Case Study: HomeLight achieves 99% document processing accuracy with Extend

HomeLight Hits 99% Accuracy and Eliminates Manual Review with Extend

Homelight, a real estate platform, faced a significant operational bottleneck as its team manually reviewed lengthy contracts, a process taking over 20 minutes per document. This manual review threatened to hinder the growth of their product line by requiring a linear increase in headcount. They initially built a prototype with OpenAI's APIs but only reached 85% accuracy, which was insufficient to remove humans from the process.

To solve this, Homelight implemented Extend's AI document processing platform. Extend provided a suite of tools for extraction and a built-in human-in-the-loop review system, allowing Homelight to go live in just two weeks. The results were immediate: accuracy jumped to over 98%, and manual review was eliminated entirely after a month of zero corrections. This allowed Homelight to scale its operations without adding staff and freed its engineering team to focus on core product development.

HomeLight

Mike Abner

Chief Technology Officer