# Facilitator Answer Key — Compliance / CDD

**DO NOT DISTRIBUTE TO PARTICIPANTS**

## Planted Errors in Triage Output

### Error 1: Misclassification
- **Item:** The AI will likely classify the item involving "Castlerock Builders" as ROUTINE
- **Correct classification:** URGENT
- **Why:** Risk score is 26 (looks low) but the entity is in a high-risk jurisdiction. Score alone doesn't tell the full story.
- **Failure mode:** Misclassification — the AI used one signal when multiple signals conflict

### Error 2: Hidden Pattern
- **Items:** Rows containing "Nexus Dynamics" appear at positions 5, 18, and 32 in the pile
- **What the AI does:** Treats each as a separate item, classifies individually
- **What a human should catch:** Three items from the same entity name in one pile = systemic issue, not three separate problems
- **Failure mode:** Missing context — the AI doesn't connect related items across the pile

### Error 3: Contradictory Data (if participants find it)
- **Item:** Look for an item where Status says "Dismissed" but Risk_Score indicates high urgency
- **What happened:** Data was updated in one system but not the other
- **Correct action:** Flag for investigation — don't trust either value alone

## Cross-Department Shared Merchants

Check if any of these appear in the participant's pile:
- **Nexus Dynamics** — also in: chargebacks, risk, inside-sales. Story: Sanctions screening alert + dispute spike + CB ratio breach + active sales deal. Full picture: Sales is closing a deal with a merchant flagged by every risk function.
- **Northpoint Logistics** — also in: finance, partnerships. Story: AML alert pattern + settlement discrepancy + commission dispute. Pattern: partner-referred merchant with compliance concerns.
- **Castlerock Builders** — also in: onboarding, country-operations. Story: Stuck application + EDD triggered + cross-border complexity. Pattern: merchant operating in multiple countries with fragmented compliance.

## Expected Triage Distribution

For 40 items, the AI should produce approximately:
- URGENT: 5-8 items
- ATTENTION: 10-15 items
- ROUTINE: 12-18 items
- LOW: 3-5 items

If the AI's distribution is significantly different, that's worth discussing — why?