03TELATRIS

Building a startup MVP with AI-assisted design-to-code

FINALThe full customer workflow, running on a front end I designed and built.

Telatris is an early-stage SaaS startup building more efficient workflows for IT service providers. I joined as founding designer, a side project alongside my full-time role at ConnectWise, and took the product from vision to a coded front end the engineering team could build on.

0→1AI
RoleFounding Designer
TimelineJun–Dec 2025 · Jun 2026–Present
TeamFounders, 3 Engineers

Problem

Savings hide in spreadsheets

Our users are partners, the IT service providers who manage software costs for dozens of customers. Finding savings is slow work, done by hand in spreadsheets and admin portals. It's too complex and too sensitive to hand to junior staff, so experts do it themselves.
An advisor's Microsoft license spreadsheet, with customer names anonymized
An advisor's license report, anonymized.

Approach

Spreadsheet to prototype

Within a week, an advisor was reviewing a working prototype built on his own data. I used AI to turn that data into mock services, so the screens ran on realistic scenarios and edge cases surfaced early. He confirmed the structure, then pointed out friction we hadn't thought to ask about.
Coded prototype screen running on mock data
A coded prototype with mock data wired in, used to get honest feedback before engineering started.

Selected Work

Foundations to features

I built the foundations first: a component library and a lightweight brand. Everything after was faster to make, which meant more rounds of feedback and something credible to put in front of investors while the product was still moving.

1

Design System

Engineers needed a front end they could build on without waiting on design. I built a token-based React component library covering colors, typography, and UI patterns. I generated components in v0 and iterated in Claude Code, then pushed the library to the engineering repo so the team could adopt it directly. The library now lives in Storybook, where I iterate on components and push updates to the team.

The first local build of the library, since moved into Storybook.
2

Brand System

I designed a lightweight brand package to give the startup credibility from day one, structured enough to hold up across product screens and investor decks.

Stealth AI startup brand identity
Logo and icon in light, dark, and brand colorways, ready for product and pitch.
3

Onboarding Flow

I designed the first-run flow around bulk sync: a provider connects Partner Center and every tenant lands in one pass, with a manual path for the ones outside it.

The first-run flow, from an empty account to every tenant synced.
4

Alerts

I turned a metric-heavy dashboard direction into an alerts-forward hierarchy. Instead of digging through spreadsheets for savings, partners get timely tasks in the app, or in their own AI tools over MCP.

Alerts grouped into Act today and Deadlines, with an expired GDAP access alert open beside the list, showing how to fix it
Alerts grouped by urgency. Each one opens with how to fix it.

Outcomes

Ramp to release

I handed over a coded front end: real components from the design system, ready for the backend to connect to. Tables, alerts, toasts, each with usage notes for the engineers picking them up. The brand went with it, giving the founders something to work from in investor conversations. The product is still pre-release.

What I learned

AI has changed every phase of my process. Prototyping a product UI is easy now, yet designing a considered UX still requires the fundamentals of connecting the user's pain point with a valid business solution. Most of the build ran in v0 and Claude Code, and I got better results giving the model an outcome and constraints than giving it instructions. I'm still working through a systematic way of leveraging AI tools to stay true to my product design craft.