Role

Strategy & Design

Product

Houzz Takeoffs Software

Houzz Pro Takeoffs: AI Symbol Recognition

Led design and strategy for an AI symbol recognition experiment in the Takeoff Count tool, testing a next-generation AI model that drove immediate engagement and helped shape Houzz Pro’s shift from targeted AI experiments to a platform-wide AI strategy.

Main Project Image

Opportunity

Symbol counting in Takeoffs was largely manual, and AutoMate, the existing AI measurement mode, was limited to floor plans. As AI-powered recognition became an expected capability among our target audience, we launched an experiment to test a faster model that could detect symbols across all plan types, while working within real technical constraints and ensuring results professionals could trust.

Large Project Gallery Image #1

Partners

Product manager, AI scientist, and engineer.

Design Strategy

To quickly validate symbol recognition without heavy engineering lift, I introduced a universal “Search” workflow that works across all plan types and operates independently from AutoMate at Houzz. This allowed AI-generated counts on both floor and electrical plans, removed technical constraints tied to floor plan detection, and moved Houzz Pro Takeoffs closer to a reliable, AI-first experience for larger firms and enterprise clients.

Impact

The results showed strong adoption and confidence in the AI-assisted workflow, with 22.7% of users returning on another day, 56% running multiple detection sessions, and an average of 12 symbols counted per session.