Harvey, the legal AI platform valued at $11 billion earlier this year, achieved a 26x increase in document processing volume over the past 12 months, according to a company blog post published on July 27, 2026. The platform now handles tens of millions of documents weekly, processing 24.8 million documents weekly—equivalent to 56 terabytes of data—compared to just under one million documents in a busy week a year ago, which consumed 1.44 terabytes of data. This translates to roughly 3.5 million documents daily, driven by expanded enterprise adoption and increased integration with document management systems such as iManage, SharePoint, Google Drive, and Box. Harvey rebuilt its document processing pipeline, splitting it into distinct stages for extraction, embedding, and indexing, and introduced a Unified Document Format (UDF) that cut extraction latency by up to 19%. The company also transitioned to a new vector-store system via a live migration involving dual-writing and shadow-read validation, and replaced JSON serialization with Arrow IPC to reduce latency. Harvey mitigates risks from OCR provider slowdowns, blob read timeouts, and rate limits through stage-specific capacity controls, dynamic task routing, and advanced observability tools.