Work / Case study

UW Triage Engine + Board

In production Updated September 22, 2026 Google Apps Script

Automated the underwriting-notice workflow end to end: LLM-powered triage, a web-based work board and two-way Slack integration for accountability and SLA tracking.

Source code is private. Walkthroughs are available on request.

0regression failures at v2.10
2-waySlack integration
1Past Due flag, measured in days

Business impact

Replaced an email inbox with an accountable, SLA-driven workflow. Every carrier underwriting notice becomes a tracked work item with an owner and a due date, protecting policy retention and giving leadership real-time visibility into what is past due.

Problem

Carrier underwriting notices arrived by email and fell through the cracks. There was no single source of truth for what was due, who owned it, or what was past due, which put policies and client retention at risk.

Solution

Built an intake-to-resolution pipeline. An Apps Script engine monitors the notices mailbox, uses the Claude API to extract structured fields, and writes work items to a central tracker. A web-app board lets agents claim and close items, with Slack DMs, a daily digest and commands issued straight from Slack threads.

Results & highlights

  • Shipped iteratively through board v2.10.5 and engine v1.4, including backfill of legacy notices
  • Built two-way Slack integration within platform constraints (polling, since Apps Script can't read Slack signing headers), with every write going through one controlled path
  • Standardized SLA visibility: one Past Due flag measured in days, plus an end-of-day digest
  • Regression suites recorded 0 failures at v2.10

Competencies demonstrated

  • Workflow automation: Built an intake-to-resolution pipeline in Google Apps Script that turns underwriting emails into tracked work items on a central board.
  • AI / LLM integration: The Claude API extracts structured fields from each notice before it reaches the board.
  • Systems integration (Slack API): Two-way Slack sync (DMs, a daily digest and commands from threads) with every write going through one controlled path.
  • SLA management & reporting: One Past Due flag measured in days, plus an end-of-day digest for accountability.
  • Iterative delivery & QA: Shipped through board v2.10.5 and engine v1.4, backfilled legacy notices, and recorded 0 regression failures at v2.10.

Architecture

  1. Notice intakeunderwriting notices
  2. LLM triageAI-assisted classification
  3. Web work boardGoogle Apps Script
  4. Two-way Slackpolling sync, one write path
  5. SLA trackingPast Due flag + end-of-day digest
Design-level view. Hostnames, addresses and account details are intentionally omitted.

Next: Audit & Compliance Scoring System →