Homewise

What I did

• Led the voice of customer research, 8 interviews synthesized into six findings that set the product's direction.
• Owned Hearth, the design system, and built it as living code, so consistency is enforced by the repo, not by review.
• Built the front end prototype in Claude Code, then handed the dev team the repo they shipped to production.
• Rebuilt it on my own as Version 2 after usability testing showed people losing track of what they had approved.

Problem

Hiring a contractor is a guessing game

Trust starts social: people ask a neighbor or a friend first, because they assume online ratings can be gamed. Then the grind begins. They repeat the same problem to contractor after contractor, book visit after visit, and wait on quotes that come back impossible to line up against each other. Three frustrations surfaced in almost every interview.

Insight

The price anxiety is really scope anxiety

When contractors propose different fixes for the same issue, their quotes describe different work. One homeowner was quoted $250 to $7,000 for the same siding repair, and he had already worked out the answer: hire someone to define the scope, then send that one scope out for comparable bids. He had described Homewise before we built it. Fix the scope, and the price becomes legible.

Solution

An agent that scopes the job before it shops it

Homewise acts on the reframe: scope first, price second. A chatbot would leave people guessing at both, so I pushed for a structured, human in the loop workflow instead. The agent works on its own between four stops, the scope, sending it to contractors, the booking, and the close out, and waits for your yes at each one, even when it is sure.

Version 1

We skipped Figma and shipped version 1 in code

With the PMs and engineers, we made an unusual call: skip Figma and design in code from day one, a real React prototype on Hearth. The other two designers worked in the same repo, and I reviewed their pull requests. The dev team built the real backend straight onto it, and we tested it in three rounds. A full walkthrough of the shipped build, four minutes.

Three decisions and their tradeoffs

Version 2

The conversation carries the job, the board keeps the decisions

Version 1 gave every step its own page. Version 2, which I rebuilt on my own, puts the whole job in one window, and each finished step folds into one line, so your decisions stay in view.

Final Design

Define the scope, in plain language

You describe the problem in your own words and send two photos. The agent drafts the scope on the board: the most likely cause and an alternative, labor in hours, and a total checked against 46 local jobs. Nothing goes to a contractor until you approve it.

A match you can trust, not a list to sort

Three licensed plumbers on the same rows, from permits to Homewiser recommendations. The pick comes with its reasons, so it does not look like a paid placement, and amber marks only what needs a look. Before anything goes out, you see what they receive: photos with no faces, and your street and ZIP but not your house number.

Compare apples to apples, then you sign off

Quotes arrive in the conversation as they come in, and the agent nudges the one running late. On the board they line up against one scope and the local price range, and the headline gives the answer. Every column offers a call first, then you pick a time and book.

A booking you can still change

The visit lands on the board: when the plumber arrives, the price, and the last time you can change it for free. The conversation shows the same deadline.

The job closes when you say so

The plumber marks the visit done with three after photos, but the job closes only when you confirm the work. Recommending them is a plain yes or no that adds to their Homewisers count, and sharing your photos stays off unless you turn it on.

A working prototype, not a mockup

Design System

Hearth, calmed so a decision reads at a glance

Version 2 keeps the palette and type of Hearth, the system I owned for the team, and adds a small kit on top, because the board only works if a decision reads at a glance. Geist at 400 to 600, five sizes, one rule color, and three washes that each mean one thing.

Reflection

People hand over the work only when they can check it

Hiring a contractor is hard because nobody can tell whether the scope or the price is right, so people put repairs off. More automation does not fix that. Homewise turns a guessing game into a decision people can check, and leaves the final yes with the homeowner. An agent earns trust by showing its work, not by doing more of it.