DOCS/IS_overview.md

Process-Aware AI: Project Overview

1. Goal & Vision

The primary goal of Process-Aware AI is to optimize Greige Issuance in textile manufacturing. The system aims to replace static, "one-size-fits-all" norms with dynamic, data-driven recommendations that minimize waste (surplus fabric) while ensuring order fulfillment (preventing shortfalls).

2. The Problem

• Static Norms: Traditional planning uses fixed rules (e.g., "Always add 5% process loss").
• Inefficiency:
  • Over-issuing: Wastes raw material (Cotton, Tencel, etc.) and increases deadstock.
  • Under-issuing: Causes "shortfalls" (orders not fulfilling demand), requiring expensive reprocessing or urgent small-batch productions.
• Lack of Feedback: Planners rarely see if their "buffer" was actually needed or if it caused waste.

3. The Solution

We have built an AI-Driven Decision Intelligence System that:

1. Analyzes History: Looks at every past order for a specific article.
2. Evaluates Outcomes: Did X% reservation succeed? Did Y% fail?
3. Recommends Precision: Suggests the exact reservation needed to succeed based on historical performance, not just a guess.

4. Key Achievements (Current Status)

• ✅ Data Pipeline: Successfully ingesting Sale Orders, Norm Rules, and Production Data.
• ✅ Interactive Dashboard:
  • Global Views: Trends, Norm deviations.
  • Article Drill-down: Deep dive into specific fabric behaviors.
• ✅ "Outcome-Based" AI Engine:
  • Moved away from simple averages (which are skewed by outliers).
  • implemented Success-Based Logic: Recommends the median reservation of successful orders.
  • implemented Failure-Safe Logic: If an article has 0% success history, analyzes why it failed and recommends a robust buffer to ensure future success.
• ✅ Scenario Playground: Allows planners to simulate "What if we changed the norm to X%?" to see financial and operational impact.

5. Value Proposition

• Reduce Waste: Identify articles where standard norms (e.g., 5%) are too high compared to actual needs (e.g., 2%).
• Prevent Failures: Identify "Same-Norm" articles that frequently fail and require higher buffers.
• Standardization: Reduce dependency on individual planner intuition by providing a standardized, data-backed baseline.