Current AI models can't trade directly - they're not trained for real-time financial decision making. I've tested many variants of this early on and they consistently fail. The right approach: AI works like professional traders work 1. Market Intelligence - AI monitors signals, identifies opportunities (momentum shifts, catalysts, pattern formations) 2. Strategy Formation - When opportunities emerge, AI generates structured trade plans in code. - Entry thesis and zones - Position sizing based on risk - Multiple take-profit levels - Stop losses (technical + volatility-based) - Invalidation conditions 3. Algorithmic Execution - Orders executed programmatically, scaled in/out, stops trailed dynamically 4. Active Monitoring - AI continuously assesses if thesis remains valid, proposes adjustments (tighten stops, take early profits, exit on invalidation) Key separation: AI analyses and strategises. Algorithms execute. AI stays in the loop via heartbeats to adapt.