Engineering Blog · Post #6

Complete Submission Audit Trail: 20+ Event Types, Field-Level Change Tracking, and One-Click Compliance Export

How InsightUW tracks every action on an Excess Casualty submission — from the moment the broker email lands to the day the policy binds — with field-level before/after snapshots and a single-click regulatory export.

The Problem

When a state regulator asks "Why did you decline this Excess Casualty submission?", you need an answer. Not a guess. Not a reconstruction from memory. Not a search through 47 emails and 3 systems.

Most underwriting workstations log some events — submission created, status changed, maybe a note or two. But the critical details are missing:

Who changed the limit from $25M to $15M, and when? — No field-level tracking
Why was the appetite score 45 instead of 72? — AI decisions aren't logged
The broker says they sent updated financials on March 3 — did we receive them? — Document receipt isn't timestamped
The underwriter quoted at $180K, but the bound premium is $165K — who approved the deviation? — Pricing authority decisions live in emails
For SOX compliance, we need to show segregation of duties between who quoted and who bound — Roles aren't tracked per action

The regulatory risk: Market conduct exams, E&O lawsuits, reinsurance audits, and DOI complaints all require a complete, tamper-evident record. If you can't produce it, you're exposed.

InsightUW's Audit Architecture

Every action in InsightUW generates an immutable audit event. Events are append-only — nothing is deleted, nothing is overwritten. Each event records who, what, when, why, and the before/after state.

graph TD subgraph Actions["User & System Actions"] A["CBP creates submission"] B["AI analyzes risk"] C["UW changes field"] D["Manager approves quote"] E["System sends notification"] F["Broker responds"] end subgraph Audit["Immutable Audit Log"] G["Event Type (20+ types)"] H["Actor (user, system, bot, AI)"] I["Timestamp (UTC, millisecond precision)"] J["Payload (before/after, metadata)"] K["Correlation ID (links related events)"] end subgraph Storage["Storage & Export"] L["Append-Only Database (no Update, no Delete)"] M["One-Click Export PDF / CSV / Json"] N["Regulatory Report DOI / SOX / E&O format"] end A --> G B --> G C --> G D --> G E --> G F --> G G --> H H --> I I --> J J --> K K --> L L --> M L --> N

The 20+ Event Types

Every event type has a category, icon, and structured payload:

#	Event Type	Category	Icon	Triggered By	Payload Contains
1	email received	Intake	mail icon	System	sender, subject, attachment count, mailbox
2	email parsed	Intake	robot icon	AI Parser	extracted fields, confidence scores
3	submission created	Lifecycle	document icon	System/User	submission number, LOB, source, initial data
4	documents stored	Documents	save icon	System	file names, sizes, storage paths, SHA-256 hashes
5	document received	Documents	paperclip icon	System/User	file name, category, source (email/upload/bot)
6	ofac screening	Compliance	shield icon	System	result (clear/match/review), lists checked
7	duplicate check	Compliance	search icon	System	result (clear/duplicate), matched submissions
8	fifo assigned	Assignment	person icon	FIFO Engine	assigned_to, team, rotation_position, reason
9	`reassigned`	Assignment	transfer icon	Manager	from_user, to_user, reason
10	notification sent	Communication	bell icon	System	recipient, channel (app/email), message_type
11	ai analysis	Analysis	brain icon	AI Engine	8 check results, scores, recommendations
12	appetite scored	Analysis	gauge icon	Appetite Engine	score, rule_matched, decision (accept/refer/decline)
13	field changed	Data	pencil icon	User	field_name, old_value, new_value, reason
14	status changed	Lifecycle	refresh icon	User/System	old_status, new_status, trigger
15	rush flagged	Priority	fire icon	User	reason, requested_by, sla_hours
16	rush removed	Priority	flag icon	User	reason, original_rush_duration
17	quote created	Workflow	calculator icon	Underwriter	quote_number, premium, terms, authority_level
18	quote approved	Workflow	check icon	Manager	approver, authority_level, deviation_pct
19	quote sent	Communication	send icon	System	recipient, delivery_method (email/portal)
20	broker response	Communication	reply icon	Broker	response_type (accept/counter/decline), details
21	binder issued	Workflow	lock icon	System	binder_number, effective_date, premium
22	`bound`	Lifecycle	check-circle icon	Underwriter	policy_number, final_premium, effective_date
23	`declined`	Lifecycle	x-circle icon	Underwriter	reason_code, reason_text, appetite_score
24	comment added	Notes	message icon	User	comment_text, visibility (internal/broker)

The Scenario

TechBridge Manufacturing needs $25M Excess Casualty coverage over a primary GL program. The submission arrives by email, goes through full underwriting, gets quoted, negotiated, and bound over 12 business days.

Here is the complete 18-event audit trail:

Field-Level Change Tracking

The most powerful audit capability is field-level change tracking. Every time a user modifies any field on a submission, InsightUW captures the old value, new value, who changed it, when, and why.

graph TD subgraph Before["Before (Mar 3)"] A["requested limit: $25,000,000"] B["attachment point: $2,000,000"] C["broker contact: david.chen@wtwco.com"] D["effective date: 2026-06-01"] end subgraph Change["Change Event (Mar 5)"] E["Robert Chen changes requested limit"] F["Reason: Broker confirmed insured revised down after board review"] end subgraph After["After (Mar 5)"] G["requested limit: $15,000,000"] H["attachment point: $2,000,000"] I["broker contact: david.chen@wtwco.com"] J["effective date: 2026-06-01"] end A -->|"changed"| E E -->|"new value"| G B -->|"confirmed (no change)"| H C -->|"unchanged"| I D -->|"unchanged"| J

Field Change History Table

For any submission, managers and auditors can view the complete field change history:

#	Timestamp	Field	Old Value	New Value	Changed By	Reason
1	Mar 3, 9:00:14 AM	insured_name	(created)	TechBridge Manufacturing Inc	AI Parser	Email parsed
2	Mar 3, 9:00:14 AM	lob	(created)	Excess Casualty	AI Parser	Keyword detection
3	Mar 3, 9:00:14 AM	requested_limit	(created)	$25,000,000	AI Parser	Email body extraction
4	Mar 3, 9:00:14 AM	effective_date	(created)	2026-06-01	AI Parser	Date regex match
5	Mar 3, 9:00:14 AM	attachment_point	(created)	$2,000,000	AI Parser	"$2M primary" extraction
6	Mar 5, 2:30:00 PM	requested_limit	$25,000,000	$15,000,000	Robert Chen	Broker confirmed revised limit
7	Mar 10, 9:45:01 AM	quoted_premium	(created)	$165,000	Robert Chen	Quote QT-2026-XS-TB-001
8	Mar 14, 4:15:00 PM	status	quoted	bound	Robert Chen	Binder issued

One-Click Compliance Export

InsightUW provides three export formats, each tailored to a specific compliance use case:

PDF Export (for DOI Market Conduct Exams)

A formatted, paginated document with:
- Cover page: submission summary, insured, LOB, outcome
- Chronological event timeline with icons and descriptions
- Field change history with before/after values
- Document receipt log with timestamps
- Decision rationale (appetite score, quote deviation, approval chain)
- Digital signature: SHA-256 hash of export contents for tamper detection

CSV Export (for Reinsurance Audits)

Flat tabular data, one row per event:

Compliance & Regulatory Use Cases

graph LR subgraph Audit["InsightUW Audit Trail"] A["18 events Field-level tracking Tamper-evident hash"] end subgraph Regulators["Regulatory Uses"] B["DOI Market Conduct Exam Why was this risk declined?"] C["SOX Compliance Segregation of duties: who quoted vs who approved"] D["E&O Defense Complete decision chain: what was known at quote time"] end subgraph Internal["Internal Uses"] E["Reinsurance Audit Prove treaty compliance: limits, territory, exclusions"] F["Management Review Underwriter decision patterns and deviation tracking"] G["Training Review real submissions for UW training scenarios"] end A --> B A --> C A --> D A --> E A --> F A --> G

SOX Segregation of Duties

For publicly traded carriers, SOX requires that the person who creates a quote is not the same person who approves a pricing deviation. InsightUW's audit trail makes this trivially provable:

Action	Actor	Role	Timestamp	SOX Control
Quote created at $165K	Robert Chen	Senior UW	Mar 10, 9:45 AM	Originator
Deviation noted: -9.3% from indicated	System	Automatic	Mar 10, 9:45 AM	Detection
Quote approved	Patricia Morgan	VP Underwriting	Mar 10, 10:00 AM	Approver

Different person. Different role. Different timestamp. Logged and exportable.

E&O Defense Scenario

Two years after binding, TechBridge Manufacturing has a $12M products liability claim that pierces the excess layer. The insured's attorney claims the carrier should have known the risk was uninsurable. The audit trail shows:

Event 9: AI appetite score was 68/100 — within appetite
Event 9: D&B credit score 72, no prior excess claims
Event 10: Limit was reduced from $25M to $15M per broker instruction
Event 14: Premium of $165K was within the indicated range ($155K-$210K)
Event 15: VP approved the deviation with documented rationale

The underwriting decision was informed, documented, and approved through proper authority channels. The audit trail is the defense.

Audit Trail Search & Filter

The audit trail UI supports filtering by:

Filter	Options	Example
Event type	Any of 24 types	Show only field changed events
Actor	User, system, AI, bot	Show only human actions
Date range	From/To	Events between Mar 5-10
Category	Intake, Lifecycle, Data, Compliance, etc.	Show only Compliance events
Field name	Any tracked field	All changes to requested limit

Key Metrics

Metric	Legacy Systems	InsightUW Audit Trail
Events tracked per submission	3-5 (create, assign, status)	18-25 (every action)
Field-level change history	None	Every field, every change, with reason
Time to produce audit export	2-5 days (manual compilation)	< 5 seconds (one-click)
Audit trail completeness	~40% (gaps in manual logging)	100% (system-enforced)
DOI exam preparation time	40+ hours per exam	2-4 hours per exam
Tamper detection	None	SHA-256 hash on every export
Segregation of duties proof	Manual attestation	Automatic from event actors

InsightUW's audit trail turns compliance from a burden into a byproduct of normal workflow. Every action logged. Every change tracked. Every export one click away. Request a demo to see the complete audit trail.