Prep: James Park — steering committee readout (Tuesday)
Simulated demo brief — in the real app this is generated live by Claude with read-only access to Calendar, Gmail, and the web, with inline source attribution.
Where things stand
- Phase 1 of the AI underwriting transformation is engaged and on track: triage model at 84% precision on holdout, pilot expanded to the inland-marine desk.
- You owe the readout deck draft by Friday (per 2026-07-16 pre-read call); James owes you the final attendee list.
- Relationship: Trusted, weekly rhythm. Grace Kimura runs day-to-day and drafts internal comms from your materials.
What's changed since last touch
- (per calendar) The steering committee moved 30 minutes earlier — now 8:00am Tuesday; CFO confirmed attending.
- (per email) Grace's underwriter feedback session happened Thursday — outcome not yet logged; ask before the readout.
- (per web) A competing MGA announced an "AI underwriting copilot" pilot this week — expect the board to ask how Atlas compares.
Objectives for this meeting
- Get the CFO to repeat the "underwriter hours returned" framing in his own words.
- Land the readout date and agenda; volunteer to present the benchmark slide.
- Seed Phase 2 (claims intake automation) without asking for it yet.
Talking points
- 84% precision → but lead with submissions-per-underwriter-day, the metric Grace's dashboard tracks.
- Anonymized benchmark: comparable program shops see 22-30% cycle-time reduction by month 4.
- The competitor pilot news is good for the narrative — Atlas started six months earlier.
Questions to ask
- What did the seniors say in Thursday's feedback session — did the routing-rule pushback resolve?
- Which board member is most skeptical, and what would move them?
- Does he want Phase 2 scoped before or after renewal season?
Watch-outs
- Don't let the readout become a model-metrics review — CFO buys hours and dollars.
- The poached-underwriters wound is fresh; keep the pro-underwriter framing explicit.
Suggested follow-up actions
- Log the steering-committee outcome and close the readout-deck action item.
- If Phase 2 interest surfaces, open a "scoping" opportunity and date the next step.
Generated 2026-07-18 08:12 · focus: Phase 1 readout dry-run — in the real app, verify anything attributed to Gmail/web before relying on it.