📖 How to read this playbook (and how I read the FigJam)
Each page below is one decision point from the top-level map — one "play." The left pane recreates the FigJam tree exactly as drawn (same branch labels, same terminal boxes); the right pane is the same tree flattened into path → terminal state rows, so every route to an outcome can be checked line by line.
How I read each board: start at the entry decision, walk every arrow to a terminal box, and record the full condition path — then capture the sticky notes separately, because they answer a different question: not "what's the rule?" but "what do we do when the field is MISSING?" (look it up, assume a default, or follow up). Where two arrows overlap and the wiring is genuinely ambiguous, I did not guess — those became the questions at the bottom of the page.
Dashed-underlined terms open a ? glossary dialog — the shared vocabulary (UWing period, Knox Box, KYC, P(F), RCE…) captured while reading. Each page also notes where an agent helps: the exact spots a decision would stall on vague or missing info, and whether the unstick is a lookup, a derivation, the board's own assume-and-verify default, or one scoped question.
Fire Simulation
When the wildfire model says Fail / Do-Not-Write, the play is a documented manual review — “legacy underwriting” — that can still reach a quote.
What makes this board special: it is the playbook's HUMAN-OVERRIDE protocol. The model can be wrong at the edges, so a Fail doesn't end the conversation — it opens a hand-check of the model's own inputs (access, vegetation, slope, 7A, neighbor distance), and even a willing-to-mitigate conversation with the client. And a successful override doesn't just quote — it produces a mitigation plan for the broker.
| Path | Terminal state |
|---|---|
| Model verdict Fail / Do-Not-Write, no override pursued | DECLINEDecline |
| Override: legacy UW review — map the risk on Google Maps/Zillow | NOTEhand-check of the model inputs, below |
| Access → ingress/egress limited · responder turn-around insufficient | DECLINEDecline |
| Access → multiple points · turn-around sufficient | OK / QUOTEContinue to quote |
| Vegetation heavy → client willing to mitigate greater clearance? | REQUIREYes → quote · No → decline |
| Vegetation moderate/light → slope gentle/level | OK / QUOTEQuote |
| … slope steep (clearance needs increase) → willingness check | REQUIREYes → quote · No → decline |
| 7A compliant → yes | OK / QUOTEproceed via slope check |
| Min distance to neighbors adequate / too close | DECLINEContinue / Decline |
| Any successful quote path | REQUIREQuote + determine a preliminary mitigation plan to discuss with the broker |
- The legacy-review evidence is gatherable before the underwriter opens the file: maps/satellite imagery for access and vegetation, parcel data for slope and neighbor distance — an agent assembles the override dossier so the human check takes minutes, not an afternoon.
- The "willing to mitigate?" step is the playbook's one crisp outbound message on this board — an agent drafts it with the specific clearance ask already computed.
- A quote here must ship WITH a preliminary mitigation plan — a document an agent can draft from the same checks that justified the override.
- Is the model's output the same P(F) the Roof/Siding boards band on, or a separate Fail / Do-Not-Write verdict?
- The “7A Compliant → No” arrow's destination is ambiguous among crossing connectors — I read it as weighing toward decline. Confirm?
- Is the legacy-UW override formally documented per case (who signed off, which checks passed)? That matters for how the mitigation plan is tracked.