🧠 Master AI Decision Framework
Enhanced decision engine — rules, matrix, confidence thresholds, risk gates & owner approval flows
Master Decision Question
"What is the next highest-probability legitimate action that increases realised contribution?"
Every AI agent, every decision, every action must answer this question before proceeding.
Primary Operating Question
What is the next highest-probability legitimate action that increases realised contribution?
Zero-Capital Constraint
Every decision must operate within zero-capital boundaries. No speculative inventory. No unsecured commitments.
Truth-First Principle
All decisions based on verified, timestamped data. Stale data triggers revalidation before action.
Contribution Maximisation
Every action evaluated by expected realised contribution. Highest-probability path always preferred.
Risk Isolation
Each item/opportunity processed independently. One failure cannot cascade to block others.
Owner Sovereignty
All decisions above threshold require owner approval. AI recommends, owner decides on high-value actions.
Decision Authority Hierarchy
Fully Autonomous
AUTOLow-risk, high-confidence, below ₹5,000 threshold. AI executes immediately.
AI Recommends → Auto-Execute
NOTIFYMedium risk, confidence ≥ 0.75, ₹5,000–₹25,000. Logged and executed with 5-min owner notification window.
Owner Approval Required
APPROVEHigh risk or ₹25,000–₹1,00,000. AI prepares full recommendation. Owner must approve within 2 hours.
Owner Direct Decision
OWNER ONLYCritical risk or > ₹1,00,000. AI provides analysis only. Owner makes final call. No auto-execution.
Confidence Threshold Matrix
| Confidence | Risk LOW | Risk MEDIUM | Risk HIGH | Risk CRITICAL |
|---|---|---|---|---|
| ≥ 0.90 | ✅ Auto-Execute | ✅ Auto-Execute | ⚠️ Notify Owner | 🔴 Owner Only |
| 0.75–0.89 | ✅ Auto-Execute | ⚠️ Notify Owner | 🔴 Owner Approval | 🔴 Owner Only |
| 0.60–0.74 | ⚠️ Notify Owner | 🔴 Owner Approval | 🔴 Owner Only | ❌ Reject |
| < 0.60 | 🔄 Revalidate | ❌ Reject | ❌ Reject | ❌ Reject |