# Depot AI Demo: Live Prompt Cheat Sheet (Notional Data)

Audience: E-9s, depot refurbishment. Upload the three CSVs, then run these in order.
All data is NOTIONAL/synthetic. No FOUO. Built to mirror WIP, supply/AWP, and schedule.

## Files
- `Depot_WIP_NOTIONAL.csv` (55 jobs on the floor)
- `Depot_Supply_AWP_NOTIONAL.csv` (parts holding up the AWP jobs)
- `Depot_Schedule_Throughput_NOTIONAL.csv` (12 months, 5 product lines)

## The "morning huddle" arc (build from simple to impressive)

1. Orientation (proves it read the data)
   "Summarize this WIP file. How many jobs, by status and by shop?"

2. The thing they care about first
   "List every job past its ECD that is not yet RFI. Sort by days past ECD,
    show platform, shop, and current status."

3. Connect two datasets (this is the Level 3 moment)
   "Cross-reference the AWP jobs in the WIP file against the supply file.
    Which jobs are stuck on a single backordered, sole-source OEM part?"

4. Turn data into a decision
   "I have 7 EXREP jobs. Given status, percent complete, and any parts holds,
    rank them by likelihood of making ECD and tell me which two need command attention."

5. The brief writes itself
   "Draft a 5-bullet production status brief for the 0700 stand-up: throughput
    vs plan, top 3 schedule risks, AWP impact, and one recommended action."

6. The mic-drop (predictive + their language)
   "Based on actual vs target TAT trends by product line, which line is
    drifting and what's the likely impact on next quarter's output?"

## Tips for the room
- Let one of them hand you the prompt. Ownership sells it harder than your slides.
- After step 5, point out: that brief took 8 seconds and zero PowerPoint.
- Keep the "human in the loop" point explicit: AI drafts, the artisan/chief decides.
- If GenAI.mil hiccups live, you still have the CSVs to talk through manually.