Skip to main content

Phase 2: Assisted Intelligence

Duration: Months 2-4 Goal: "System helps operators notice what matters"

Overview​

Phase 2 introduces AI-powered detection heuristics that highlight potential anomalies for operator review, establishing the human-in-loop workflow that ensures accuracy while building trust in the system.

Deliverables​

DeliverableDescription
AI HeuristicsInitial detection models for core use cases
Confidence ScoringProbability indicators on all detections
Event ListsPrioritized anomaly queues for operator review
Feedback LoopOperator validation workflow for model improvement
Alert PipelineNotification system for high-priority events

Success Criteria​

  • AI heuristics operational for 10+ use cases
  • Confidence scores calibrated against operator validation
  • Event management workflow functional
  • False positive rate quantified
  • Operator feedback incorporated into model updates

Key Activities​

Week 1-4: Model Development​

  • GroundingDINO + SAM2 ensemble deployment
  • Use case-specific model training
  • Confidence calibration
  • Event classification logic

Week 5-8: Operator Integration​

  • Event management interface
  • Feedback capture workflow
  • Alert pipeline implementation
  • Operator training sessions