Allow admins to upload a CSV file containing up to 1,000 users. The system should validate each row, show a preview, and import valid users...
Flow
- 1.Admin uploads valid CSV file
- 2.System validates all 1,000 rows
- 3.Preview shown, admin confirms
Flow
- 1.Admin uploads CSV with missing fields
- 2.Rows flagged with inline errors
- 3.Option to fix or partial import
Flow
- 1.Non-admin user accesses import
- 2.Access blocked, error surfaced
- 3.Redirect to request access flow
Flow
- 1.Import submitted, server times out
- 2.Error state with retry option
- 3.Partial state preserved safely
The Problem
Every feature needs edge cases — but writing them is slow, repetitive, and often skipped under time pressure. Designers and product managers spend hours manually documenting what happens when things go wrong: validation failures, empty states, permission blocks, API timeouts, partial successes. These scenarios are critical for quality, yet they're the first thing cut when a sprint gets tight.
The result: edge cases discovered late in development, inconsistent UX across error states, and handoffs that leave engineers guessing what the design intent was.
What If AI Could Do This in Seconds?
"Turn a feature spec into review-ready edge cases with visuals and copy — instantly."
I designed and built UX Intelligence Partner — a tool that takes any PRD, FSD, or Jira ticket as input and generates a comprehensive set of edge-case scenarios, each with step-by-step user flows, system messages, recovery actions, and live screen previews built with enterprise UI components.
This is an AI-first approach to design productivity — not just using AI as a feature inside a product, but using it to reimagine how design work gets done.
Solo — Designed & Built End to End
I conceived, designed, and built this tool entirely on my own — using AI-assisted development in Cursor. This project is itself a proof of what an AI-first design approach looks like in practice: using Gen-AI not just to produce outputs, but to ship real, working software that makes the entire design team faster.
From the product concept and UX to the component architecture, API integration, and deployment — every part of this was built by a designer, with AI as the development partner.
Spec In. Scenarios Out.
The tool is deliberately simple to use — no setup, no training needed. A designer or PM can go from a raw spec to a full edge-case review in under a minute.
Paste your spec
Drop in a PRD, FSD section, or Jira ticket. Optionally add product context (user roles, platform type) to tailor the output.
AI generates scenarios
The AI analyses the spec and generates 6–8 typed scenarios: happy path, validation errors, API failure, permissions, timeouts, empty states, partial success.
Review with live UI
Each scenario card shows the full user flow, system messages, and recovery actions. Click "View UI" to open a live screen preview built from enterprise components.
Export & hand off
Copy the JSON spec directly, or export a formatted Markdown file — ready for Confluence, Jira, or a design handoff document.
Designed for Speed and Trust
The design challenge was to make an AI tool feel reliable and structured — not like a black box. Every decision was made to give designers and PMs confidence in the output.
Typed scenarios with colour-coded badges — each scenario is classified (Happy Path, Validation Error, API Failure, Permission, etc.) so reviewers can scan and prioritise at a glance, without reading every line.
Live screen previews from JSON — the "View UI" button renders actual enterprise UI components from the scenario data, not static images. This means every scenario has a visual that reflects its real state.
Graceful fallback to built-in demo — if no AI key is configured, the tool seamlessly shows a rich built-in sample instead of failing. This made it demos-ready at any time, in any environment.
Export for any workflow — Copy JSON spec for developers, or export Markdown for Confluence and Jira. The output works with existing team tools without requiring any process change.
Enterprise design system throughout — all UI components follow an enterprise MDS-inspired system: compact, accessible, no gradients. This ensures the tool feels native to the professional environments it's used in.
Also Shipped as a Cursor Skill
Beyond being a standalone web app, UX Intelligence Partner was packaged and published as a reusable Cursor agent skill — making it available directly inside the Cursor IDE as an AI capability that any designer or developer can invoke.
This means any team using Cursor can trigger the edge-case generator from within their existing workflow, without switching tools. The skill understands how to run the generator locally, extend its scenario types, add new UI components, and troubleshoot common issues — all through natural language in the AI chat.
This was a deliberate choice to make the work compound: not just a tool, but a reusable building block that others can extend and build on.
What This Changes
UX Intelligence Partner directly addresses the most consistently skipped part of the design process — and makes it the easiest part.
Hours reduced to seconds — what used to take a designer half a day to document manually now takes under a minute, freeing time for higher-value design work.
No edge case left behind — the AI consistently surfaces scenario types that humans miss under time pressure — permissions, partial successes, timeout recovery — building more complete, resilient designs.
A designer who ships — this project demonstrates that a designer with an AI-first mindset can conceive, build, and hand over a working, functional tool end-to-end — not just a mockup or prototype.
Reusable as a Cursor skill — packaged for the Cursor IDE so any team can invoke the generator directly from their AI coding workflow, multiplying its reach beyond a single deployment.