Engineering Blog · Post #64

AI Trend Detection: Proactive Alerts on Broker Volume, Hit Ratios, and Submission Patterns

InsightUW doesn't just process submissions — it watches for patterns. When a key broker's volume drops 50%, you'll know before your competitor does.

The Problem

A key broker has been submitting 8 risks per month for 2 years. Last month they submitted 3. This month, 2. By the time the UW manager notices the drop in a quarterly report, the broker has already moved their business to a competitor. The pattern was visible in the data months ago — but no one was watching.

Similarly, a 20% decline rate might be acceptable, but if it spikes to 35% in Cyber submissions, something is wrong with appetite alignment. These trends hide in aggregate data until it's too late.

The InsightUW Approach

InsightUW's AI Trend Detection runs 6 pattern analyses across all submissions and generates proactive alerts when anomalies are detected. Alerts are delivered as in-app notifications — catching trends before they become problems.

graph TD subgraph Data["Submission Data"] ALL["All submissions across all brokers, LOBs, and time periods"] end subgraph Detectors["6 Trend Detectors"] D1["Broker Volume Drop 30d < 50% of average"] D2["Broker Volume Spike 30d > 150% of average"] D3["LOB Mix Shift Share change > 10pts"] D4["High Decline Rate Overall > 15%"] D5["Low Hit Ratio Broker < 5% w/ 10+ subs"] D6["Velocity Slowdown Week < 40% of average"] end subgraph Alerts["Proactive Alerts"] A1["🔴 Marsh volume down 50%"] A2["🔴 Gallagher hit ratio 0%"] A3["🔵 Cyber share up 10pts"] end subgraph Notify["Notifications"] Bell["Bell icon alert to UW/manager"] end ALL --> D1 ALL --> D2 ALL --> D3 ALL --> D4 ALL --> D5 ALL --> D6 D1 --> A1 D5 --> A2 D3 --> A3 A1 --> Bell A2 --> Bell A3 --> Bell

Six Alert Types

Alert	Trigger	Severity	Example
Broker Volume Drop	30d submissions < 50% of monthly avg	Warning	"Marsh: 3 subs vs 6.0 avg — 50% decrease"
Broker Volume Spike	30d > 150% of average	Info	"New broker submitting 5x historical rate"
LOB Mix Shift	Share change > 10 percentage points	Info	"Cyber: 8% → 18% in last 30 days"
High Decline Rate	Overall decline rate > 15%	Warning	"Decline rate 18% vs ~10% industry"
Low Hit Ratio	Broker < 5% with 10+ submissions	Warning	"Gallagher: 0 of 18 bound (0%)"
Velocity Slowdown	This week < 40% of weekly average	Warning	"2 subs this week vs 8 average"

Broker Trend Dashboard

The broker trends API returns per-broker analytics:

Broker	Total	30d	Avg/Mo	Trend	Hit Ratio	Top LOB
Lockton	20	4	6.7	Stable	5.0%	D&O
Gallagher	18	4	6.0	Stable	0.0%	Cyber
Marsh	18	3	6.0	Decreasing	0.0%	Mixed

LOB-Specific Example

Trend Detection Run — 2 Alerts Generated:

The system detects:

Low Hit Ratio: Gallagher (0.0%) — "Only 0 of 18 submissions bound. Consider reviewing submission quality and appetite alignment with this broker."

Action: UW manager schedules a meeting with Gallagher to understand why none of their 18 submissions converted. Are they submitting outside appetite? Is our pricing non-competitive? Are they testing the market?

Marsh Volume Decreasing — "Submitted only 3 in last 30 days vs. 6.0 monthly average — 50% decrease."

Action: Relationship manager reaches out to Marsh to ensure the partnership is still strong. Early detection prevents losing a key broker relationship.

What This Means for Underwriters

Proactive, not reactive — trends detected before they become problems
Broker relationships — volume drops caught early, relationships preserved
Appetite alignment — high decline rates and low hit ratios signal misalignment
Market intelligence — LOB mix shifts reveal changing market demands
Scheduled in production — runs daily via cron, no manual trigger needed

What's Next

Next: Claims Bordereaux Reminders — AI Reminds You 30 Days Before Renewal

Want AI watching your submission patterns? Request a demo.