RezPro
AI-Powered Home Service Estimates
01The Challenge
Home repair pricing is a black box. Homeowners wait days for estimates, get vague quotes, and have no way to know if the scope matches the price until a technician is already in the house.
For the business, every estimate requires a senior technician to assess the job in person — an expensive use of skilled labor that could be spent on billable work instead.
02The Solution
We built a Claude Vision pipeline that fetches uploaded photos, converts them to base64, and runs them through a structured assessment — producing a scope of work, a labor block recommendation, itemized materials with markup, a confidence score, and technician prep notes.
The admin portal includes a prompt management system with versioning and A/B traffic splitting so the team can iterate on estimation quality without touching code.
A block-based CMS handles all marketing pages with draft/publish/schedule workflows, giving the business full control over their web presence.
The Claude Vision pipeline converts uploaded photos to base64 and runs structured assessment — scope, labor block, materials, confidence, and tech notes.
From Photo to Field Dispatch
Photo Upload
Homeowner uploads photos of the repair job via the website
Client-side compression + S3
Vision Analysis
Claude Vision identifies issues, assesses severity, and scopes the work
Base64 pipeline → Claude structured output
Scope & Materials
Structured assessment: labor block, itemized materials, confidence score
Typed JSON schema validation
Pricing
Staff-editable service blocks and material markup produce the flat-rate estimate
Admin settings → pricing engine
Jobber Sync
Approved quotes sync to Jobber for field dispatch and scheduling
Jobber OAuth + REST API
Photo Upload
Homeowner uploads photos of the repair job via the website
Client-side compression + S3
Vision Analysis
Claude Vision identifies issues, assesses severity, and scopes the work
Base64 pipeline → Claude structured output
Scope & Materials
Structured assessment: labor block, itemized materials, confidence score
Typed JSON schema validation
Pricing
Staff-editable service blocks and material markup produce the flat-rate estimate
Admin settings → pricing engine
Jobber Sync
Approved quotes sync to Jobber for field dispatch and scheduling
Jobber OAuth + REST API
03Key Features
Claude Vision Estimation
Analyzes uploaded photos to identify issues, assess severity, and output a structured job assessment with scope, materials, and a confidence level
Prompt Management with A/B Testing
Admin portal for versioning AI prompts, splitting live traffic across variants, and tracking accuracy metrics per version
Block-Based CMS
Marketing pages built from typed content blocks stored in Postgres, with draft/publish/archive states and scheduled publishing
Block CMS stores typed content blocks in Postgres with draft/publish/schedule states — the team manages marketing pages without touching the codebase.
04Results
Photo-to-estimate
A/B traffic splitting
Staff-editable pricing
Jobber sync