Challenge
Manual review made inconsistent referrals difficult to scan and compare quickly.
PRODUCT CASE STUDY · 2026
Triagen: AI-Assisted HaematologyTriage SystemLive Prototype
A clinical decision-support concept that helps triage nurses prioritize referrals, inspect the evidence behind an AI recommendation, and retain final authority.
At a Glance
A focused decision-support concept for haematology referral review
Triagen brings the referral queue, AI-supported triage pathway, clinical evidence, and nurse actions into one workspace. It was designed as a team project for a healthcare customer-experience course.
Design direction
Make every recommendation traceable without taking the decision away from the nurse.
My contribution
UX research, workflow framing, interaction design, and high-fidelity prototyping.
Problem
Each referral required the same careful detective work
A nurse had to read the letter, find the relevant lab values, decide what mattered, and then assign urgency. The information was available, but it was scattered across the case with no shared interface to guide the review.
One thing we would not compromise onThe AI could organize the evidence. The nurse still made the call.
The interface could suggest urgency and explain why, but accepting or changing that recommendation always stayed with the nurse.
Research & Direction
The research pointed to transparency and clinician control
We combined a literature review, stakeholder mapping, low-fidelity prototyping, and feedback from domain experts. Three practical requirements shaped the final direction.
Show the evidence
A summary alone was not enough. Reviewers needed the clinical factors and source information behind a recommendation.
Make authority obvious
Accept and Modify actions needed to be explicit, documented, and clearly owned by the nurse.
Support quick questions
Domain experts wanted a way to query the case without searching through the full referral again.
Design Development
The workflow became more specific with each round of feedback
The early concepts tested two core moments: finding the cases that need attention and understanding why a case was flagged. Expert feedback then helped us make the nurse's authority more explicit.
Started with the urgent-case queue
The first concept focused on helping nurses see which referrals needed attention first.
Experts asked to see the evidence
A summary was not enough. Reviewers needed the clinical factors and source data behind every recommendation.
Made nurse actions explicit
Accept and Modify became primary actions, making it clear that the nurse—not the AI—owned the decision.
Added a traceable decision record
Recommendations, changes, notes, and supporting evidence came together in the case history.
Solution
AI surfaces a suggested pathway. The nurse checks the evidence and decides.
Triagen turns unstructured referral information into a reviewable case workspace. The recommendation is visible, but it never stands alone.
Final Interface
Evidence stays connected to the decision
Supporting information is organized around the review task, so nurses can move from recommendation to verification without losing context.
Anticipated Impact
What should feel different for the nurse
Design intent; clinical impact has not yet been measured.
Reflection
The hardest part was not generating a recommendation. It was making that recommendation trustworthy.
Designing Triagen changed how I think about AI in clinical workflows. A useful system cannot simply produce an answer. It must show its reasoning, make uncertainty visible, and leave room for the clinician to disagree.