Live · fixli.aiFounding workAI service

A repair request both sides can trust before anyone commits.

Fixli is a home repair marketplace. A homeowner snaps a photo and describes the problem in plain words, and AI turns it into a clear brief and a cost estimate that a local pro can judge quickly.

Years
Jun 2025 – now
My role
Co-founder · Design & engineering
Team
2 co-founders
Status
Live · fixli.ai
Fixli’s landing page: ‘Snap. Send. Fix.’ with a photo upload card asking ‘What’s broken?’ and repair categories below.

What I covered

Overall role Co-founder Steps 1–4

  1. 01FindTrust-gap problem
  2. 02ResearchHomeowner interviews, Trust UX
  3. 03DirectionDesign lead, AI behavior design
  4. 04BuildHomeowner screens, Portfolio builder, Design system
  5. 05OperateJust launched
1Find the problemTrust-gap problem

Finding a pro got easy. Committing to one still takes a week of calls.

Platforms made discovery easy, yet homeowners still call three contractors and wait a week for quotes. A homeowner can’t tell whether a quote is fair or the pro is right for the job, and a pro can’t tell whether the job is real or worth the trip.

Diagram: discovery is solved, but homeowners need quote, fit and timing confidence and pros need job, scope and trip confidence. Fixli adds a shared, AI-assisted context layer between them.
The trust gap: both sides lack the context they need to commit, at different moments.
Why now

The home services market is large and fragmented. When a homeowner finds a pro, they still have to explain the problem, wait for a follow-up and judge whether the answer is fair. When a pro gets a request, they still have to decide whether it’s real, clear and workable before spending time on it. AI tools can now take some of that ambiguity out: a homeowner describes the need in their own words, and the product turns it into context both sides can use.

2Research itHomeowner interviewsTrust UX

Homeowners and pros decide to trust at different moments.

A homeowner decides before spending money: when choosing which pro to contact or which answer to accept. A pro decides before spending time: when a request arrives and they choose whether to respond. The signals they need are different, so serving one side doesn’t automatically serve the other.

Homeowner interviews2rounds

with 3–4 homeowners

Guerrilla interviews2more

quick sessions with 1–2 people each

Pro advisor1pro

meets the team often to give feedback on the pro side

Homeowner, before spending money: “Can I trust this option?”

  • The problem feels understood
  • The pro looks credible
  • The scope feels right
  • Price and timing feel fair

Pro, before spending time: “Is this request worth answering?”

  • The inquiry looks real
  • The need is specific enough
  • Fit is easy to judge
  • The next step is quick
3Set directionDesign leadAI behavior design

A homeowner writes in plain words. The pro gets a brief they can act on.

Homeowners shouldn’t need trade vocabulary to ask for help, and pros shouldn’t have to guess what they’re being asked to quote. The hard question was how much the product should help before either side feels locked in.

From a vague need to a clear next step
  1. Describe the need
  2. Clarify the context
  3. Compare options
  4. Decide the next step
  1. 01Reduce the translation burdenA homeowner should be able to start in plain language. The product turns that into useful context without making them an expert.
  2. 02Make uncertainty visibleA repair request often has unknowns. The interface surfaces them at the moment they’re easy to resolve.
  3. 03Design for both sides at onceThe same detail can build confidence on one side and hesitation on the other, so neither side is optimized alone.
  4. 04Keep the human decision in viewAI can make things clearer, but the homeowner still makes the call, and the product has to feel accountable for what it suggests.
Protected addendumDesigning the AI’s behavior

Four decisions that shaped how the AI works: what it asks versus infers, how it shows what it understood, where confirmation sits, and what the pro receives. Password available on request.

Open the addendum →
4BuildHomeowner screensPortfolio builderDesign system

Snap, send, fix: five steps from a photo to a pro.

I designed the two-sided workflows and built the homeowner side: the flow from the main screen to the result page, the pros’ portfolio builder, and the design system they share. Fixli is live at fixli.ai.

Building your Fixli quote: a four-step progress card (reviewing details, calculating labor and materials, validating scope, finalizing) with the line: we’d rather take 30 seconds than show you a wrong number.
While the quote builds, each step says what it’s checking, and the copy explains the wait: “we’d rather take 30 seconds than show you a wrong number.”
Your Instant Quote: an estimate of $202–$289 at 90% confidence, broken into labor hours and itemized materials, with a Request Pro Quotes button.
The result: a cost range with its confidence, labor and materials itemized, and room to add details for the pro before requesting quotes.
The homeowner flow
  1. Take a picture
  2. Pick the AI’s problem statement
  3. Answer follow-up questions
  4. Get the quote
  5. Find pros
Saved quote page: the job, its cost estimate marked as an AI estimate and not a contractor’s quote, the homeowner’s photo, job and schedule details, a share link to find a pro, and a status timeline.
The saved quote. The estimate is labeled “AI estimate, not a contractor’s quote,” and a share link lets the homeowner bring in pros from their own networks.
Pro workspace: the Portfolio page with an Add Portfolio button and an example entry for a full kitchen renovation tagged carpentry, electrical and plumbing.
The pro side: the portfolio builder, where pros show finished work so homeowners can judge fit.
What I’m learning

AI features and trust are different things. A flow can get faster and feel less human at the same time. The harder design work is deciding when the product should help, when it should ask, and when it should hand control back to the person deciding.

Every trust mechanism for the homeowner changes something for the pro, and the reverse. Yerbba taught me to design the service before the screens. Fixli takes that into a consumer setting where the product has to make an offline, high-friction service feel understandable before anything has happened.

5Operate

Just launched Fixli went live at fixli.ai on November 1, 2026.

Next projectResearchDataGovOne access path to restricted research data from 16 federal agencies.