Industry: Health Tech / AI Product Design
Executive Summary
Developed as part of the Hustle Badger: Build with AI course, Wound Wizard is an AI-assisted concept designed to support non-specialist carers with wound assessment. The project focused on identifying responsible AI use cases, designing human-in-the-loop experiences, and framing AI as a supportive tool rather than a clinical authority in a high-risk domain.
My Role: Product Designer (AI Use Case Definition, Prompt Engineering, Experience Design, Risk Mapping).
Context and Challenge
Wound assessment is frequently performed by non-specialists or carers who lack formal medical training. Current guidance is often overly clinical, fragmented, and difficult to interpret, leading to anxiety and delayed action.
-
The Design Challenge: How do we use AI to provide clarity without the user over-relying on it as a diagnostic tool?
-
The High-Risk Barrier: In healthcare, AI must be framed with extreme care to maintain safety, user trust, and clinical boundaries.
My Approach: Responsible AI Discovery
I followed a risk-aware discovery process to ensure the AI added value without crossing into “medical advice.”
Activities included:
-
Problem Definition: Setting clear boundaries to avoid diagnostic claims, focusing instead on “confidence and understanding.”
-
Risk & Over-reliance Mapping: Explicitly mapping where a user might “blindly trust” the AI and designing friction points to prevent it.
-
Prompt Engineering: Designing and iterating on structured prompts that gather symptoms and context while ensuring the output remains cautious and explainable.
-
Human-in-the-Loop Design: Ensuring every AI output was paired with an escalation path to a human professional.
The Solution: AI as Decision Support
Wound Wizard is a guided experience that uses AI to surface patterns and considerations in plain language, rather than providing a “final answer.”
-
Structured Dialogue: Guides users through specific questions to gather consistent data.
-
Pattern Recognition: Uses AI to highlight potential risk signals (e.g., “This pattern often indicates inflammation”) without making a definitive diagnosis.
-
Uncertainty Framing: The UI was designed to highlight what the AI doesn’t know, encouraging users to seek professional help when signals are unclear.
-
Calm & Reassuring Tone: Following the “CosimaCreates” philosophy of replacing anxiety with confidence through clear, human language.
The Iteration: Reframing Authority
During the validation phase, I iterated on the “voice” of the AI. Initial prompts felt too “authoritative,” which increased the risk of user over-reliance.
-
The Feedback: Users were taking AI suggestions as absolute facts rather than “considerations.”
-
The UX Improvement: I adjusted the prompt structures and UI components to use “hedging” language (e.g., “This might suggest…” or “Consider checking…”) and added persistent disclaimers.
-
The Result: Testing showed that users were more likely to use the tool as a secondary check while still planning to consult a clinician.
The Outcome
The project resulted in a robust, testable product concept that demonstrates how to navigate the ethics of AI in sensitive environments.
-
Strategic Impact: Proven ability to identify where AI shouldn’t be used, as much as where it should.
-
Design Impact: A strong example of responsible AI design, prompt engineering, and the practical application of “human-in-the-loop” principles.
-
Technical Application: Demonstrated a structured approach to prompt design and the framing of AI-generated insights in a high-stakes UX.