Turn every request into a completed workflow. - enterprise SaaS workflow automation dashboard UI/UX case study
Designing an enterprise workflow automation platform that turns complex operational work into a clear, scalable experience.
01The product
Complexity was the raw material. Clarity was the design goal.
NexusFlow is an enterprise workflow automation platform built around one promise: a request goes in, a completed workflow comes out. Between those two points sit approvals, teams, analytics, reporting, AI-assisted insights, onboarding and administration.
I led the end-to-end UX and UI: every one of those surfaces, plus the design system that holds them together. The job was to make a deep product feel obvious from the first interaction, without taking control away from the people who live in it all day.
02The problem
Enterprise power creates enterprise friction.
Six kinds of people touch the same request, each with a different job and a different definition of “done”. The challenge wasn’t designing more screens. It was designing one system all six could understand.
The test for every screen
03The core idea
From request to resolution.
Before any screen, I designed the loop every request travels. Each stage has one owner, one visible status and one place in the product.
- RequestHas an ID, requester, category, priority and status from the first moment. Lives in Requests.
- RouteWork is categorized and sent to the right workflow and the system says what it did. Lives in AI routing.
- ApproveOne queue ranked by urgency, with approve and reject inline. Lives in the approvals queue.
- ExecuteAutomated steps report back as events with a source. Lives in Workspace Insights.
- TrackFour signals pair a number with a direction. Lives in the operational overview.
- ReportActivity re-expressed as performance for people accountable for it. Lives in Executive Reports.
- ImprovePolicy changes like SLA thresholds appear in the same feed as everyday work.
What the reports reveal feeds the next request.
Automation that explains itself is automation people trust.04The experience
One system. Different perspectives.
Same data, same components, same navigation. What changes is which question each screen answers first.
05Requests
The source of truth is a table.
Requests is the screen everyone eventually lands on, so it does the least decoration and the most work. Hierarchy and status carry the meaning.
- Status firstAll, Open, In Progress, Approved, Rejected are the top-level filters, because status is the first thing anyone asks about.
- A count you can trust“84 total” sits beside the title and pagination repeats it.
- Color plus wordsPriority and status are paired chips and every color carries a text label.
- One primary actionNew Request is the only filled button on the page.
06The command center
A dashboard that answers before it asks.
The goal wasn’t packing more onto the screen. It was a reading order: scan first, investigate second, following the order in which an operations lead actually thinks.
- ScanFour KPIs, each a number with a direction, put open work and pending approvals ahead of vanity totals.
- UnderstandVolume over 30 days and the category split explain the numbers.
- Get contextThe activity feed names people and actions, so changes have an owner.
- ActThe approvals queue is the only place that asks for a decision, last in the reading order on purpose.
The hierarchy survives compression.
On the marketing site the command center had to fit in one panel. The same order held: three signals on top, then the latest automated actions with their source.
- When a layout can shrink without losing its logic
- the logic is probably right.
07Executive visibility
From operational activity to business insight.
Executives don’t live in the workflow, so the report changes the language, not the system.
- Active requests
- → Total operations volume
- Avg. resolution time
- → SLA resolution rate
- Pending approvals
- → System escalations
- Activity feed
- → Department SLA breakdown
KPI overview, a 12-month trend, a category split and a department table: four ways to read performance, ordered from headline to evidence.
08AI
AI should reduce the work of understanding.
In NexusFlow, AI interprets incoming work so people don’t have to triage it by hand. It doesn’t make the decision. The design keeps those two jobs visibly separate.
09The design system
Consistency is a product feature.
Foundations were built so a new screen never meant a new visual language.
- Type
- Geist, from Display 48 to Overline 11, for high-density readability
- Elevation
- Four levels: shallow, floating, alert, modal
- Components
- Core and complex sets paired with the foundations page
10The details
Trust lives in the edges.
Enterprise software is judged outside the happy path. Sign-in and settings were designed as part of the product, with the same structure and calm.
Also designed in the same file: 404 and Access Denied states, a notifications panel and center, team roles and permissions, team invites and mobile responsive layouts.
11The design process
Designed in systems, not isolated screens.
The Figma file is organized by product area, not by presentation order.
- StructureMap the product into areas and the one navigation that holds them.
- UXGive every role a primary question and every key screen a reading order.
- UIHigh-fidelity screens on one shell: dark rail, light canvas, status chips.
- SystemPull repeated decisions into foundations and component sets.
- PrototypeConnect flows end to end, from sign-in to reporting.
- RefinementCover errors, notifications, roles, invites, settings, billing and mobile.
12Proof, designed
Confidence is a surface too.
For the marketing site, trust content was designed with the same rules as the product: a named person, a specific before-and-after and figures set as scannable cards using the KPI pattern from the dashboard.
The quote, name and figures shown are product content within the NexusFlow design, not measured results of this project.
13The outcome
Ten disconnected screens? One system.
Marketing site, sign-in, dashboard, requests, reports, settings and foundations: one shell, one set of tokens.
- Out-01 · Structure
- Requests have ID, owner, category, priority and status from the first moment.
- Out-02 · Visibility
- Every automated step reports what it did.
- Out-03 · Context
- Activity names people, queues rank urgency.
- Out-04 · Clarity
- Executives see performance in their vocabulary.
- Out-05 · Scale
- Foundations and components carry every new screen.
Complex products don’t need complicated experiences.