FD Fabian Delven
Case study
Product concept · Caregiving

Private
Nurse

A calm, human-centered caregiving system that helps families organize complex health information — without pretending to replace medical professionals.

Discipline
Product & systems design
Focus
Human-centered AI
Role
Concept, UX, visual
01 — Context

Context

Private Nurse is a caregiving coordination concept for families responsible for medications, appointments, symptoms, documents, provider communication, and daily care decisions.

Millions of people quietly run a second, unpaid job as the coordinator for a parent, partner, or child. The project focuses on reducing their emotional and informational overload — while staying firmly on the right side of the line that separates organizing from advising.

02 — The problem

Care information lives in too many places at once.

Family caregivers routinely manage important information across a scatter of sources:

Medication bottles Printed instructions Patient portals Appointment notes Text messages Insurance documents Personal memory

This fragmentation makes it hard to see what has changed, what needs attention, and what to raise with a provider.

03 — User & operational needs

What caregivers actually need.

See what changed

A single, trustworthy view of the latest state — not seven apps and a memory of a phone call.

Know what needs attention

Surface the few things that matter today, and let the rest stay quiet until it's relevant.

Walk into visits prepared

Arrive with organized questions and a clear brief, instead of remembering the concern in the parking lot.

Share without re-explaining

Let a sibling or partner step in with the same context, so care isn't trapped in one person's head.

04 — Before the solution

The workflow today.

STEP 1
Notice something

A symptom, a refill, a form.

STEP 2
Hunt for context

Across portals, texts, memory.

STEP 3
Improvise a note

On paper, in an app, or not at all.

STEP 4
Try to recall it

At the visit, under pressure.

RESULT
Things slip

Details lost, stress compounds.

05 — Design principles

Principles that shaped it.

01
Calm over complete

Show the few things that matter now. Everything else stays quiet.

02
Plain language always

No jargon, no cold clinical framing. Words a worried person can read at 2am.

03
Organize, never advise

Structure the information; leave the decisions to people who are qualified to make them.

04
Human warmth, not medical cold

Warm paper tones and soft type instead of the sterile dashboard.

05
Shared by design

Built for more than one caregiver from the start — not a single-user silo.

06
Trust is the feature

Every AI action is transparent and confirmable. Nothing happens silently.

06 — Proposed system

Nine calm, understandable spaces.

Everything a caregiver holds maps to one plain-language space — no dashboards to decode, no medical taxonomy to learn.

01
Medications
Doses, schedules, changes
02
Appointments
Dates, prep, follow-ups
03
Symptoms
What changed, and when
04
Provider questions
Held until the visit
05
Health history
Conditions and context
06
Documents
Records, forms, insurance
07
Care notes
Daily observations
08
Emergency information
Fast access when it counts
09
Family coordination
Shared, not siloed
07 — Role of AI

AI does the organizing — quietly, in the background.

Explaining medical information in plain language

Organizing questions for providers

Summarizing care notes

Preparing appointment briefs

Identifying information that may need clarification

Connecting related events in a care timeline

08 — Human confirmation points

The system organizes. A human always confirms.

Private Nurse draws a clear line between organizing information and providing medical guidance. Every consequential moment routes back to a person.

Confirm
Before it's saved

AI-drafted notes and summaries are reviewed by the caregiver before they become part of the record.

Confirm
Before it's shared

Appointment briefs and provider questions are approved by a person before they leave the app.

Refer out
When it's a decision

Anything resembling advice is redirected to the patient, caregiver, and a qualified professional.

Health decisions remain with the patient, the caregiver, and qualified healthcare professionals — the system never stands in for them.

09 — Key screens

Key screens & diagrams.

10 — Outcome

Human-centered AI that reduces the burden of caregiving — while preserving trust, clarity, and appropriate boundaries.

11 — What I learned

What I learned

The hardest problem wasn't the features — it was restraint. Caregivers don't need another dashboard; they need fewer things demanding attention.

Trust turned out to be a design material. Where AI touched anything consequential, the interface had to slow down and hand control back to a person — and that visible boundary is what made the whole concept feel safe.

12 — Next iteration

Next iteration

Test the confirmation flow with real caregivers to find where friction helps and where it just gets in the way.

Prototype the multi-caregiver handoff — how context passes cleanly between family members.

Explore accessibility for older and lower-vision users, who are often the caregivers themselves.

FD
Fabian Delven
Design & AI workflow — fabiandelven.com

Visual documentation

How the experience works.

A practical visual set for reviewing the workflow, responsibilities, system boundaries, and human-control points. Screen imagery remains upload-ready where final interface captures are not yet available.

Before-and-after workflow

From fragmented effort to one guided, reviewable sequence.

Search→Rebuild context→Act manually
Organize→Guide next step→Human confirms

User journey / service blueprint

The user journey aligned with visible system support.

User
ArriveUnderstandReview
System
GatherStructureRecord

Understandable system architecture

A high-level view of inputs, support logic, review, and output.

People + source material→Organizing layer→Review gate→Approved output

Important decision flow

Consequential actions pause for context and explicit confirmation.

Is the information complete?
No → request contextYes → human reviews
Approve, revise, or stop

Screens with annotations

Reserved for final product screens and concise callouts.

Future screen uploadAdd 3–5 interface captures with numbered annotations describing the user need, system response, and confirmation state.

Where AI helps

AI reduces repetitive synthesis without owning the decision.

Raw input→SortSummarizeSuggest→Proposal

Where humans confirm

People retain authority at every consequential boundary.

AI proposal→Check accuracy→Edit / approve→Use or share

What is intentionally not automated

The product supports judgment; it does not replace accountability.

No autonomous sending, publishing, or consequential decisionsNo silent changes to approved informationNo replacement for qualified human expertise