AI Hierarchy Explained
How the five-team AI stack gates every trade — and where the Manager AI holds the final veto
No single AI has unchecked authority. Five specialist teams own one slice of the trade lifecycle each; the Manager AI validates all five gates before any ticket ships.
Market Intel AI
Reads the market type, indices, liquidity, volatility, market health, time-of-day, and the economic event calendar — answers "are conditions tradable?"
Option Intel AI
Decodes the option chain — open interest, put-call ratio, implied volatility, Greeks, and expiry structure — to find the right strikes and strategies.
Decision AI
Computes probability, confidence, and risk. Position size is half-Kelly by default; capital allocation scales with conviction.
Execution AI
Validates readiness, gates on slippage tolerance, monitors the live ticket, and manages exits — hard dashboard, soft target, or market type break.
Learning AI
Reviews every closed trade, attributes P/L, detects repeat patterns, and feeds improvements back into the stack — namespaced, never cross-contaminates other markets.
Manager AI
Final authority. Validates all five gates and can optionally invoke an LLM veto to reject any signal that fails quality bar.
The engine runs as a five-team hierarchy. Each team owns one slice of the lifecycle and reports up to the Manager AI — which holds the final veto before any ticket routes to the broker.
Reads the market market type, liquidity, volatility, and the economic event calendar. Answers the only question that matters first — are conditions tradable at all?
Decodes the option chain — open interest, put-call ratio, implied volatility, and the Greeks — and proposes strikes and strategies that fit the market type.
Computes probability and confidence. Position size comes from a half-Kelly model; capital allocation scales with conviction and risk limits.
Validates readiness, gates on slippage tolerance, monitors the live ticket, and manages exits — hard dashboard, soft target, or market type break.
Reviews every closed trade, attributes P/L to the gate that drove it, detects repeat patterns, and feeds improvements back into the stack — namespaced so Hero Zero learning never bleeds into Forex, India, or Crypto.
Holds the veto. Validates all five gates, optionally invokes an LLM second opinion, and either signs the ticket through to the broker — or kills it. No ticket ships without the Manager's signature.