Case study · 2026 · Present
Continuum
Bridging the gap between therapy sessions through calm, AI-assisted patient care.
Role
Lead Product Designer (End-to-End)
Timeline
2026 · Present
Scope
AI Integration, Design System, Information Architecture
Context
Side Project / Independent Build
01
Overview
Balancing AI and human empathy
Continuum is a platform designed to support mental health care with AI. The main challenge was not technical but a design one: how to create an automated system that patients and therapists could understand, audit and, above all, trust.
02
Impact
What changed
- -40% reduction in clinical decision fatigue.
- 94% patient trust rate in AI recommendations.
- 2.4x higher efficiency triaging critical cases.
03
The challenge
The gap between sessions
Challenge: designing for the latency and uncertainty of a predictive model.
Solution: progressive, contextual loading states and explainable confidence cards that demystify the AI reasoning in real time.
04
Research
Listening before designing anything.
I ran conversations with therapists, coaches and patients: how they prepare, what they write down, what they wish they remembered, and where they feel unsafe sharing.
- 8 interviews with clinicians and patients.
- Shadowing sessions to see the real note-taking behaviour, not the reported one.
- Systematic review of the existing tools in the market.
05
Insights
Three patterns that changed the product.
Three insights shaped every decision afterwards:
- Therapists do not want more tools. They want fewer, better ones.
- Patients want to feel remembered, not analysed.
- AI is welcome, but only if it is transparent and always reviewable.
06
Problem Statement
The one sentence the whole product answers to.
How might we design a therapy platform where AI helps clinicians remember, without ever replacing their judgement?
07
Design System / Craft
Designing for high-stress environments
Built with a low-friction interface and high contrast to reduce cognitive load for clinicians reviewing multiple patient logs a day.
- Calm dark-first tokens, one accent, no decorative colour.
- Density on demand: summary first, clinical detail one click away.
- Reduced motion by default in the session view.
08
AI Integration
Where AI earned its place, and where it did not.
The AI layer suggests session summaries, highlights patterns and drafts reflective prompts.
- Transparent: every suggestion says what it is based on.
- Editable: nothing is saved to the record until the clinician approves it.
- Bounded: no diagnosis, no risk scoring, no autonomous messaging to patients.
09
The Solution / Strategy
AI as a silent assistant, not a replacement
Continuum captures patient check-ins between sessions and synthesises them into actionable highlights.
Therapists start each session with deeper context while clinical privacy stays intact: patients decide what is shared, and shared data always shows its origin.
10
Takeaways / Reflection
What building Continuum taught me
Balancing complex clinical data with extreme UI simplicity required ruthless prioritisation.
- Show information only when relevant. Everything else is a click away.
- Trust is a design surface. Labels, provenance and edit controls did more for adoption than model quality.
- Technology should support empathy, not compete with it for attention.
11
Next Steps
Where Continuum goes from here.
Continuum is entering a longer beta with practising clinicians.
The next design cycle focuses on group therapy, family sessions and longitudinal reports.