Slide 1 of 15
⌂ Presentations
Slides — PDF
Slide 1 of 15
Flame AI — Collective System Design in Healthcare. Clement Tan Yong Wei and Dr. David S. Cochran, Purdue University Fort Wayne Systems Engineering Center.
The Challenge: a powerful methodology, hard to scale — expertise-bound, time-consuming, expensive; Franciscan Health (18,000 team members) as the scaling test bed; research hypothesis.
What is Collective System Design? Aligning Leadership/Tone, Thinking, Structure and Work into one sustainable system.
The Flame Model — four nested layers, Tone to Work, with Diagnosis and Design arrows.
An unambiguous design taxonomy: FR (Functional Requirement), FRm (FR Measure) and PS (Physical Solution).
The Twelve Steps grouped into Problem (1–3), Design (4–6), Refinement (7–9) and Implement (10–12).
How Flame AI is built: OpenWebUI hosted VM with system prompt, RAG knowledge base, OpenAI LLM and Google PSE live search.
XML-gated viewpoints: siloed instructions and context checkpointing in the system prompt.
Sample problem statement: medical-surgical supply chaos, described by a nurse (illustrative, not a real deployment).
Viewpoints 1–3 (Problem): Root Cause Analysis, Collective Agreement, Flame Analysis — and the 5th-why root cause.
Viewpoints 4–6 (Design): use cases, the main FR0 'Enable rapid supplies access', and FR0 measures with thresholds and goals.
Viewpoints 7–8 (Solutions): physical-solution alternatives and weighted-table selection; leading PS is color-photo labels.
Viewpoints 9–10 (Decomposition): axiomatic decomposition and the coupled FR–DP matrix driven by DP5.
Viewpoint 11: automated decoupling — best-effort lower-triangular reordering and a revised uncoupled identity matrix.
The future of system design: what Flame AI demonstrated (speed, depth, rigor), the roadmap, and the 1979 IBM quote.