion, creative swap)
• Krapheno evaluates the proposal against policy thresholds (MIC constraints)
• Verdict issued: ALLOW, ESCALATE, or BLOCK
• Decision recorded permanently in SmritiTree — append-only, hash-chained ledger
• Performance outcome linked back to the decision for causal traceability

Real governance example

Campaign: blissedpilgrims_gov_001. AI proposed 30% budget increase on day 1. MIC-01 fired — confidence 78% below threshold 95%. MIC-02 fired — sample 55 below threshold 1000. Verdict: ESCALATE. Hold applied. Campaign improved 44% in CPV over 4 days with zero intervention. Every decision permanently inscribed with SHA-256 record hash.

What this proves

• Runtime policy enforcement — governance happens before execution, not after
• Causal purity — holding a decision preserves attribution clarity
• Platform recommendation governance — AI governance can evaluate and reject platform actor suggestions on evidence grounds
• Permanent accountability — every verdict is auditable via decision_id and record_hash