Case Study

Persistent Issue Detection for Multi-Unit Housing

A property management software company approached me with a recurring operational challenge: maintenance teams were overwhelmed by duplicate work orders, making it difficult to identify and prioritize unresolved issues.

My role was to explore the root cause of this behavior and design an AI-assisted maintenance prioritization system that reduces duplicate requests while improving operational clarity and tenant trust, without removing human oversight from the workflow.

Mergent Property OS analytics dashboard showing deduplication performance and trends

Role

Product Designer

Focus

AI/UX, Systems Thinking

Skills

UX Research, UI Design, Product Strategy

Project Overview

The client observed that tenants frequently submitted multiple requests for the same problem when they didn't receive timely updates or visibility into the status of their original submission. As a result, maintenance queues became cluttered, issue tracking became fragmented, and property managers struggled to identify which problems required immediate attention.

The client wanted to explore how AI could help identify recurring issues without removing human oversight from the workflow, reducing triage effort while improving operational visibility and tenant confidence.

The Challenge

The client's existing system treated every maintenance request as an independent task, regardless of whether the issue had already been reported. This created several compounding problems:

Duplicate maintenance requests cluttering the queue
Increased triage workload for property managers
Fragmented issue histories
Delayed prioritization of critical problems
Reduced tenant confidence in the maintenance process

My Approach

Through research and analysis, I reframed the problem from one of duplicate detection to one of unresolved user uncertainty. I identified that repeated maintenance submissions were often signals of deeper issues:

Lack of status visibility

Unresolved tenant frustration

Operational bottlenecks

Building-wide maintenance failures

Based on this insight, I designed an AI-assisted work order consolidation system that helps property managers identify persistent unresolved issues while maintaining full control over final decisions.

The Solution

The proposed system uses AI to assist with maintenance triage, reducing effort while improving operational visibility, rather than replacing human decision-making.

Detecting semantically similar work orders
Recommending potential duplicate groupings
Identifying persistent unresolved issues
Escalating long-running maintenance problems
Detecting recurring issues across multiple units
Providing explainable recommendations with human approval controls

AI-powered Duplicate Detection screen grouping related work orders into reviewable clusters with severity levels and manager actions — The Duplicate Detection workspace groups related submissions into clusters, flags severity, and keeps “Mark Reviewed” and “Not a duplicate” decisions in the manager's hands.

Prioritization Framework

To determine urgency and escalation priority, the system evaluates multiple signals together rather than relying on raw ticket volume.

Signal	Purpose
Safety keywords	Identifies urgent hazards and risks
Time unresolved	Detects long-standing issues
Duplicate frequency	Surfaces recurring operational problems
Multi-unit recurrence	Identifies potential building-wide failures
Semantic similarity	Groups related requests
Tenant sentiment	Detects escalating frustration

Issue detail modal showing grouped tenant reports, escalation signals, and an explainable AI Analysis with a confidence rating — Drilling into a cluster reveals the underlying reports alongside an explainable AI analysis: Semantic, spatial, and temporal signals with a confidence rating, so managers can trust the recommendation before acting.

Key Design Principles

Transparency

Users should understand why issues are being grouped or escalated.

Human Oversight

Property managers retain final authority over work order consolidation.

Trust Building

Tenants need visibility into issue status to reduce repeat submissions.

Cognitive Load Reduction

Interfaces should prioritize meaningful problems rather than raw ticket volume.

Proposed Features

AI-powered duplicate detection suggestions
Persistent issue timelines
Escalation indicators for unresolved problems
Building-wide pattern detection
Explainable AI confidence scoring

Intended Outcomes

For Property Managers

Lower triage workload and faster issue prioritization through assisted, explainable recommendations.

For Tenants

Improved confidence and transparency, with clear visibility into issue status to reduce repeat submissions.

For Operations

Reduced duplicate maintenance requests and a cleaner, more reliable issue history.

For the Property

Earlier detection of infrastructure-level problems before they become building-wide failures.

Reflection

This project reinforced the importance of looking beyond surface-level user behavior to understand the underlying systems problem. What initially appeared to be a duplicate ticket issue revealed a broader challenge involving communication gaps, operational ambiguity, and user trust.

By reframing the problem, I was able to design a solution that balanced AI assistance with human oversight while addressing both operational efficiency and tenant experience.

Want to see more of my work?

View All Projects