Fyxr Launches an AI-Enhanced Contractor Platform

Fyxr today rolled out a platform-wide update that brings targeted AI assistance into three places where homeowners and contractors have historically lost the most time: writing a clear job description, triaging incoming work, and reviewing quotes side by side.

The release does not replace any part of the marketplace. Contractors still set their own prices, homeowners still pick who they work with, and licensing checks remain a human step. The goal is narrower: remove the friction that used to live between a homeowner thinking "I need a plumber" and getting a contractor on site.

What changed

Three flows received AI-assisted layers in this release:

Job intake

Homeowners describing a job in their own words now get follow-up prompts that ask for the small details contractors actually need — wall material, access to the panel, age of the equipment. The structured form behind the conversation is unchanged, so the contractor still sees a normalized brief rather than a transcript.

Contractor triage

When a job lands in a contractor's inbox, a one-line summary now appears alongside the full brief: scope, urgency, and whether the address falls inside the contractor's primary service area. Contractors who reviewed the change in beta said the summary was most useful at the top of their day, when they're deciding which jobs to quote first.

Quote review

Homeowners comparing multiple bids see a structured comparison of what each contractor included — materials, labor, timeline, and warranty terms. The platform doesn't recommend a winner. It just makes the bids legible.

What we are not doing

A few things were deliberately left out of this release:

• No price suggestions. Bids are between the contractor and the homeowner. The platform does not generate a "fair price" estimate.
• No automated quoting. Every quote on Fyxr is written by a contractor.
• No personal data goes to the model. Job descriptions are processed; names, addresses, payment details, and message threads are not sent to the LLM provider.

How it was built

The platform uses a small, well-defined set of prompts behind feature flags, each with a human-reviewed evaluation suite. We can roll back any individual surface without affecting the others, and the AI layer can be turned off at the account level for homeowners and contractors who prefer the unassisted flow.

The bar was simple: if a feature didn't save real time for a real user we tested with, it didn't ship.

We expect the system to keep improving as the dataset of cleaned-up job descriptions grows. For now, the win is concrete — homeowners write less, contractors read less, and the quotes that come back are easier to compare.

What's next

The next release will focus on scheduling. Today, locking in a job date still requires a few back-and-forth messages. We're testing a calendar-aware intake step that lets homeowners propose two or three windows up front and lets contractors confirm one with a single tap.