66% faster report review
66% faster report review
with verifiable AI
with verifiable AI
Public water inspectors lose critical hours manually extracting data from complex incident reports. I designed a transparent AI RAG tool that summarises events and classifies severity risk, allowing teams to verify the system's reasoning and make faster, more confident intervention decisions.
Public water inspectors lose critical hours manually extracting data from complex incident reports. I designed a transparent AI RAG tool that summarises events and classifies severity risk, allowing teams to verify the system's reasoning and make faster, more confident intervention decisions.
Role
Lead UX/UI designer
Timeline
6 weeks
Team
Small cross-functional team
Type
AI Product Design
Role
Lead UX/UI designer
Timeline
6 weeks
Team
Small cross-functional team
Type
AI Product Design
Role
Lead UX/UI designer
Timeline
6 weeks
Team
Small cross-functional team
Type
AI Product Design
Challenge
A National Water Regulator receives 500-600 water safety incident reports annually. With reports spanning up to 60 pages, a small team of inspectors was losing critical hours to manual triage. This administrative bottleneck delayed severity classifications, stretching resources and risking slower responses to high-severity public safety events.
Process
I led the UX discovery process for a rapid pilot, co-facilitating workshops with inspectors to map their as-is workflow. I introduced blue sky ideation to surface ambitious possibilities, then helped the team distil those into a ruthless MVP scope around one need: verifiable AI assistance.
Solution
An AI-powered document reader built on a Retrieval-Augmented Generation (RAG) architecture that summarises complex incidents, classifies severity risk, and explicitly links its recommendations to the source text. Built on principles of human-centred transparency, the interface allows inspectors to instantly audit the AI's reasoning alongside the original PDF.
Impact
The pilot proved that AI could safely augment expert inspection. In early user testing, review time dropped by roughly 66% (from an average of 1.5 hours to 30 minutes per report). The lead inspector praised the generated summaries and the in-document highlighting, noting that the tool's initial classifications were highly accurate.
Challenge
A National Water Regulator receives 500-600 water safety incident reports annually. With reports spanning up to 60 pages, a small team of inspectors was losing critical hours to manual triage. This administrative bottleneck delayed severity classifications, stretching resources and risking slower responses to high-severity public safety events.
Process
I led the UX discovery process for a rapid pilot, co-facilitating workshops with inspectors to map their as-is workflow. I introduced blue sky ideation to surface ambitious possibilities, then helped the team distil those into a ruthless MVP scope around one need: verifiable AI assistance.
Solution
An AI-powered document reader built on a Retrieval-Augmented Generation (RAG) architecture that summarises complex incidents, classifies severity risk, and explicitly links its recommendations to the source text. Built on principles of human-centred transparency, the interface allows inspectors to instantly audit the AI's reasoning alongside the original PDF.
Impact
The pilot proved that AI could safely augment expert inspection. In early user testing, review time dropped by roughly 66% (from an average of 1.5 hours to 30 minutes per report). The lead inspector praised the generated summaries and the in-document highlighting, noting that the tool's initial classifications were highly accurate.
Challenge
A National Water Regulator receives 500-600 water safety incident reports annually. With reports spanning up to 60 pages, a small team of inspectors was losing critical hours to manual triage. This administrative bottleneck delayed severity classifications, stretching resources and risking slower responses to high-severity public safety events.
Process
I led the UX discovery process for a rapid pilot, co-facilitating workshops with inspectors to map their as-is workflow. I introduced blue sky ideation to surface ambitious possibilities, then helped the team distil those into a ruthless MVP scope around one need: verifiable AI assistance.
Solution
An AI-powered document reader built on a Retrieval-Augmented Generation (RAG) architecture that summarises complex incidents, classifies severity risk, and explicitly links its recommendations to the source text. Built on principles of human-centred transparency, the interface allows inspectors to instantly audit the AI's reasoning alongside the original PDF.
Impact
The pilot proved that AI could safely augment expert inspection. In early user testing, review time dropped by roughly 66% (from an average of 1.5 hours to 30 minutes per report). The lead inspector praised the generated summaries and the in-document highlighting, noting that the tool's initial classifications were highly accurate.
