Turn AI capability into a bounded, inspectable workflow for a clinical operator.
MY CONTRIBUTION
Product narrative · originating requirements · user experience framing
WHO IT SERVES
Clinical and service operators
THE PRODUCT VISION
Make useful knowledge easier to apply without obscuring who is responsible for the decision.
An anonymized experience narrative. Demonstrations use reconstructed flows and invented inputs; internal metrics and artifacts are excluded.
01 / THE STORY
A use case worth solving.
RECONSTRUCTED SCENARIO
An operator needs help understanding a procedure while handling a live task. Sometimes the reference material is sufficient. Sometimes context is missing. Sometimes the request asks for a decision the assistant should not make. A convincing product must handle all three.
Context
My experience includes originating concepts and requirements for clinical assistance: contextual knowledge, summarization, and workflow support. The product needed to help the operator’s task while making the relationship between suggestion, review, and action understandable.
What I owned
I authored product narratives and requirements and worked through the operator experience. My contribution was in framing the problem, defining the scope of assistance, and connecting capabilities to a reviewable workflow; the wider implementation was a team effort.
THE STARTING CHALLENGE
Search while doing the task
The operator locates, reads, and combines relevant guidance while keeping the interaction moving.
→
THE PRODUCT DIRECTION
Assistance within the task
A bounded assistant organizes relevant information for the operator to inspect and decide how to use.
02 / THE DECISIONS
The reasoning behind the product.
DECISION 01
Start with the operator’s task
Design around a specific job instead of a general chat surface.
The tradeoff
Fluent conversation is not evidence that the product helps someone complete the work.
DECISION 02
Make evidence inspectable
Connect a suggestion to relevant reference material and show when context is insufficient.
The tradeoff
A confident answer without support can increase the reviewer’s burden rather than reduce it.
DECISION 03
Make the boundary visible
Give out-of-scope requests a clear handoff and let people review or dismiss assistance.
The tradeoff
A boundary written in a product document is incomplete until it changes the user experience.
03 / EXPLORE THE FLOW
Follow a use case through the decisions.
Care Navigator uses scripted synthetic scenarios. It makes no clinical decisions, accesses no patient records, and represents a proposed experience rather than a live AI service.
Enable JavaScript to play the synthetic scenarios and inspect the decision flow.
04 / TEST THE PRODUCT THINKING
What would convince us it works?
A reconstructed evaluation plan showing the questions I would make explicit. These are evaluation criteria, not reported test results.
Use case
Expected behavior
Evidence to inspect
Supported question
The draft is relevant, inspectable, and useful for the task.
Source support; usefulness; correction effort.
Insufficient evidence
The assistant asks a relevant clarifying question.
Unsupported assertions; quality of clarification.
Outside the assistive scope
The experience hands off with context instead of taking authority.
Boundary adherence; handoff clarity; user understanding.
05 / OUTCOME & LEARNING
What the work changed.
The documented work includes the originating product concept and requirements for clinical assistance. The interactive Care Navigator below is a new, independent portfolio concept inspired by that experience. It is not a reproduction of the employer product and has not been validated with users.
WHAT I WOULD CARRY FORWARD
The next experiment I would run is whether an operator can accurately explain what the assistant knows, what it cannot decide, and how to recover from an unhelpful suggestion.