Industry: Real Estate
Use Case: Document Processing
Location: San Francisco, CA
States Title is a San Francisco-based company that specializes in commercial and residential real estate settlement services. States Title’s mission is “to apply machine intelligence to the title and escrow process, enabling faster and safer real estate transactions.” To that end, they are looking at every aspect of the process, including underwriting, document processing, and communication.
Anyone who has gone through a real estate transaction can attest to how document-heavy the title and escrow process can be. These documents are necessary to establish insurance coverage, specify fees, and officially convey ownership of property. However, identifying and parsing these documents manually is incredibly time-intensive and error-prone. Automating this process will lead to major leaps in efficiency and response time, and reduce the frequency of human error when performing data entry or quality control checks. This will also lead to reduced costs and a better experience for all parties including lenders, realtors, and consumers. States Title’s patented machine learning technology, combined with specialized closing expertise will make closing a mortgage simpler and more efficient for buyers.
Title and escrow documents are a difficult subject matter, often filled with arcane legalese and a substantial amount of nuance. “We needed a partner that had the capability to train their annotation team and guarantee performance metrics,” said Brian Holligan, Senior Manager of Data Science. The States Title team initially worked with offshore annotation companies in their early data science modeling efforts. A persistent theme that emerged, however, was the difficulty in ramping up annotator performance. “We found ourselves constantly auditing and resubmitting documents for annotation, and reached a point where our internal algorithms to annotate documents were outperforming the work provided by our vendor,” added Holligan. That’s when States Title decided to go with Scale – a technology-first company building solutions to bring automation to industries like these with a world-class ML team and ML-augmented operations.
“Scale has provided the fuel to put our machine learning systems on overdrive. They make sure the highest quality training data is there in time to meet our aggressive roadmap. Lenders and borrowers will experience faster and more efficient closings sooner as a result.”
Chief Data Science Officer, States Title
The States Title team chose Scale AI and the Scale Document product because “Scale checked all of the boxes we were looking for when it came to data annotation,” said Holligan. By taking a hybrid approach of combining the best of OCR technology with human insight, Scale Document, “Guaranteed robust performance, could manage a variety of annotation tasks, were capable of spinning up both small and large projects to accommodate our needs and had dedicated team members available to troubleshoot any issues that emerged,” Holligan concluded.
By incorporating feedback early in each project and regularly updating the annotator training curriculum, Scale has maintained 95%+ accuracy on all of States Title’s projects. Working with Scale has also, “Freed up our data science team to spend more time on actual data science work, instead of getting bogged down in arduous auditing cycles. The highlight of working with Scale has been the technical support we’ve received. They’ve helped us debug code, build custom audit feedback, and adapt the design of the product based on our project requirements,” Holligan concluded. States Title customers can experience the future of real estate settlement services with effective natural language processing models such as Instant Underwriting and Instant Closing Disclosure.
“The highlight of working with Scale has been the technical support we’ve received. They’ve helped us debug code, build custom audit feedback, and adapt the design of the product based on our project requirements.”
Senior Manager, Data Science, States Title