Terr.ai
From light fittings to full renovations, Terri handles the search, the outreach, and the quotes. I built the product from 0>1.
What I Built
Complexities
The Outcome
Reflection
Home maintenance and repair is one of the most friction-filled consumer experiences that exists. Whether it's hanging a light, fixing a leak, or managing a full renovation, the process is the same: search, shortlist, call, wait, follow up, chase, compare. Repeat. Most people either give up, overpay for convenience, or spend hours doing work that shouldn't be their job.
Terr.ai was conceived to solve this at the root. Not by building a better directory, but by building an agent that handles the entire process on your behalf. Property owners and managers describe what they need. Terri takes it from there.
I designed the full Terr.ai product from zero, working within tight time constraints at an early stage of AI-native product development, building out a complete token architecture and component system in Cursor as part of the process.
The core user journey: a conversational intake flow where users describe their job, type, urgency, location, budget, in natural language. Terri processes the request, identifies and contacts relevant trusted providers, and returns a structured set of quotes for the user to review and act on. No calls. No chasing. No comparison paralysis.
For property managers handling multiple jobs simultaneously, I designed a dashboard layer that gave visibility across all active requests, status, providers contacted, quotes received, and actions required, in one coherent view.
The design system was built to support the AI-native interaction patterns the product required: conversational states, loading and thinking states, confidence indicators, and the particular UX challenge of helping users trust an agent acting on their behalf.
Designing for agentic AI is a fundamentally different problem from designing for traditional software. The user isn't completing a form or navigating a flow, they're delegating. That delegation requires trust, and trust requires transparency.
The hardest design question was: how much does the user need to see of what Terri is doing? Show too little and it feels like a black box. Show too much and the value proposition collapses, if users have to review every step, they might as well have made the calls themselves.
I designed around confident minimalism: clear confirmation of what Terri had understood, a lightweight status layer showing progress without demanding attention, and a structured quote comparison view that made the final decision easy without burying the user in detail.
Working at the early stages of AI-native product design also meant navigating genuine uncertainty about what was technically feasible week to week. Building a flexible token and component system from the start was essential, it meant the UI could adapt as the product's capabilities evolved without requiring full redesigns.
Terr.ai shipped as a live product with active users. The platform handles the full job lifecycle, intake, provider outreach, quote collection, and comparison, without the user making a single call.
For property managers, the time saved across a portfolio of properties is significant: what previously required multiple hours of calls and follow-ups per job is reduced to a description and a decision. The design system established during the zero-to-one phase provides a scalable foundation for the product's continued development.
Terr.ai pushed me into territory that most product designers haven't had to navigate yet: what does good UX look like when the product is an agent, not an interface? The answer I landed on, confident minimalism, transparent progress, structured decision points, feels like a design pattern that's going to matter a great deal as AI-native products become the norm.
Building the token architecture in Cursor at this stage of AI tooling was also a formative experience. The workflow between design intent and implementation has changed permanently, and being early to that shift has shaped how I approach every project since.