01Workflow friction
User problem
Repetitive work, unclear ownership, and constant context switching slowed every review.
More effort. Less focus.
Product workflow case study
Lennar Mortgage
How an AI-driven application simplified underwriting by automating document review, validation, due diligence, and exception routing—accelerating decisions, improving productivity, and saving approx $100,000.
This case study presents selected work. Non-disclosure agreements prevent me from sharing the full set of project artifacts.
One workspace for underwriting
The calendar, pinned reviews, and document-processing queue bring daily priorities into a single view.
Project Details
My Role
Information Architect & Design Lead. I was solely responsible for the UI deliverables, experience strategy, stakeholder approvals, UI design, and its implementation.
Collaboration
Cross-functional partnership across underwriting operations, product, design, and engineering.
AI Execution
AI agents run multiple tasks in parallel while generating accurate reports that reveal bottlenecks and operational issues.
The Impact
Reimagined underwriting operations around intelligent orchestration. AI agents now process multiple tasks and subtasks in parallel—improving productivity by up to 60% and saving approximately two hours per task.
The Problem Statement
Underwriters were forced to manage complex, interdependent loan reviews through manual assignment and a maze of disconnected tools—fragmenting workflow context across nearly six applications.
02Operational impact
Business problem
Longer cycles raised operating costs, reduced capacity, obscured performance, and constrained revenue.
More delays. Less capacity.
03Industry economics
Market pressure
In 2024, lenders spent $11,076 per loan for just $443 in profit. Manual work and disconnected systems put already thin margins under pressure.
Higher costs. Thinner margins.
Industry context
MBA benchmark. A digital credit-risk programme reduced processing costs by 30–40%. McKinsey case study
What I did
I led the experience and built agents to extend it.
User research and business context.
Human leadership Led by me
Understand the real work
- Conduct user research and stakeholder interviews.
- Capture business information, goals, and operational constraints.
- Identify repetitive tasks, friction, and opportunities for AI.
Agent support Built by me
User research agent
I created an agent to support user research—helping turn research material into organised findings that I review and interpret.
Human direction. Agent assistance.
What this creates
Research insights & business context
Explore each phase to see what I led, what I built, and how the work came together.
Product research
I mapped tasks, systems, handoffs, and wait states to understand where underwriting expertise mattered—and where the product could take on the repetitive work.
Nearly6applications in one workflow
The central findingExpert judgement.
Expert judgement.
Fragmented attention.
Underwriters balanced high-value risk decisions with document sorting, data entry, follow-ups, and queue administration across disconnected applications.
Context switchingManual coordinationRepetitive work
Business findings. Product priorities.
I grouped business findings into primary, secondary, and tertiary tasks. This hierarchy guided feature prioritisation and shaped how the product workflow comes together.
- 01Business findings
- 02Task priorities
- 03Product features
- 04Product workflow
| Assess and select risk | Triage new submissions and establish review priority | Monitor routine status notifications and reminders |
| Determine coverage structure | Review statements of values, loss-control reports, and supporting documents | Classify, bookmark, rename, and file documents |
| Set pricing, deductibles, and terms | Validate exposure values, business-income limits, and rating inputs | Extract data from documents and enter it into systems |
| Decide whether to quote, modify, decline, or bind | Investigate exceptions, discrepancies, and missing information | Reconcile routine worksheets and correct administrative data errors |
| Apply underwriting judgment to complex or unusual risks | Prepare referrals and obtain approval when authority limits are exceeded | Maintain queues, assign reviewers, and update workflow statuses |
| Own the final underwriting decision and its rationale | Confirm subjectivities, required documents, and bind readiness | Chase standard documents and perform routine follow-up that can be automated or delegated |
Primary priority / Design direction
Keep the underwriter’s decisions at the center.
Feature priorities
Risk assessment, exception review, and a clear decision workspace with evidence and rationale close at hand.
Role in the workflow
Review the prepared case, resolve exceptions, and make the final decision. Human ownership stays explicit.
Research methods and findings.
Mapped the loan journey
Mapped stages, systems, and handoffs to expose dependencies and wait states.
Underwriter mental model
Linked document gaps, shifting rules, and coordination friction to the need for trusted automation.
Note: I create a wide range of research and design artifacts, including personas, journey maps, A/B testing reports, analytics reports, harmonisation reports, and testing documentation. Non-disclosure agreements prevent me from sharing the complete set of deliverables; the examples above offer a glimpse into my process.
Solution & implementation
One workspace for priorities, evidence, and decisions. AI prepares and validates documents; underwriters review exceptions and own the final call.
PrepareValidateDecide
From automated checks to human review
Classify, extract, and validate in parallel. Route exceptions to a reviewer with the source evidence attached.
From design to experience
Preview Working Prototype (opens in a new tab)Take a closer look.
Step inside the working prototype to explore the underwriting experience.
Better tools. More room for human judgment
Useful automation gives experts the context and time to make better decisions.
Protect attention
Automate routine preparation so underwriters can focus on risk and exceptions.
Make trust visible
Keep source evidence and reasoning close to every recommendation.
Keep ownership clear
Show who acts next, what needs review, and where the human decision belongs.
