AI-Augmented Contract documents Analysis Platform

Designed trust into AI-powered contract analysis for high-pressure insurance workflows

Overview

My Role

Product Designer (Sole Designer)

Key Outcomes

Streamlined contract review during critical renewal periods
Designed Continuous AI Training in workflow that felt natural to users
Built user trust through transparent AI verification design

Additional Complexity

Remote collaboration with a dev team in France
Tech constraints: LLM latency and inability to build a specific layout
Ecosystem consistency: aligned with two products and fed insights into a core product

The Product

An internal underwriting application for insurance teams to review and triage submissions, manage workflows, and make faster and more confident decisions

The Problem

Fragmented, High-Stakes Workflows

During busy renewal periods, underwriters analyse hundreds of reinsurance and insurance contracts under tight deadlines. Their workflow is chaotic—constantly switching between:

Contract PDFs
Email chains
Notes documents
Slack conversations
Dropbox files

Need a solution that could

Make contract review instantaneous, not time-consuming
Simplify checking against complex underwriting rules
Reduce compliance risk and financial exposure

How might we accelerate contract analysis without compromising thoroughness, accuracy, or user confidence in AI-generated insights?

Research and Discovery

I conducted stakeholder interviews with underwriters and product managers to understand their workflows, pain points, and needs during high-pressure renewal periods.

Key Insights that shaped design

Navigation must be instantaneous and mirror document’s natural flow
Users question AI accuracy, design should communicate transparency and verifiable sources
AI assistance should feel like productive work, not extra overhead

Design Evolution

I went through three distinct layout approaches, testing each with PMs and users to land on the optimal information hierarchy that also fit within the dev team's technical constraints.

Early Prototype

This is what the UI looked like when I first joined the project.

#Christmas tree #Unclear IA
#Long scroll #Content gets lost

Explored Modal-Based Details (users’ fav)

Rebalanced the information hierarchy across multiple layers of detail

Consolidated all detailed information in a modal to minimise visual noise

(blocked by dev team due to tech constrains)

Hybrid Tabs + Accordion (final)

Parent accordions organised by document section (mirroring contract order)

Enable users to do side-by-side comparison with one click

Design Decision

Detected wording is compared against preferred language in a clear side-by-side view with concise risk summaries, enabling decisions in seconds.

AI-Powered Comparison with Source Verification

Human-in-the-Loop Machine Training

Underwriters can quickly verify or relink AI-detected text within the “Text” view, embedding feedback directly into their normal review workflow. Each correction improves model accuracy over time, ensuring the system evolves with real user judgement.

Natural Review Workflow

Key validation actions are surfaced at the point of review, enabling one-step decisions and preserving underwriters’ analytical flow.

Action List for Faster Collaboration

Instead of jumping between components to recall what needs discussion, underwriters get a consolidated view of every flagged item, suggested alternative, and comment at a glance. When everything checks out, a single 'Finish Validation' click closes the loop.