# A 53% capacity uplift per Underwriter, without removing human judgment

_An assisted-underwriting agent runs the intelligent intake, reads the documents, runs the eligibility and risk-appetite checklist, and arrives at the Underwriter with a pre-opinion, risk summary, and follow-up questions — every step audit-trailed._

## The client

Underwriting is judgment work, but most of the day around it is not. Document analysis, pendency-chasing, and eligibility checklists eat the Underwriter's hours before they ever apply their expertise. Proposals advance with incomplete data, decisions stay loosely standardized, and the audit trail is rebuilt by hand at the end.

## The problem

Subscription decisions depend on manual document analysis, proposals advance with pendings and incomplete data, and the decision becomes hard to standardize or trace. The Underwriter spends judgment hours on rote work, which caps how much business the team can write per month.

## What we built

HomoDeus builds an assisted-underwriting agent with intelligent data, document, and context intake. The agent reads the documents and flags pendencies, runs the eligibility and risk-appetite checklist, and delivers a pre-opinion, risk summary, and follow-up questions to the Underwriter. Every decision is audit-trailed end to end with sources, versions, justifications, and approvals — the Underwriter applies judgment instead of chasing data.

## Outcomes

- +53% — capacity per Underwriter, with human control preserved
- ~71% — of document analysis pre-processed before reaching the Underwriter
- 4 layers — audit trail: sources, versions, justifications, approvals
- 0 ungoverned decisions — every decision traceable end to end

## Key outcome

The Underwriter applies expertise where it matters and signs off a defensible decision. The carrier writes 53% more business per Underwriter without easing risk discipline.

## Stack

- Intelligent intake (data, documents, context)
- Automatic document reading + pendency detection
- Eligibility + risk-appetite checklist
- Pre-opinion + risk summary generator
- Follow-up question agent
- Audit trail (sources / versions / justifications / approvals)

## Time to production

6-8 weeks to first production agent

## Questions buyers ask

### Does AI replace the Underwriter's decision?

No. The agent runs the intake, document reading, eligibility, and risk-appetite checklist, and delivers a pre-opinion and follow-up questions. The Underwriter applies the judgment and signs off. The Underwriter's expertise is reallocated from data-chasing to risk decisions.

### How is governance preserved when the agent does more of the work?

Every step is audit-trailed end to end: sources, versions, justifications, approvals. The carrier can defend any decision the way it would defend a fully manual one — with the evidence already in place rather than reconstructed after the fact.

### What is the realistic capacity uplift per Underwriter?

In this engagement, capacity per Underwriter rose 53% without easing risk discipline, because around 71% of document analysis and pendency work was pre-processed before the case reached the Underwriter's desk.

## Related

- https://www.homodeus.me/work/insurance-product-channel-copilot.md
- https://www.homodeus.me/work/insurance-claims-nps-radar.md
- https://www.homodeus.me/work/insurance-regulatory-copilot.md
- All impact studies: https://www.homodeus.me/work