Ward

Role

Product designer

Team

1 Product Manager, 2 Developers, 1 Designer

Timeline

Jan 2026 - May 2026

Skills

Interaction Design, System Design, UX Research

Context

Ward is a conversational AI designed to support coordination between home care workers and family caregivers.

Problem

Home care coordination relies on fragmented communication, verbal updates, handwritten notes, and memory across multiple caregivers.

We need a solution that can support caregiver coordination, task continuity, and communication.

Solution

Home care coordination conversational AI that supports communication, task continuity, and collaborative decision-making.

It introduces summaries, shift handoffs, reminders, and low cognitive load interaction for shared caregiving environments.

Care Overview

Task Coordination

Ward conversational AI notes and shift handoff screen

Notes & Shift Handoff

Reminder System

User Interviews

Interviews were conducted to better understand how home care coordination currently works.

Participants included home care workers and family caregivers who had varying levels of AI familiarity.

Findings From Interview

Relied heavily on texting, verbal updates, handwritten notes, and memory

Information frequently broke down during shift handoffs

Managing appointments, reminders, and tasks were a burden

Involved emotional labor beyond operational task management

Wanted tool that felt supportive rather than authoritative

Emphasized the importance of preserving human judgment

Information Architecture

The flow was designed to support tasks across multiple caregivers with shared visibility, task coordination, context persistence, reminders, appointment tracking, and escalation pathways.

The system was structured to support ongoing communication, caregiver transitions, and collaboration.

Ward information architecture diagram showing login flow, first shift scenario, and during-shift scenario with page, content, status, and button states

System & Interaction

System Design & Interaction Guidelines

Operating in emotionally sensitive and high-stakes environments, the interaction behavior itself needed to be intentionally designed.

This created several challenges:

How should the AI speak to caregivers?How much information should it provide at once?How could it reduce stress rather than increase it?

I created an interaction guideline system defining the AI's conversational structure, behavior, tone, and error handling.

Design Principles

Acknowledge user input, use affirmative language, remain supportive rather than authoritative

Conversational Structure

Acknowledgement layer (confirm user input), action layer (explain what the system is doing), support layer (offer contextual follow-up support)

Designing for Low Cognitive Load

Create short responses, use calm tone, give progressive disclosure, lightweight conversational pacing

Design

Initial Care Overview

This was designed to help new caregivers quickly understand information.

When beginning a shift, caregivers could ask, "What's the status of Ms. Rivera?"

The system responds with:

Current condition summariesPrevious caregiver notesEmotional contextSuggested priorities

Ward conversational AI screen showing the daily task overview flow, walking through a caregiver's first task and additional details

Daily Task Coordination

This was designed to guide caregivers through contextual conversational flows.

The system supported:

Meal coordinationMedication trackingAppointment schedulingTask progression

Ward conversational AI screen showing the daily task coordination flow

Notes & Shift Handoff

This was designed to help preserve continuity between shifts and reduce information loss

Caregivers could leave contextual notes for future caregivers:

Emotional observationsBehavioral patternsIncomplete tasksCare preferencesFollow-up reminders

Ward conversational AI screen showing the notes and shift handoff flow

User Testing

Prototype Evaluation

We ran moderated conversational walkthrough sessions where participants reviewed care summaries, corrected AI misunderstandings, logged tasks, left caregiver notes, managed reminders, and coordinated appointments.

Users Appreciated...

Immediate acknowledgements

Structured summaries

Reminder support

Users Wanted...

Shorter information summaries

Faster access to important updates

Clarity of complete/incomplete tasks

Iteration

Reducing Information Density

Before

Responses felt too long during caregiving tasks.

After

Shorter conversational summaries and progressive information disclosure.

Before

After

Prioritizing Important Updates

Before

Displayed all information with similar visual weight.

After

Prioritized urgent or time-sensitive updates first to improve scanability.

Before

After

Improving Task Visibility

Before

Completed and incomplete tasks were visually grouped together.

After

Clearer task status separation and visual hierarchy. Incomplete tasks notified through a reminder system.

Before

After

Results & Impact

Research submitted to ACM CHI

The resulting prototype was submitted to ACM CHI, contributing research on conversational AI as supportive infrastructure for human caregiving.

Reflection

Learnings

This project shifted my perspective from designing screens to designing AI behavior and conversational systems in high-stakes human environments.

Designing AI behavior, not just interfaces

I learned that designing conversational AI requires thinking beyond visual interfaces and focusing on how systems communicate, guide decisions, handle ambiguity, and build trust through interaction.

Designing for human-centered AI collaboration

Working on Ward reinforced that AI systems in caregiving environments should support caregivers rather than replace human judgment. Small interaction decisions around tone, uncertainty, and conversational pacing directly influenced trust, emotional comfort, and cognitive load.