Designing Actionable Care Insights in a Healthcare Platform AI Agent

Turning complex behavioral data into clear, trustworthy decisions

 

Role

Solo Product Designer (UX/UI)

Collaboration

PM, engineers, data scientists, researchers, annotation team

Duration

6 months

Context

This is a healthcare SaaS platform supporting care managers in residential homes. The system uses AI (thermal sensors + skeleton tracking) to monitor resident behavior and detect potential health risks.

I led the UX/UI design within a cross-functional team of 30 over 6 months, owning the redesign of the insights & alerts dashboard.

 

Opportunity

The platform generated large volumes of behavioral data, but:

  • Insights were fragmented and data-heavy
  • Events were presented without context or prioritisation
  • Care managers struggled to distinguish urgent vs non-urgent and meaningful signals vs noise

As a result:

Dashboard usage was limited (2–3 checks/day)

Low trust in AI outputs

Delayed response to critical events

Poor handover between night and day staff

The opportunity was not only to improve the UI, but to transform how AI insights support real-world decision-making.
Key strategic principles:

minimum information, minimum time,
minimum action
= Maximum value

Minimum Information

Reduce cognitive load through clarity

Minimum Time

Enable understanding at a glance

Minimum Action

Build trust by filtering uncertainty

Maximum Value

Support, not replace, professional judgement

Strategy

| Single events don’t matter, but patterns do.

I positioned the product as a decision-support system, not a monitoring dashboard.

Instead of focusing on individual incidents, I reframed the system around:

  • Pattern recognition over time

  • Actionable narratives instead of raw data

  • Clear prioritisation to guide attention

Methods

| Translating Data into Narrative Insights

I worked closely with marketing and data teams to:

  • Refine how events are generated from raw sensor data
  • Align on thresholds for surfacing insights
  • Ensure the system adapts to individual behavioral differences

| Designing a Clear Prioritisation System

To reduce ambiguity, I classified:

  1. Alert → immediate action required
  2. Insight → emerging behavioral pattern
  3. Flag → backend review required

This enabled quick scanning and faster decision-making, while also improving signal quality and overall trust in the information.

| Stakeholder Alignment

There was concern around AI replacing professional judgement in medical diagnoses.

I facilitated workshops and created service blueprints to align stakeholders to:

  • Positioning AI as a “co-worker”
  • Ensuring final decisions remain with care professionals

Outcomes

Behavioral Impact
  • Dashboard usage increased from 2 hours/day to up to 9 hours/day
  • Shift from reactive checking to continuous monitoring
Efficiency Gains
  • Time to identify critical issues reduced from minutes to seconds
  • Improved clarity during shift handovers
COMPLIANCE & STANDARDS
  • WCAG AA Accessibility Compliance
  • Integration with Existing Care Systems
  • CQC Regulatory Alignment
  • Privacy-First Data Handling

Key learnings

This project taught me that simplicity in healthcare comes from managing complexity. Building trust is essential, especially in AI-driven systems, and many of the most important design decisions happen at the system level, not only in the interface.

It also highlighted a gap in my process. I realised the need to bring in more structured quantitative validation earlier, such as measuring response time, decision accuracy, and usage patterns.

Ultimately, strong UX decisions should be supported by both qualitative insights like trust and clarity, and quantitative evidence that shows real behavioral impact.

Scroll to Top