AI Experience Design  ·  2024–25
How do you design care without control?

AI Family
Support
System

A care product that gives adult children peace of mind about elderly parents — without making either of them feel surveilled.

RoleLead Product Designer
DeliverablesStrategy · Design System · Hi-Fi · Blueprint
UsersAnna (caregiver) · Jan (elderly)
AI layerInvisible · background only
Project question

How do you design
care
without control?

Most care technology answers the wrong question. It monitors people when it should connect them. It generates data when it should produce peace of mind. This project began with a different premise: the problem is not information — it is the absence of trust.

00 / Role & Contribution
Role
Lead Product
Designer
Concept Product
Responsibilities
Research & Synthesis
Service Design
Product Strategy
UX Architecture
UX / UI Design
AI Experience Design
Prototyping
Design challenge
"Build a product two people will trust for different, deeply personal reasons."
Anna needs reassurance. Jan needs dignity. The system must serve both simultaneously — without ever asking them to compromise.
01 / Problem

Most care technology
solves the wrong problem.

1 in 3
adult children report daily anxiety about elderly parents living alone
"The problem is not that Anna doesn't have enough information. It's that she has no way to turn information into peace of mind."
The real problem
Anna and Jan are close. Distance has made that closeness anxious. This product removes the anxiety — not the distance.
Current solutions feel clinical. Cameras, sensors, emergency buttons — they solve safety but destroy dignity.
Elderly users resist monitoring. They interpret surveillance technology as a signal that their independence is being questioned.
Caregivers spiral into compulsive checking. Without reliable context, every unanswered call feels like an emergency.
✦ Caregiver
Anna
Adult daughter · 42 · Lives 80km away · Works full-time
"I just need to know he had a good morning. That is enough."
◈ Elderly user
Jan
Retired engineer · 78 · Lives alone
"I am fine. I just like knowing she is there. I don't want her to worry about me every minute."
Most care products monitor people. This product protects relationships.
→

This reframing shaped the research agenda. The question became: what does the emotional experience of distance actually feel like — and what would it take to change it?

02 / Research & Key Insight

What we heard that
changed everything.

Interviews surfaced a consistent pattern: the emotional experience of distance is the core design problem, not the practical one.

“
I don't need to know everything. I just need to know he had a good morning. That's enough.
Anna — adult daughter, 42, lives 80km away
“
I am fine. I just like knowing she is there. I don't want her to worry about me every minute.
Jan — retired engineer, 78, lives alone
“
I check my phone every hour. If he doesn't pick up, my mind goes to the worst place immediately.
Research participant — daughter, 38, primary carer
F01
Anxiety scales with distance, not actual risk. Caregivers who lived further away reported higher daily anxiety regardless of their parent's actual health status.
F02
Dignity is the primary barrier to adoption. Every participant over 70 rejected products that felt like surveillance. The word "monitoring" caused immediate negative response.
F03
Voice messages outperform all other formats. Elderly participants rated voice messages from family as "the thing I look forward to most" — above phone calls, texts, or video.
F04
Routine, not incident, is the signal caregivers need. "If I know he had breakfast and went for his walk, I can get on with my day." Ordinary information is reassuring. Anomalies are alarming only in context.
The insight that changed everything
"The problem is not information.
It is the absence of peace of mind."
This reframing changed every design decision that followed. Two separate apps. One invisible AI layer. No companion persona. No health dashboard. A product that connects two people who already love each other.
The problem is not information. The problem is uncertainty.
→

The research didn't just clarify the problem — it dictated the architecture. If dignity was the barrier and routine was the signal, the system needed to be invisible to Jan and meaningful to Anna simultaneously.

03 / UX Architecture

Four principles that
made the hard calls easy.

Invisible by default
The AI has no interface, no name, and no relationship of its own. It operates entirely in the background, surfacing meaning without announcing itself. If Jan ever thinks about the technology, the design has already failed.
Anna · Jan
Dignity before safety
Jan will accept technology that feels like a natural part of his home. He will reject anything that makes him feel observed, managed, or diminished. Whenever dignity and safety conflict, dignity wins — because without dignity, Jan opts out entirely.
Jan-first
Peace of mind as a design metric
Anna's interface is measured against one question: does opening the app make her exhale? Alert inflation is a design failure. Clarity under cognitive load is a design success. If Anna checks compulsively, the product has failed — even if every alert was accurate.
Anna-first
The AI serves the relationship
This product is not about Jan's health data or Anna's notifications. It is about a daughter and her father. Every feature must exist in service of their connection — not as a separate digital relationship with a product.
Core principle
◈ Elderly user
Family
Companion
Ambient presence. Voice messages. Clock. No settings.
Passive
routine
AI LAYER — invisible
Context
→ Meaning
Pattern recognition · Natural language · No active data collection
Human
language
✦ Caregiver
Peace of
Mind App
Status hero · Signal tiles · Alert framework.
03 / System Architecture

One product.
Three layers.

This is not two apps — it is one system. The AI layer is the design challenge: how do you translate Jan's lived routine into language that gives Anna peace of mind?

Jan's world
Family Companion
Ambient tablet device. No configuration. No monitoring interface. A warm presence that shows family messages.
AI Layer — invisible
Context → Meaning
Passive routine sensing. Pattern recognition. Contextual summarisation. No active data collection.
Anna's world
Peace of Mind App
Morning status hero. Four signal tiles. Alert framework. 60 seconds to peace of mind.
Jan's experience
Never onboards. Never configures. His morning is simply his morning.
Jan makes coffee. Goes to the garden. The system reads the rhythm. Anna reads the meaning.
Anna's experience
Opens the app. Sees "Jan had a calm morning." Puts the phone away. Goes back to her day.
Alert Escalation Framework
L1 · Critical
Emergency
Hardware button pressed. No routine detected for 18+ hours with deviation from baseline.
Full-screen override · Haptic · Push · Phone call
L2 · Concern
Watch
Three consecutive missed breakfasts. Sleep pattern deviation over 4+ days.
Push notification · Persistent in-app card
L3 · Passive
Note
Weekly summary ready. Recommendation available based on recent patterns.
In-app card only · Optional push
L4 · Positive
Moment
30-day milestone. First garden walk of spring. Something worth celebrating.
In-app toast · Auto-dismisses · Never push
"A single missed breakfast is never an L2. Three consecutive missed breakfasts may be. Alert inflation destroys the trust that makes the whole system work."
Trust is designed through the system, not through individual screens.
→

The system architecture created a new design problem. How does the AI translate sensor data into language a daughter would trust? That question required an entirely different design discipline.

04 / AI Experience Design

Designing for
invisible intelligence.

The AI's job is to disappear. These three design decisions determined exactly how.

Decision 01
From sensor data to human language
Raw inference cannot reach Anna. Every data point must be translated into a sentence a daughter would actually feel. This is the AI's primary design challenge: producing meaning, not metrics.
Not this: Routine deviation detected. Activity sensors inactive 08:00–11:00.

This: Jan had a quiet morning at home. No activity in the garden — cooler today.
Decision 02
Confidence without false certainty
The AI never overstates what it knows. Summaries use qualified language when the signal is partial. A false reassurance is worse than no reassurance.
Not this: Jan is healthy and following his routine.

This: Jan's morning looks calm based on usual patterns. A quieter day than typical.
Decision 03
The AI never acts. Humans decide.
In every alert scenario, the AI surfaces information and stops. It does not call emergency services. It does not make care decisions. It alerts a human, then waits. This is both an ethical and a dignity decision.
AI role: "Jan's button was pressed at 14:32. Anna has been notified."

Human role: Every decision after that moment.
"The AI has no relationship with either user. It enables their relationship with each other."
This reframing changed everything downstream — no companion persona, no AI branding, no chatbot interface. The AI is infrastructure. The product is connection.
Input
Passive sensing
Movement, routine, device interaction — collected silently in the background.
08:20 · kitchen activity
08:47 · door opens
Compare
Baseline match
Today's pattern compared against Jan's established personal baseline.
Breakfast: on time
Walk: yes, 47 min
Interpret
Contextual inference
Partial signals qualified. Confidence level assessed. Nothing assumed.
Pattern: calm
Confidence: high
Translate
Human language
Data becomes a sentence a daughter would actually feel — not a metric she has to interpret.
"Jan had a calm morning. Walked in the garden."
Deliver
Peace of mind
Anna opens the app. Sees "All calm." Exhales. Puts the phone away. Goes to her meeting.
Anna closes the app.
The product succeeded.
The AI has no interface of its own. Its role is to translate routine into reassurance.
→

With the AI logic defined, the design challenge became surface-level. How do these principles manifest in two completely different interfaces — one for a caregiver scanning between meetings, one for a 78-year-old retired engineer?

05 / The Two Experiences

Two apps.
One family.

The contrast between them is intentional. Anna's app is designed for 60 seconds of meaningful scanning. Jan's interface is designed to disappear into his routine entirely.

✦ Anna — Peace of Mind App

"Open the app. Exhale. Put the phone away."

09:14
●●●●
1
All calm
Jan had a calm morning.
Walked in the garden · Breakfast at 8:20
Updated 12 minutes ago
8:20
Breakfast
47'
Outside
6.8h
Sleep
Calm
Pattern
Today's moments
🌿
Morning garden walk
08:42
☕
Breakfast routine
08:20
📻
Listening to radio
07:55
⌂
Today
📈
Patterns
✉
Messages
⚙
Settings
A
Status hero answers the question in one glance. "All calm" — the green badge is the product's core promise delivered before Anna reads a single word.
B
Four signal tiles, not a dashboard. Sleep, food, movement, pattern — the minimum meaningful context. No raw sensor data ever reaches this surface.
C
The pulsing sync dot is the AI made visible. The only moment in the entire product where the system announces its presence — and it does so quietly.
◈ Jan — Family Companion

"Jan never knows a summary was generated."

09:14
Thursday, a good morning
Good morning, Jan.
Anna is thinking of you. Your family is nearby.
A
Anna · voice message"Papa, thought of you this morning…"
Press and hold to reply with your voice
A
The ambient orb is the system's only visual language. It breathes gently, signals presence, never demands attention. No alerts, no numbers, no status indicators for Jan.
B
Voice-first communication respects Jan's comfort. Research showed voice messages were "the thing I look forward to most." The primary CTA is always a family voice message — never a form.
C
Jan never configures, onboards, or manages technology. The companion was set up by Anna. Jan's entire interaction surface is this screen. Nothing else exists for him.
05 / Service Blueprint

One morning.
The whole system visible.

The full blueprint covers 12 journeys. This is the one that defines the product: a quiet Tuesday morning, Jan's breakfast routine, Anna's 60-second check-in. Nothing dramatic. This is what the system is built for.

Scenario — Jan's morning routine · Tuesday · 08:15–09:30
Layer
Jan wakes & moves
Breakfast routine
Garden walk
Anna checks in
Jan
Wakes naturallyPuts kettle on. Radio plays. No device interaction.
Eats breakfastSits at the table. Morning as always. No action required from Jan.
Goes outside47 minutes in the garden. Companion screen shows time quietly.
Sees Anna's message"Papa, thinking of you…" He presses play.
AI layer
Passive sensing beginsMovement detected. Routine baseline comparison starts silently.
Breakfast confirmedPattern matched to baseline. Generates: "breakfast at 8:20."
Activity loggedDuration, time, context. Composes morning summary in natural language.
Summary delivered"Jan had a calm morning." Pushed to Anna's app.
Anna
Still asleepNo notifications. No early morning anxiety. The system handles it.
Morning commuteNo alerts. Nothing is wrong. Anna does not need to check yet.
Opens app (09:14)Status hero: "All calm." Green badge. Signal tiles confirm. She exhales.
Sends voice message"Papa, thinking of you this morning." Leaves for her meeting.
What the blueprint revealed that screen design could not: The most important moment in this system is not an interaction — it is Anna closing the app. The product succeeds when she stops engaging with it. That insight could only be seen in the blueprint.
Reassurance happens between touchpoints, not inside a single interaction.
06 / Prototype

Not just a story.
A real product to explore.

The prototype was built to test one question: does the status hero actually produce the feeling it is designed for? Participants were asked to check on a parent they hadn't heard from. We measured time, expression, and their first spoken words after opening the app.

"Jan had a calm morning."
Experience what it feels like to read that.
Anna App · Family Companion · Both perspectives inside
✦ Anna — Morning check-in flow
60 seconds to peace of mind
01
App opens to status hero — "All calm" badge visible before the screen finishes loading. No navigation required.
02
Signal tiles load below — four contextual data points confirm the summary. Each colour-coded by deviation from baseline.
03
Timeline shows the morning — optional deeper read. Anna can scan the moments of Jan's morning as a gentle narrative. Not data — life.
04
Send a voice message — one tap. Anna records. Jan receives it. The product becomes a channel for their relationship.
◈ Jan — Ambient moment + response
The product disappears into routine
01
Companion shows time + ambient orb — no notifications, no badges. The orb breathes. Jan's companion is always present, never demanding.
02
Anna's message surfaces gently — a warm card below the orb. "Anna · voice message." No alarm. No urgency. A presence, not an alert.
03
One tap to play — Jan hears Anna's voice. No headphones, settings, or configuration needed. It just plays.
04
Press and hold to reply — Jan holds the screen. Speaks. Releases. His voice message arrives in Anna's app with a gentle notification.
→
What testing confirmed: Participants who saw the green "All calm" badge stopped reading at that point in 7 out of 9 tests. The status hero worked. The signal tiles were used for reassurance-seeking, not primary communication. The prototype validated the hierarchy.
07 / Outcome

What this product
is designed to change.

<60s
Time for Anna to answer "Is Jan okay?" from app open
Design target — status hero
3×
Projected increase in family message exchanges vs. phone calls alone
Based on voice message research finding
0
Times Jan needs to understand or configure any technology
Design principle — Jan never onboards
+NPS
Primary metric: caregiver-reported peace of mind score
The product's core promise, measured directly
"The hardest outcome to design for is not engagement — it is relief. A product that succeeds when people stop using it."
The product's success metric is not sessions, time-in-app, or notification open rate. It is whether Anna checks compulsively or checks confidently. Those look identical in analytics. They feel completely different in life.

The full Experience Blueprint documents 12 user journeys, the complete alert logic, emergency flow design, and system architecture. Available on request.
Anna's experience
<60s
to confirm Jan is safe, from the moment she opens the app
0
unnecessary alerts — alert inflation is a design failure, not a feature
3×
projected family message frequency versus phone calls alone
Jan's experience
0
onboarding steps, configuration decisions, or settings to manage
1
interaction surface — the companion screen is Jan's entire product experience
∞
mornings lived as normal — the system reads context. Jan just lives his life.
A dual-user care ecosystem designed around dignity rather than surveillance.
08 / Reflection
The hardest design decision in this project was not visual. It was philosophical. Deciding what the product is not — not a monitoring tool, not a companion robot, not a healthcare dashboard — cleared the path for everything it is.
I would involve Jan-profile users earlier. Most of the research surfaced caregiver perspectives. The design decisions about the companion device were made largely by inference — what elderly users said they didn't want, more than direct testing of what they responded to. The next iteration begins there: six weeks with Jan.
I would also challenge the two-app model earlier. There is a version of this product where Anna's interface lives in a family shared space — a kitchen screen, not just a phone. The companion and the caregiver interface may not need to be separate products. That question deserves an answer before engineering begins.
Open tension 01
"Jan agreed to be 'connected.' But did he fully understand what that means? Consent in the context of cognitive change is a design problem we did not fully solve."
Open tension 02
"What happens when Jan has a genuinely bad day — and the system normalises it? The baseline is built over time. Early patterns may become invisible."
Open tension 03
"How does the AI learn what 'calm morning' means for this specific Jan, not a statistical Jan? Personalisation at the inference layer is the product's next design challenge."
The hardest challenge was balancing independence and reassurance — for two people who want different things from the same system.
09 / Visual Design System

Two visual languages.
One coherent system.

The design system was built to serve two fundamentally different emotional contexts — while remaining a single, coherent product. Every token was chosen to feel warm, human, and deliberately not clinical.

✦ Anna — Peace of Mind App
Primary
Soft
Accent
BG
Surface
Ink
Muted terracotta — warm, grounded, not alarm-adjacent. Nordic sage accent for positive signals. No clinical blues. No alert reds.
◈ Jan — Family Companion
Primary
Soft
Heather
BG
Surface
Ink
Nordic teal-grey — fjord, not corporate blue. Heather accent for warmth and familiarity. The palette references home, not hospital.
Display
Jan had a calm morning.
Cormorant Garamond
300 Italic · Display
Summaries, quotes, hero copy
App UI
Familje  Good morning, Jan.
Outfit 200–600
App headings, greetings, clock
Body
Walked in the garden · Breakfast at 8:20 · Pattern normal
DM Sans 300–500
Body copy, captions, sub-labels
Meta
Updated 12 min ago · L2 · Concern
JetBrains Mono 300–400
Timestamps, labels, codes
Status Hero — Anna App
All calm
Jan had a calm morning.
Walked in the garden · Breakfast at 8:20
Alert Level — L1 Emergency
L1 · Emergency
Jan pressed his button
14:32 · Notification sent to all contacts
Call Jan
Emergency
Ambient Orb — Jan Companion
Good morning, Jan.
Your family is nearby
✦ Anna's interface tone
Information-dense but scannable in 60 seconds
Data translated into human language — always
Alert hierarchy calibrated to emotional weight
Designed for moments of anxiety, not leisure
Success = Anna closes the app and exhales
◈ Jan's interface tone
Ambient, unhurried — designed to disappear
No numbers, alerts, or status indicators
Voice-first — interaction is a voice message
Feels like home, not a medical device
Success = Jan never notices the technology
Next project
Smart Zoo
Companion
Designing discovery across physical and digital environments. A spatial navigation system that transforms a zoo visit into a layered, personalised experience.
View Project →
Portfolio Justyna Awtuch
Interested in the full blueprint
or the design system? Let's talk.
Get in touch justyna.awtuch@gmail.com