Cargo® DemoTest Index p.(1-21)Email, Instagram
Cargo® Demo

Praesent bibendum a elit non efficitur. Etiam et velit vitae quam aliquet fermentum. Quisque sit amet nulla dictum, cursus erat a, tempus est. Sed laoreet, magna eget mattis tincidunt, felis tellus vestibulum augue, in lobortis mauris risus eu nunc. Vivamus sed condimentum leo. Nunc efficitur, magna condimentum gravida egestas, ligula mauris posuere neque, sit amet sagittis dui velit vitae neque. Donec in dapibus arcu. Aliquam efficitur suscipit magna, vel pretium mi. 
rakan.site
Rakan AlkhattabiUser experience designer Three cases on the wall. Click a poster to read one.
Rakan AlkhattabiPlatform02Patient Kiosks← Back to the platform

02
Patient Kiosks
Service design

Self check-in for patients who are tired, rushed or unwell.

Case 02rakan.site
Scroll to read ↓

Patient Kiosks

Everything that makes an app easy is absent in a corridor. One user at a time, a queue behind them, and a machine most of them rarely use.

  • Role: Lead, design & research
  • Team: Business analyst, developers, stakeholders
  • Constraints: Hospital IT and security rules, existing hospital systems, a fixed launch date
  • Year: 2025
  • Platform: Kiosk, landscape touchscreen
  • Type: Service design
  • Impact: 70 kiosks live, 2 ways to check in, 00:20 to a clean reset

① At a glance

The problem. Check-in happens in a public corridor, one person at a time, with a queue behind them and a machine most of them rarely use.

What I did. Designed the whole check-in, not only the screen: the queue token, the printed armband, the handoff to the reception desk, and a 00:20 reset.

What changed. Live on 70 kiosks in Arabic and English as mirrored layouts, with the desk kept for the cases that need a person.

70kiosks live
2ways to check in
00:20to a clean reset
Idle, the language choice
English and Arabic, one tap
ONE WELCOME SCREEN, ENGLISH OR ARABIC IN ONE TAP

01 · context

② The queue is part of the interface

Someone walks up to the row of kiosks and stops. They do not touch anything. They look at the machine, then at the reception desk, then at the person behind them, and after a few seconds they join the desk queue instead.

That is the failure, and no usability metric catches it. The kiosks stand in public areas at KFSH&RC to take work off a reception desk. The person in front of one is standing, often unwell or accompanying someone who is, with people waiting behind them. And the interaction does not end on screen. It ends with a printed token and a band around a wrist.

So the problem was never the check-in screen. It was the hesitation before it, the queue during it, and the hardware after it.

One check-in spans four things. Only one of them is a screen
ArriveFinds the kiosks, unsure it is the right machine
ScreenIdentifies, picks the appointment, checks in
Token slipA queue number, printed
ArmbandPrinted, worn, carried to the waiting area
The highlighted step is the only part a screen design normally covers
THE KIOSK IN CONTEXT: A QUEUE BEHIND, A PRINTER BELOW, A DESK STILL OPEN
METHODMEASUREWHEREWHEN
Timed observation at the kiosks, liveDwell per person and queue growth, not task successPublic corridors, KFSH&RC Riyadh2025, before launch

Dwell was the number that mattered. Two behaviours dominated: hesitation before touching anything, and walking away mid-flow.

RESEARCH: WHAT WAS MEASURED AT THE KIOSKS, AND WHY IT WAS NOT TASK SUCCESS
STAGEARRIVEIDENTIFYCHECK INWAITLEAVE
DOINGFinds the kiosks, unsure it is the right machineMedical record number if they know it, finger if they do notConfirms details, standing, at about arm’s lengthStands while a token and an armband printTakes the token and the band and looks for the waiting area
WHERE IT BREAKSHesitates, asks the person behind, or joins the desk queue anywayFingerprint reads often fail, and failure feels personalOne extra decision per screen and the line behind growsSilence reads as failure. A print takes seconds with nothing on screenWalks away mid-flow and leaves a record on a screen in a public corridor
OPPORTUNITYA bilingual welcome that asks for one thing first: the languageTwo ways in, and finger choice offered up front rather than after a failureWide landscape screen, one decision per screen, action where a hand restsExplicit wait states: the machine says what it is doingA visible 00:20 inactivity countdown and a full clear on expiry
CURRENT STATE: ONE CHECK-IN, FIVE STAGES. ONLY ONE OF THEM IS A SCREEN

02 · approach

③ Design the machine and the queue, not only the screen

WHAT WE SAWWHAT IT MEANTWHAT CHANGED
People stood at the kiosks without touching anything.The machine did not announce what it was for.A bilingual welcome that asks for one tap, the language, before anything else.
Fingerprint reads failed and people retried the same finger.Recovery was invisible, so failure looked like rejection.A finger-selection diagram, offered before the failure rather than after it.
Sessions were abandoned halfway, records left on screen.A public screen holds a stranger’s record until someone clears it.A visible 00:20 countdown and a complete reset on expiry.
People waited through silent prints and assumed a fault.The hardware was invisible to the person relying on it.Explicit wait states for the token and the armband.
SYNTHESIS: FOUR OBSERVATIONS FROM THE QUEUE, AND THE DECISION EACH ONE FORCED
before the screen

The corridor, the queue, the walk-away

01Measure dwell, not task success

Live observation at the kiosks, timing how long one person occupies a machine and watching what makes the line behind them grow. Task success would have scored this product well. Dwell told the truth. Two behaviours dominated: people unsure they were even in the right place, and people who walked away mid-flow and left their record on a screen in a public corridor.

Plan view: row of kiosks, main corridor
Kiosk
Kiosk
Kiosk
Reception desk
  • Dwell, not task success: how long one person holds the machine
  • “Am I even at the right machine?”, the pause before the first touch
  • Walk-aways that leave a medical record on screen in a public corridor
Dots are people waiting. The desk stays open for cases that need a person
QUEUE AND DWELL: WHAT MAKES THE LINE GROW

02A kiosk is four machines in one cabinet

A screen, a fingerprint reader, a token printer and an armband printer. Half the design is telling someone what the cabinet is doing while nothing is happening on screen. A print takes seconds, and seconds of silence read as failure to a person who is already unsure.

Elevation: one cabinet, four things to explain
ScreenWide landscape, read standing at arm’s length
Fingerprint readerThe common failure point. Recovery is a different finger
Token slipQueue position, printed first so the place is secured
Armband printerFour to six seconds of silence that has to be narrated
The screen is one of four outputs. Three of them are physical
THE CABINET: SCREEN, READER, TOKEN SLIP, ARMBAND
LANEARRIVEIDENTIFYCHECK INPRINTLEAVE
PATIENTApproaches the kiosksEnters MRN or presents a fingerConfirms detailsWaitsPuts the band on
FRONTSTAGEBilingual welcomeTwo entry paths, finger diagramOne decision per screenToken number, then wait stateConfirmation and instruction
BACKSTAGEIdle loop, session clearedReader poll and matchRecord lookupToken printer, then armband printerSession clears, returns to welcome
EVIDENCEThe cabinet itselfA read or a retryThe details on screenA printed tokenA band around a wrist
SERVICE BLUEPRINT: WHAT THE PERSON DOES, WHAT THE SCREEN SHOWS, WHAT THE CABINET IS DOING
at the screen

Standing, watched, rarely used

03Wireframe at standing distance, not desk distance

The screen is a wide, short canvas read at arm’s length by someone on their feet. Low fidelity first: one decision per screen, the action placed where a right hand already rests, and the hospital information rail running the length of the flow. Every extra decision per screen is a person added to the line. Text and buttons are sized to be read and pressed while standing, and every line is written in plain Arabic and English. The service menu offers more than check-in, but check-in is what most people come for, so this case follows it.

LOW-FIDELITY: ENTRY, CAPTURE, HANDOFF
LOW-FIDELITY: ENTRY, CAPTURE, HANDOFF

04Two ways in, because readers fail

Medical record number for people who know it, fingerprint for everyone else. The finger-selection diagram exists because fingerprint reads often fail, and the recovery is to try a different finger. So the interface offers that choice up front rather than after a failure, because a failed read in public does not feel like a technical fault. It feels personal.

Medical record number
Medical record number
Finger selection
Finger selection
Capture
Capture
Success
Success
CHECK-IN: THE MRN PATH, THEN THE FINGERPRINT PATH FROM SELECTION TO SUCCESS

05A visible countdown is a message to the queue

A 00:20 inactivity timer runs in the header and restarts with every touch. It is visible rather than hidden, because the person waiting behind can read it too. On expiry the session clears completely and returns to the bilingual welcome. No half-finished state, and no previous patient’s record left on a screen in a public corridor. Every screen in the flow runs fully mirrored in Arabic.

Service menu in English, 20 second countdown
Service menu in English, 20 second countdown
The same menu in Arabic, fully mirrored
The same menu in Arabic, fully mirrored
IDLE AND RESET: ONE MENU IN ENGLISH AND ARABIC, WITH THE COUNTDOWN
after the screen

A slip, a band, and a desk still open

06The last screen is about a physical object

Token number first, so the person has their place in the queue before anything else. Then the armband print with an explicit wait state. Then a confirmation that says what to do with the thing in their hand: put it on, go to the waiting area. The screen is finished before the visit is.

Token printed
Token printed
Armband printed
Armband printed

03 · outcome

④ Checked in without waiting at the desk

Live on 70 kiosks in public areas across the hospital, in Arabic and English as fully mirrored layouts. The person from the opening, the one who looked at the machine and joined the desk queue, now gets one question first: which language. Two ways to identify, a countdown the queue can read, and a token and armband that print with the screen saying what is happening. The reception desk keeps the cases that actually need a person.

Results are being measured now, including dwell time and the share of check-ins done at a kiosk instead of the desk, and will be added here soon.


04 · what I’d do differently

⑤ What I’d do differently

Test it standing, with three people waiting behind. Live observation showed the queue, but every prototype session I ran was seated and unhurried, which made the interface look calmer than it is. The queue is the hardest part of this product to simulate and the part that decides whether anyone uses it.

Next stop

Case 03 · Discharge Summary Generator

Nobody trusts a machine to write a clinical document. So the tool only drafts: thirteen sections, each traceable to the note it came from.


03
Discharge Summary Generator
● AI tool

Case 03→
Rakan AlkhattabiPlatform03Discharge Summary Generator← Back to the platform

03
Discharge Summary Generator
AI tool

Turning clinical notes into a summary people can act on.

Case 03rakan.site
Scroll to read ↓

Discharge Summary Generator

Nobody trusts a machine to write a clinical document. So the tool only drafts: thirteen sections, each traceable to the note it came from.

  • Role: Lead, design & research
  • Team: Business analyst, developers, stakeholders
  • Constraints: Patient data and security rules, integration with ICIS, a fixed pilot date
  • Year: 2026
  • Platform: Web: ICIS integrated
  • Type: Clinical AI
  • Impact: 73 → 11 notes in one encounter, chosen by the clinician, 13 sections, each traceable, 3 steps to a draft

① At a glance

The problem. The least interesting writing in medicine sits on the critical path, and nobody trusts a machine to produce a clinical document.

What I did. Built a tool that drafts thirteen sections from notes the clinician chooses, each section traceable to the note it came from.

What changed. In pilot. The clinician stays the author, and an empty section says so instead of guessing.

73 → 11notes in one encounter, chosen by the clinician
13sections, each traceable
3steps to a draft, 5 to save
The generated summary
Scroll
The generated summary, with its sources

01 · context

② The least interesting writing in medicine, on the critical path

End of a shift. One clinician, one encounter, seventy-three notes to read back through before a word gets typed. The patient is ready to go home. The discharge cannot complete until the summary exists.

It gets typed from scratch in ICIS, the hospital’s clinical records system, or pasted together from earlier notes, which carries their errors forward. It varies by whoever happens to write it. And it is the last thing anyone wants to do at the end of a long day.

Writing quality was fine. What hurt was that this writing sits on the critical path, and that nobody trusts a machine to take it off.

One encounter, one clinician, end of a shift
73notes on this encounter
13sections to fill
0of them written yet
Every note on the encounter. The eleven that matter are in here somewhere
Finding them is the work. Writing is quick once they are found
WHERE THE TIME GOES: READING BACK, RETYPING, AND THE WAIT THAT FOLLOWS
METHODWITHWHEREWHEN
Contextual inquiry, sitting through the task6 clinicians writing discharge summariesOn the ward at end of shift, KFSH&RC Riyadh2026, before the pilot

The reading is the work. Typing is quick once the facts are found, so the tool had to earn its keep at retrieval, not composition.

RESEARCH: WATCHING THE SUMMARY GET WRITTEN, RATHER THAN ASKING ABOUT IT AFTERWARDS
STAGEOPEN THE ENCOUNTERREAD BACKASSEMBLECHECKSAVE
DOINGOpens an encounter that can hold dozens of notesReads back through the notes to find the factsTypes from scratch in ICIS, or pastes from earlier notesRereads for anything missing or wrongSaves, and the discharge can complete
WHERE IT BREAKSNo sense of how much reading is aheadThis is the real cost, and it is invisible to everyone elsePasting carries earlier errors forward with themNothing marks which claim came from which noteUntil it exists the discharge cannot complete
OPPORTUNITYShow the note count and let the clinician choose the source setThirteen named sections, each one sourced separatelyGenerate a draft from the chosen notes, never from all of themView Source on every section, opening the notes beside the draftRegenerate per section, and nothing reaches ICIS without a person
CURRENT STATE: WRITING ONE SUMMARY. THE READING IS THE WORK, AND IT SITS ON THE CRITICAL PATH

02 · approach

③ Draft it, but never let it speak for the clinician

WHAT WE SAWWHAT IT MEANTWHAT CHANGED
Clinicians spent the time reading, not writing.The bottleneck was retrieval, not composition.The tool intervenes at the source set, before a word is generated.
Summaries were pasted together from earlier notes.Errors propagated forward invisibly.Per-section provenance, so a claim can be traced to its note.
Nobody trusted a machine to write a clinical document.Trust needed to be earned per claim, not per document.Thirteen fields instead of one blob, each editable and regenerable.
A confident paragraph with no source behind it was the fear.A plausible invention would end the pilot.“Not found in the selected notes”, with a manual field instead.
SYNTHESIS: FOUR OBSERVATIONS AT THE DESK, AND THE DECISION EACH ONE FORCED
before the draft

Seventy-three notes, and the eleven that matter

01Watch someone write one

Sitting with clinicians through the actual task, not asking about it afterwards. The reading is the work. The typing is quick once the facts have been found. That one observation moved the whole product: the tool had to earn its keep at retrieval, not at composition.

How one summary gets written today
01Read backthrough 73 notes on the encounter
02Retypethe same facts into ICIS by hand
03Or pastefrom an earlier note, errors included
04Dischargecannot complete until it exists

The reading is the work. The typing is quick once the facts have been found.

The last step of an admission, and the one on the critical path
THE OLD PATH: WHERE A DISCHARGE SUMMARY ACTUALLY COMES FROM

02Thirteen fields, not one blob

A field can be empty, sourced, regenerated or edited on its own. A blob can only be accepted or rejected whole, and nobody accepts a whole clinical document from a machine. Almost every other decision in this product follows from that one.

Thirteen sections, each with its own source, state and regenerate
Date of admissionAlways
Attending physicianAlways
Principal diagnosisAlways
Secondary diagnosisOften
Past medical historyOften
ProceduresSometimes
ConsultationsSometimes
History of present illnessOften
Physical examinationSometimes
Condition on dischargeSometimes
Discharge instructionsSometimes
Follow-upSometimes
Discharge medicationsOften
Always presentOften presentSometimes empty, and allowed to say so
THE SECTION MODEL: THIRTEEN FIELDS, EACH WITH ITS OWN STATE
STEP1 · ENCOUNTER2 · SOURCE SET3 · GENERATE4 · REVIEW5 · SAVE
CLINICIANOpens patient and encounterPicks which notes feed the draftWaitsReads, edits, regenerates a sectionPresses Save
THE TOOLShows 73 notes, grouped by typeShows a running count and a time estimateDrafts thirteen named sectionsOffers View Source on each sectionSaves the approved summary into ICIS
GUARDRAILNothing is chosen silentlyCommit is explicitOnly the selected notes are usedEmpty sections say “not found”Nothing saves without a person
DECISION FLOW: WHERE CONTROL STAYS WITH THE CLINICIAN, AND WHAT THE TOOL IS NOT ALLOWED TO DO
the draft

Thirteen sections, not one blob

03Wireframe the argument, not the output

Patient and encounter, then notes, then review. Low fidelity, because the argument was about how much control to hand over before generation, not about how the output looked. The output was never the contested part.

LOW-FIDELITY: ENCOUNTER, NOTES, REVIEW
LOW-FIDELITY: ENCOUNTER, NOTES, REVIEW

04Make the source set a decision, not a default

Seventy-three notes on one encounter. The clinician picks which ones feed the summary, grouped by type, filterable, with a running count and a time estimate before they commit. A tool that silently chose for them would be faster to use and far harder to trust, and this one only works if it is trusted.

Select notes
Scroll
The source set, chosen by the clinician
Find patient by MRN
Scroll
Step 1 · Find patient by medical record number
Patient and encounter
Scroll
Step 1 · Patient and encounter
Select the notes
Scroll
Step 2 · Select the notes to summarise
SELECT NOTES: 11 OF 73, GROUPED, WITH AN ESTIMATE
after the draft

Provenance and refusal

05Provenance per section, because that is the unit of disagreement

View Source sits on each of the thirteen sections and opens the notes that produced it beside the draft. Per-document provenance would be easier to build and useless in practice. A clinician does not disagree with a summary. They disagree with one line in Medications.

Review summary
Scroll
The draft under review
Generating
Scroll
Step 3 · Generating
Summary with sources
Scroll
Step 4 · The summary, with the note each line came from
REVIEW: PER-SECTION SOURCE, COPY AND REGENERATE

03 · outcome

④ A draft in three steps, with its sources attached

In pilot with a small number of departments. The clinician from the opening no longer reads back through seventy-three notes. They choose the eleven that matter, get a draft in three steps, and check each section against the note it came from. Where the notes are silent, so is the tool. Accuracy is checked by clinicians in the pilot: every draft, section by section, against its source notes, before it is saved into ICIS.

Results are being measured in the pilot now, including time to a finished summary, and will be added here soon.


04 · what I’d do differently

⑤ What I’d do differently

Design the refusal before designing the output. We built the generation first and the not-found case second, which meant the empty state arrived after the pattern for a full one had already set. On any tool that writes into a clinical record, the sentence it says when it does not know is the load-bearing one, and it should be drawn first.

Next stop

Case 01 · KFSH Employees

A hospital runs on shifts. The old app was organised by department, so 16,000 staff kept leaving it to get things done.


01
KFSH Employees
● Mobile app

Case 01→
Rakan AlkhattabiPlatform01KFSH Employees← Back to the platform

01
KFSH Employees
Mobile app

The staff app for a hospital that never closes.

Case 01rakan.site
Scroll to read ↓

KFSH Employees

A hospital runs on shifts. The old app was organised by department, so 16,000 staff kept leaving it to get things done.

  • Role: Lead, design & research
  • Team: Business analyst, developers, stakeholders
  • Constraints: Hospital IT and security rules, existing HR and attendance systems, a fixed launch date
  • Year: 2025–now
  • Platform: iOS & Android
  • Type: Product design
  • Impact: 16K+ employees on it, 6 tasks moved into the app, 40+ screens shipped

① At a glance

The problem. A hospital runs on shifts. The employee app was organised by department, so six of the most common tasks happened somewhere else.

What I did. Rebuilt it around the shift. Attendance, credentials, HR self-service, the directory, oncall escalation and a policy assistant, behind one login.

What changed. Live on iOS and Android with more than 16,000 employees on it. The walk to HR for a payslip is a tap. A missed 02:40 punch can now be fixed in the app.

16K+employees on it
40+screens shipped
6tasks moved into the app
Home
Scroll
Home
PolicyGPT
Scroll
PolicyGPT
Digital badge
Scroll
Digital badge

01 · context

② The app opened. Then the walking started.

Someone comes off a night shift and opens the app to correct a punch they missed at 02:40. There is nothing in there for it. The correction is an email to a manager, the manager reads it in the morning, and the fix lands on a payslip a fortnight later.

KFSH&RC already had an employee app. It was dated and partial. You could open it, look at a little, and then still go somewhere else for the thing you actually came for. A payslip. An employment letter. A policy answer. Proof of who you were at a locked door. A clinical workforce does not carry a laptop around a hospital, so every gap in the app became a phone call, a walk, or a ticket for someone in HR to close.

The app was not missing features. It was ordered by who owns a service rather than by when a person needs one.

Still done somewhere else
Payslipwalk to HR
Employment letterraise a ticket
Proof of IDplastic badge
Policy questionphone a colleague
Missed punchemail a manager
Who is oncall?printed rota
Six of the most common tasks: none of them in the app
METHODWITHWHEREWHEN
Shadowing and semi-structured interviewsClinical and non-clinical staff  11 staffKFSH2025

The day has a shape: arrive, prove who you are, work, need an answer, leave. The old app was ordered by department instead.

RESEARCH: HOW THE SHAPE OF A SHIFT WAS ESTABLISHED
STAGEARRIVEPROVE WHOWORKNEED AN ANSWERLEAVE
DOINGPunch in at a wall terminal, or in the old appCarry a plastic badge for doors and checksNeed a payslip, a letter, the rotaAsk a colleague what the policy saysPunch out, hope it registered
WHERE IT BREAKSA missed punch has no path. It becomes an email to a manager.Badge left at home means a locked door and a phone call.Each one lives elsewhere: an HR office, a ticket, a printed rota.The answer depends on who you ask, and carries no source.Same gap as arrival, discovered a fortnight later on the payslip.
OPPORTUNITYPunch in and out above the fold. Justify Absence gets its own route.The phone becomes the badge: door access, mask fit test, safety cards.One Home, quick access ordered by how often a thing is needed.PolicyGPT: answer first, source second, policy document is authority.The day’s record sits in the same place the punch does.
CURRENT STATE: ONE SHIFT, FIVE STAGES. FRICTION MARKED WHERE THE APP STOPPED AND THE BUILDING STARTED

02 · approach

③ The unit of design is a shift, not a department

WHAT WE SAWWHAT IT MEANTWHAT CHANGED
Staff opened the app, then still walked to an office or phoned someone.The app was a directory of departments, not a container for the day.Five areas built around the shift, not the org chart.
Every person asked about policy asked a colleague first.The trusted source was a person, and people are not available at 3am.PolicyGPT, with the source document shown beside every answer.
A forgotten plastic badge stopped someone at a door.Identity was the one thing that could not fail, and it lived on plastic.Digital badge and door access, states drawn to be unambiguous.
A missed punch became an email to a manager, found weeks later.The error path was outside the product, so it was invisible and slow.Justify Absence as a first-class route, not a dead end.
SYNTHESIS: FOUR OBSERVATIONS, WHAT EACH ONE MEANT, AND THE DECISION IT FORCED
the shift, not the department

Where the day actually goes

01Start with the day, not the backlog

Shadowing first, then interviews, with clinical and non-clinical staff. The question was never what people wanted from an app. It was what a day contains: arrive, prove who you are, work, need an answer, leave. The old app served pieces of that, in an order that matched the org chart. Nobody’s day is ordered by department.

One shift, end to end
ArrivePunch inJustify a missed punch
IdentifyDigital badgeDoor access
WorkOncall escalationDirectoryReport incident
AskPolicy questionPayslip / letter
LeavePunch out
SHIFT MAP: THE TASKS A WORKING DAY ACTUALLY CONTAINS

02Five areas, each named for a moment

Home for the day, HR for the record, PolicyGPT for the question, Directory for the person, Oncall for the escalation. The six missing tasks land here: payslip and letter in HR, the badge and the missed punch on Home. HR keeps its name because a record is the one thing people look for by owner. Everything else sits on Home as quick access, ordered by how often a thing is needed. Ordering by owning department is easier to defend in a meeting and worse for everyone who has to use the result.

KFSH Employees
Home
  • Attendance
  • Punch in / out
  • Justify absence
  • Quick access grid
  • Report incident
HR
  • My information
  • Payslip
  • Letters
  • Leave balance
PolicyGPT
  • Ask a question
  • Answer + source
  • Generated-text notice
  • Error state
Directory
  • Search staff
  • Department
  • Contact card
  • Extension
Oncall
  • Today's rota
  • Level one
  • Level two
  • Escalate
Five areas in the bottom nav. Everything else reached from Home
IA MAP: FIVE AREAS AND WHAT SITS UNDER EACH
HOMEHRPOLICYGPTDIRECTORYONCALL
Punch in / outPayslipAsk a questionSearch staffToday's rota
Justify absenceLettersAnswer + sourceDepartmentLevel one
Quick accessMy informationGenerated noticeContact cardLevel two
Report incidentLeave balanceError stateExtensionEscalate

Everything else sits on Home as quick access, ordered by how often it is needed rather than by which department owns it.

FIVE AREAS INSTEAD OF A MENU: EACH ONE NAMED FOR A MOMENT IN THE DAY, NOT A DEPARTMENT

03Argue about sequence before anyone argues about colour

Low fidelity, in grey, on purpose. Attendance, the policy assistant and the credentials set were drawn and redrawn before any visual design began. Three flows carry most of the traffic, and the other thirty-seven screens follow whatever those three establish.

LOW-FIDELITY: ATTENDANCE, POLICYGPT, CREDENTIALS
LOW-FIDELITY: ATTENDANCE, POLICYGPT, CREDENTIALS
what they opened it for

Attendance, policy, credentials

04Attendance: the thing most people opened it for

Punch in and punch out sit above the fold with the time and the date. The missed punch, the case that used to become an email, gets its own route: Justify Absence. An error path that lives outside the product is invisible to the people who could fix it, and invisible means slow.

Home
Scroll
Home
Punched in and out
Scroll
Punched in and out
Justify absence
Scroll
Justify absence
ATTENDANCE: PUNCH IN, PUNCH OUT, JUSTIFY ABSENCE

05PolicyGPT: an answer you are allowed to check

An assistant over HR policy is only useful if people trust it, and trust here means being able to see where an answer came from. Answer first, source second, and a standing notice that the text is generated and the policy document is the authority. A one-time consent comes before the first question. I tested it with staff who ask policy questions, because the trusted source before this was a colleague, and colleagues are not available at 3am.

Answer and consent
Scroll
Answer and consent
Notice
Scroll
Notice
System error
Scroll
System error
POLICYGPT: ANSWER, SOURCE, AND THE STANDING NOTICE

06Credentials: the phone becomes the badge

Digital badge, door access, mask fit test records and safety reference cards in one place, designed for the conditions they get used in. One hand. A corridor. A locked door. Someone waiting behind you. The door access states are drawn to be unambiguous, because a maybe at a door is worse than a no.

Digital badge
Scroll
Digital badge
Door access
Scroll
Door access
Safety reference card
Scroll
Safety reference card
CREDENTIALS: DIGITAL BADGE, DOOR ACCESS STATES, SAFETY CARDS

03 · outcome

④ One app that holds the whole shift

Live on iOS and Android, with more than 16,000 employees on it. The night-shift worker from the opening no longer emails a manager. The missed 02:40 punch is justified in the app before the shift ends, and it reaches the payslip on time instead of a fortnight late. The payslip, the letter, the badge and the policy answer sit in the same place, behind one login.

Results are being measured now, including HR tickets and missed-punch corrections, and will be added here soon.


04 · what I’d do differently

⑤ What I’d do differently

Test the policy assistant with the people who answer policy questions today, not only the people who ask them. HR officers know which questions actually get asked and which answers get argued with. I tested with the askers and shipped something that answers them well. The people who would have caught the hard cases were down the corridor the whole time.

Next stop

Case 02 · Patient Kiosks

Everything that makes an app easy is absent in a corridor. One user at a time, a queue behind them, and a machine most of them rarely use.


02
Patient Kiosks
● Service design

Case 02→
Profile (sample)
Vivamus sed sapien fermentum, fermentum metus ut, efficitur nunc. Pellentesque sit amet ex sed est dapibus vestibulum. Donec ornare urna metus, ut mollis risus dignissim in. Phasellus tristique sed leo et dignissim. Nam lobortis vitae leo at finibus. Aliquam eget efficitur lacus.
                Mauris leo erat, placerat quis semper eu, pulvinar in est. Nam eget justo quis ligula faucibus pretium ac nec mi. Aliquam lacus est, auctor sed arcu at, luctus condimentum lectus. Quisque vitae eleifend quam, vel suscipit nisl. Duis efficitur est nunc, eget efficitur felis porttitor in. Donec facilisis ultrices mattis. Quisque blandit ac enim non accumsan. Donec euismod magna turpis, ut blandit purus tempor sit amet. Suspendisse et hendrerit mi. Integer gravida, sapien quis ullamcorper auctor, metus eros aliquet ante, ut dictum ipsum tortor eget elit. Maecenas eu rutrum dui, ut faucibus lectus. Vivamus porttitor bibendum tortor.`


Education

Vestibulum Euismod Ultrices (Congue)
Suspendisse Aenean Lectus
2023

Vestibulum Euismod Ultrices (Tincidunt)

Aliquam Faucibus Tellus
(3 years)


Employment

Vestibulum Ultricies
Suspendisse Aenean Vestibulum
2023 - 2024

Pellentesque Habitant
‘Morbi Tristique Senectus’
Nunc Euismod Faucibus,
Porta Libero
2024

Volutpat Ultricies
Curabitur Fringilla Ornare,
Duis Quis
(since 2020)

Accumsan Dui
‘Integer Consequat’
Purus Euismod Facilisis, Pharetra Massa
2022

TEMPUS, TEMPUS, TEMPUS
Aliquam Faucibus Tellus, Euismod Ornare
Tincidunt Quam
2021

Referrals

Tincidunt Nibh
Ultricies Vestibulum
Facilisis Ornare,
Porta Libero Universitas,
tincidunt.nibh@portolibero.edu.au
0491 570 006

Dictumst Gravida
Auctor Ligula / Founder
Aliquam Aliquam Congue, Felis
dictumst.gravida@gmail.com
0491 570 156

Rhoncus Mattis
Suspendisse Professor
Facilisis, Bibendum Ornare & Congue
rhoncus.mattis@facilisis.edu
0491 570 157

Sollicitudin Euismod
Founder
Pellentesque Mounds, Porta Libero
sollicitudin@vestibulum.com
0491 570 158

Egestas Convallis
Deputy Parturient Dictum
Duis & Ultricies Vestibulum
egestas.convallis@vestibulum.edu
0491 570 159

Skills

Vestibulum Ultrices
Congue Nisl
Tincidunt Euismod
Auctor Gravida
Molestie Vulputate
Tempor
Parturient
Nibh Ultricies
Aliquet Pellentesque
Rhoncus Tempor
Ullamcorper / Ornare
Porttitor Quisque
Lacinia
Bibendum Facilisis
Suspendisse Tristique
Viverra Dictum
Duis Convallis Et Congue
Semper Facilisis


Exhibitions

(upcoming) Tincidunt TBA
Ultricies Ultricies,
Congue Nisl
2024

(upcoming) Vestibulum-Vitae
Auctor Ipsum,
Dictumst Ligula,
Porta Libero
2024

Suspendisse
Molestie Aenean,
Facilisis Tellus,
Porta Libero
2024

Ultrices
Parturient Felis,
Pharetra Nibh,
Porta Libero
2023

Bibendum
Adipiscing Nunc,
Tincidunt Tristique Congue,
Vestibulum Porta Libero
2023

Eu & Fringilla, Collaborative
Imperdiet with Quam Sodales,
Donec Tristique Porta Libero
2023

Semper
Cursus Tempus,
Porta Libero
2023

Dictum, Elit. Vestibulum
Gravida Congue
2023

Rhoncus Congue
Ullamcorper Suspendisse,
Duis Ornare Porta Libero
2023


Awards

Tempus Egestas Nibh Vestibulum
Facilisis: Ultricies Ornare Ligula
2021

Rhoncus Auctor Ornare; Tincidunt Tristique, Porta Libero
2020


Press

‘Vestibulum Inceptos Gravida’, Ornare Review,
by Tincidunt Nisl
2024

‘Ullamcorper Faucibus (Congue) Euismod’, by Ligula Parturient, Bibendum Review Porta Libero
2023

Dictumst Tristique, Facilisis Magazine, Edition Two, Porta Libero
2021

Auctor Adipiscing Art Magazine, Porta Libero
2020

Sollicitudin Ornare Magazine,
Porta Libero
2020                  

Profile (Sample)

Vivamus sed sapien fermentum, fermentum metus ut, efficitur nunc. Pellentesque sit amet ex sed est dapibus vestibulum. Donec ornare urna metus, ut mollis risus dignissim in. Phasellus tristique sed leo et dignissim. Nam lobortis vitae leo at finibus. Aliquam eget efficitur lacus.
               Mauris leo erat, placerat quis semper eu, pulvinar in est. Nam eget justo quis ligula faucibus pretium ac nec mi. Aliquam lacus est, auctor sed arcu at, luctus condimentum lectus. Quisque vitae eleifend quam, vel suscipit nisl. Duis efficitur est nunc, eget efficitur felis porttitor in. Donec facilisis ultrices mattis. Quisque blandit ac enim non accumsan. Donec euismod magna turpis, ut blandit purus tempor sit amet. Suspendisse et hendrerit mi. Integer gravida, sapien quis ullamcorper auctor, metus eros aliquet ante, ut dictum ipsum tortor eget elit. Maecenas eu rutrum dui, ut faucibus lectus. Vivamus porttitor bibendum tortor.
       
Email, Instagram ↗


Education

Vestibulum Euismod Ultrices (Congue)
Suspendisse Aenean Lectus
2023

Vestibulum Euismod Ultrices (Tincidunt)

Aliquam Faucibus Tellus
(3 years)


Employment

Vestibulum Ultricies
Suspendisse Aenean Vestibulum
2023 - 2024

Pellentesque Habitant
‘Morbi Tristique Senectus’
Nunc Euismod Faucibus,
Porta Libero
2024

Volutpat Ultricies
Curabitur Fringilla Ornare,
Duis Quis
(since 2020)

Accumsan Dui
‘Integer Consequat’
Purus Euismod Facilisis, Pharetra Massa
2022

TEMPUS, TEMPUS, TEMPUS
Aliquam Faucibus Tellus, Euismod Ornare
Tincidunt Quam
2021

Referrals

Tincidunt Nibh
Ultricies Vestibulum
Facilisis Ornare,
Porta Libero Universitas,
tincidunt.nibh@portolibero.edu.au
0431 070 713

Dictumst Gravida
Auctor Ligula / Founder
Aliquam Aliquam Congue, Felis
dictumst.gravida@gmail.com
0404 965 983

Rhoncus Mattis
Suspendisse Professor
Facilisis, Bibendum Ornare & Congue
0415 843 581
rhoncus.mattis@facilisis.edu

Sollicitudin Euismod
Founder
Pellentesque Grounds, Porta Libero
0422 221 072
sollicitudin@vestibulum.com

Egestas Convallis
Deputy Parturient Dictum
Duis & Ultricies Vestibulum
egestas.convallis@vestibulum.edu
0402 254 956

Skills

Vestibulum Ultrices
Congue Nisl
Tincidunt Euismod
Auctor Gravida
Molestie Vulputate
Tempor
Parturient
Nibh Ultricies
Aliquet Pellentesque
Rhoncus Tempor
Ullamcorper / Ornare
Porttitor Quisque
Lacinia
Bibendum Facilisis
Suspendisse Tristique
Viverra Dictum
Duis Convallis Et Congue
Semper Facilisis


Exhibitions

(upcoming) Tincidunt TBA
Ultricies Ultricies,
Congue Nisl
2024

(upcoming) Vestibulum-Vitae
Auctor Ipsum,
Dictumst Ligula,
Porta Libero
2024

Suspendisse
Molestie Aenean,
Facilisis Tellus,
Porta Libero
2024

Ultrices
Parturient Felis,
Pharetra Nibh,
Porta Libero
2023

Bibendum
Adipiscing Nunc,
Tincidunt Tristique Congue,
Vestibulum Porta Libero
2023

Eu & Fringilla, Collaborative
Imperdiet with Quam Sodales,
Donec Tristique Porta Libero
2023

Semper
Cursus Tempus,
Porta Libero
2023

Dictum, Elit. Vestibulum
Gravida Congue
2023

Rhoncus Congue
Ullamcorper Suspendisse,
Duis Ornare Porta Libero
2023


Awards

Tempus Egestas Nibh Vestibulum
Facilisis: Ultricies Ornare Ligula
2021

Rhoncus Auctor Ornare; Tincidunt Tristique, Porta Libero
2020


Press

‘Vestibulum Inceptos Gravida’, Ornare Review,
by Tincidunt Nisl
2024

‘Ullamcorper Faucibus (Congue) Euismod’, by Ligula Parturient, Bibendum Review Porta Libero
2023

Dictumst Tristique, Facilisis Magazine, Edition Two, Porta Libero
2021

Auctor Adipiscing Art Magazine, Porta Libero
2020

Sollicitudin Ornare Magazine,
Porta Libero
2020                  
Top Cargo® Demo Site