WorkSortie

Sortie A project cockpit, flown by agents

A project cockpit that shows a product team who owns each piece of work, what happens next and what is blocking it.

Client
Own product
Role
Product designer and builder, solo
Type
Product design case study
Scope
Problem framing, Personas, Journey map, Information architecture, User flows, Wireframes, Design system, UI, Working app
Tools
Figma, FigJam, Claude Code, React, TypeScript
Status
Working app with a demo workspace. Five core screens designed in dark and light modes.
Year
2026

Screens show a demo workspace. Project names, tasks and people are sample data.

5
Core screens, designed in dark and light modes
6
Areas in the navigation, regrouped from about twenty screens
3
Autopilot levels per project: Suggest, Draft or Act
Solo
From problem framing to a working app
The Sortie Cockpit in dark mode: a greeting, four status counts, the ranked “What needs me today” list and the active projects table
Product film, 1 min 38 s, with music.

01 · Background

Built first, then designed properly

I built a working version first, to solve my own problem. Then I went back and did the product design properly: who it is for, what it must do, and how it should look. This case study follows that second pass.

The working Sortie app: the Daily Operations home screen with a progress dial, counts for high-priority, in-flight and completed work, the current focus and next step, and cards for the active projects
The working app with its demo workspace: Daily Operations, the home screen.View full size (opens in a new tab)
The working Sortie app: the Collaboration view, a network of people and projects around the current sprint, with a side panel listing what needs attention and who is most connected
The working app: Collaboration.View full size (opens in a new tab)
The working Sortie app: Mission Control, with counts for agents on duty and changes awaiting approval, an approval inbox, an autopilot level for each project and a risk radar
The working app: Mission Control.View full size (opens in a new tab)

02 · The problem

No tool held the connection between the people

Nobody on my team could answer “where is this project, and what happens next?” without asking three people.

I was a senior UI designer at KUBRA, carrying several projects across multiple products. Each project lived in pieces: the ticket in Jira, the design and its feedback in Figma, decisions in Slack, priorities with the PM, build status in an engineer’s standup update. Every tool did its own job. None held the connection between the people.

The company saw the problem and hired an operations manager to run projects, sprints and progress tracking. That helped the team leads. The designers, writers and researchers doing the work still pieced the picture together themselves.

03 · The patterns

Four patterns sat under every symptom

  • No owner of the connection. The designer became the human API between PM, engineering and stakeholders.
  • Status was pulled, not pushed. Information moved only when someone asked or a meeting was called.
  • Decisions had no memory. The reason behind a change lived in a thread nobody would find again.
  • The fix stopped at the lead level. Roll-ups reached managers. Clarity did not reach the people doing the work.

Published research suggests this is common. Asana’s 2023 survey of 9,615 knowledge workers found 58% of the day goes to “work about work”. Atlassian’s 2025 survey of 12,000 knowledge workers and 200 executives found a quarter of the week goes to searching for information.

04 · Who it is for

The people a process fix did not reach

Sortie is built first for the people a process fix did not reach: designers, writers and researchers.

PersonaRoleWhat they need
Maya Chen, primarySenior Product DesignerTo know what needs her today in under a minute, and to stop relaying status
Daniel Reyes, primaryUX WriterThe full context of a ticket, and the current design file, before writing a word
Aisha Rahman, primaryUX ResearcherTo know what is coming, and to see whether a finding changed a decision
Priya Nair, secondaryUX Team LeadTo see who is stuck and who is overloaded before a date slips
Marcus Bell, secondaryProduct ManagerTo know if design is on track without booking a meeting

Engineering leads, the operations manager and stakeholders read from Sortie more than they write to it.

These personas are composites drawn from my own experience on the team. They are assumptions to validate with interviews, not research findings.

05 · Process

From the problem outward

I worked from the problem outward, so every screen traces back to a named gap.

  1. 01

    Product brief

    Why it exists, the problem, the audience and the projected impact, with cited research.

  2. 02

    Personas

    Five personas in three tiers, plus three supporting roles.

  3. 03

    Journey map

    One project across seven stages, from request to status report. Today’s journey sits above the same journey with Sortie.

  4. 04

    Information architecture

    Every screen grouped into six areas: Cockpit, Projects, Sprints, People, Agents and Settings.

  5. 05

    User flows

    Daily check-in, new request to sprint, and agent proposal and approval.

  6. 06

    Wireframes

    Five core screens in greyscale, to settle layout before style.

  7. 07

    Design system

    Tokens, type, components, icons and logo, mapped to the app’s code.

  8. 08

    UI

    The five screens in dark and light modes.

Journey map of one project across seven stages, from intake to status. The top band shows what happens today and where it hurts; the bottom band shows the same journey with Sortie
Journey map: one project across seven stages. Today’s journey sits above the same journey with Sortie.View full size (opens in a new tab)
Information architecture diagram: Sortie branches into six areas (Cockpit, Projects, Sprints, People, Agents and Settings), each with its screens beneath it
Information architecture: every screen grouped into six areas.View full size (opens in a new tab)
User flow for the daily check-in: open Sortie, work down the ranked list by setting a next step, clearing a blocker or adding an update, then generate and share the standup update
User flow: the daily check-in, from opening Sortie to a shared standup update.View full size (opens in a new tab)
Five greyscale wireframes: Cockpit, Project page and Sprint planner on the top row, Collaboration orbit and Mission Control beneath
Wireframes: five core screens in greyscale, to settle layout before style.View full size (opens in a new tab)

06 · Key design decisions

Five decisions carry the product

Five decisions carry the product. Each answers one of the patterns above.

  1. 01

    Every active project must have a next step

    Jira’s “In progress” says nothing about what happens next. Sortie requires a next step with a due date on every active project, and flags any that are missing. On the project page, the next step and the blocker sit side by side at the top, above everything else.

  2. 02

    Status is pushed, not pulled

    The Cockpit opens on “What needs me today”: a ranked list with one action per row. The standup update is generated as a Slack post, an email or an executive summary, so nobody rebuilds status from memory.

  3. 03

    One record per project, beside Jira and not instead of it

    The record holds the ticket link, the design files, the team, the decisions and the handoff. Sortie links out to Jira and Figma. Replacing them would have meant asking a team to migrate, and the problem was never the tools themselves.

  4. 04

    Agents propose, people approve

    Upkeep is the reason trackers go stale, so agents do it. But an agent never changes the record silently. Each proposal shows a field-level diff with Approve and Reject. An autopilot dial per project sets how far agents may go: Suggest, Draft or Act. Every change can be undone.

  5. 05

    Show the team as a system

    The collaboration orbit puts one person or project at the centre, with direct collaborators on the inner ring. It surfaces overload, single points of failure and who work is waiting on. A lead sees this before a one-on-one, not after.

User flow for agent proposals: agents re-check the workspace after a change, the autopilot level decides whether a finding waits in Mission Control or is applied, a person approves or rejects the diff, and any applied change can be undone from history
Agent proposal and approval: how agents keep the record current without taking control away from people.View full size (opens in a new tab)

07 · Design system

One vocabulary for design and code

The design system and the code share one vocabulary: each semantic token carries the CSS variable it comes from.

  • Colour. Primitives feed semantic tokens for background, text, border and shadow. Components only ever use the semantic layer.
  • Two modes, one system. Dark pairs black surfaces with a single gold accent. Light pairs white surfaces with a blue accent. Switching mode changes token values, never components or layouts.
  • Type. Twelve Roboto text styles. Headings, labels and navigation are uppercase, which gives the interface its instrument-panel feel.
  • Components. Button, Chip, Avatar, Nav item, Input, Progress bar, Card with a content slot, Segmented control and Table row.
  • Icons and logo. Twenty icons taken from the app’s own icon font, and the Delta mark and wordmark from the app’s logo files.

One restraint shaped the look. Gold is the only strong colour in dark mode, so it always means “act here” or “this is active”. Red is reserved for blocked and at-risk work.

Sortie foundations on a dark background: background, text and border colour tokens each labelled with the CSS variable it maps to, twelve Roboto type styles, a spacing scale, corner radii and two elevation styles
Foundations: colour tokens with the CSS variable each maps to, then type, spacing, radius and elevation.View full size (opens in a new tab)
Sortie components on a dark background: primary, secondary and ghost buttons in default, hover and disabled states, four status chips, active and inactive nav items, an avatar, the command input and a progress bar
Components: buttons, chips, nav item, avatar, input and progress bar.View full size (opens in a new tab)
The Sortie Cockpit in dark mode: a greeting, four status counts, the ranked “What needs me today” list and the active projects table
View full size (opens in a new tab)
The Sortie Cockpit in light mode: the same layout and components on white surfaces with a blue accent
Same components, same layout, two modes.View full size (opens in a new tab)

08 · The screens

Five screens cover the daily loop for every persona

The Sortie Cockpit in dark mode: a greeting, four status counts, the ranked “What needs me today” list and the active projects table
View full size (opens in a new tab)

01 / 05

Cockpit

Start my day, four status counts, the ranked “What needs me today” list, and the active projects table.

  • Serves designers, writers and researchers

09 · Projected impact

A projection, not a measured result

The target is 4.4 hours saved per person per week, or 11% of a 40-hour week. This is a projection built on published research, not a measured result.

Where the time goesResearch baselineTarget reductionHours saved per week
Unnecessary status meetings2.8 h50%1.4
Searching for information10 h20%2.0
Reorienting after switching appsJust under 4 h25%1.0
Projected total per person——4.4

For a team of eight, that is about 35 hours a week. The figure sits under the 4.9 hours that workers themselves told Asana better processes would save. The reduction percentages are my own assumptions.

An early version helped my team see tickets, links and statuses more clearly, but the time it saved was never measured. A four-week pilot with one product team would replace the projection with real numbers: hours spent on status and searching, standup length, time to clear a blocker, and design questions after handoff.

10 · What I learned

Deciding what Sortie should not be

The hardest part was deciding what Sortie should not be.

  • Sit beside the tools, do not replace them. The gap was between people, so the product is a connecting layer and links out to everything else.
  • A rule beats a reminder. Requiring a next step did more than any alert asking for one.
  • Trust in agents comes from the diff. People accept automated upkeep when they can see exactly what will change and can undo it.
  • Building first has a cost. The working app grew features faster than structure. Doing the design pass afterwards regrouped about twenty screens into six areas.

11 · What comes next

Validate, measure, then extend

  • Validate the personas with interviews
  • Run the four-week pilot and measure the impact
  • Add sync so a team can share one workspace; data is local to one device today
  • Design the remaining screens and the mobile layouts
  • Bring the six-area navigation into the app

Sources

Want this kind of thinking on your product?