Skip to main content
Back to all work
№ 003 Product Design

From prompt to production.

FlowState is an AI experience builder that turns a plain-language idea into a working app, published with enterprise governance built in.

Role
Lead UX Designer
Timeline
Concept to v1.0
Team
PM, 4 engineers, AI platform
Platform
Web · Desktop
The brief

Good ideas waited on dev teams.

People across the business had tool ideas that died in backlogs. Building even a simple internal app meant a repo, a pipeline, and weeks of engineering time, so most ideas never made it past a document.

I led design for FlowState from zero to v1.0: the brand, the end-to-end flows, the component states, the accessibility annotations, and the developer handoff.

At a glance
The problem
Building internal tools required engineering time most teams did not have, so good ideas stalled.
My role
Facilitated design workshops and conducted competitive analysis, design workshops on to creating a concept through v1.0 handoff: flows, screens, states, accessibility, and brand.
The focus
Making building feel like describing: prompt in, working app out, governance handled by the platform.

Hide the pipeline, keep the trust.

The vision set a high bar: importing should feel like opening a file, editing should feel like a design tool, and publishing should feel like hitting export. Behind that one confident action the platform generates code, validates the build, and opens the pull request. My job was to hide the machinery without hiding what it does.

How it came together
01CONTEXT

From PRD to flows.

Allowing AI to analyze an attached PRD provides context so that it can build out according to the criteria provided to it and give a visual representation that hits the mark.
02COMPONENT COMPLIANCE

Using design system components

The beauty about this application is that it uses AI and logic in regard to our local design library components to compose screens and flows helping to minimize the cost of AI token spend.

03BUILD & TEST

Prompt in, app out.

Composed design experiences can be easily shared and tested with users and then revised and polished to get ready to for handoff to engineers to pull into their workstream. 

04QA & LAUNCH

Handoff, annotated.

The finalized output is a fully compliant design that is 1:1 with code, enabling engineers to run pull requests and prepare the build and push it live to production.

Publishing should feel like hitting export. The platform sweats the pull request, the validation, and the pipeline, so the builder never has to.

The bar we set for v1.0
The outcome

From described idea to deployed app.

V1.0 shipped to developer handoff with the full journey designed: prompt, canvas, preview, validation, and publish. These are the measures that will tell us it works.

AI that builds with your design system
It composes screens from approved components, so its output ships instead of stalling before production.
One tool, three very different users
Made for non-technical builders, wary engineers, and admins at once.
Democratized, but governed
Anyone can build — with review, audit, and sandbox-to-production guardrails built in.
What I’d carry forward

The craft that mattered most was invisible: states, empty screens, loading, validation, and accessibility annotations. AI gets you a draft in seconds. The details are what make it trustworthy.