September 2026 Summaries

Finch Legal, founded in 2024, supports personal injury firms with an AI-enabled, services-first model in which in-house paralegals use technology to manage substantially larger caseloads while engineers rapidly refine workflows based on direct feedback. Its main challenge is reducing the time paralegals spend reconstructing a case’s status from roughly 250 documents per file, including medical records, insurance information, correspondence, and reports. Reducto processes every incoming document into structured, machine-readable text with page-level metadata and citation locations, allowing Finch Legal to classify documents, update case information, detect required actions, and flag risks as new files, emails, or messages arrive. Citations are particularly valuable because they let paralegals and partner firms trace extracted information back to the original evidence, combining automation with human review. After initially operating within established firms’ processes, Finch Legal is expanding toward handling cases end to end, with document ingestion and traceable automation remaining central to its approach.

The company has opened a New York City office to strengthen its East Coast presence and support growing enterprise demand for its document intelligence platform, which extracts structured data from complex documents for production AI applications. Following more than $108 million in funding announced last year, it reports processing billions of pages for major enterprises and AI teams and recently introduced r-1, a unified parsing model. The office is intended to enable closer in-person collaboration with customers throughout evaluation, system architecture, and deployment, particularly in industries where accuracy and integration into existing workflows are essential. The company also plans to hire New York-based Enterprise and Growth Account Executives, Business Development Representatives, and Solutions Engineers.

Reducto has introduced r-1, a preview document-parsing model designed to improve accuracy, speed, and cost efficiency for complex files such as financial statements, contracts, dense tables, and irregular layouts. Built on a new architecture that processes layouts, formatting, tables, citations, reading order, and difficult visual features such as strikethroughs and watermarks in a single pass, r-1 reportedly reduces errors by 20% compared with the company’s legacy agentic pipelines. The model costs one cent per page, compared with the prior three to six cents per page, and is part of Reducto’s effort to provide simpler, transparent pricing without hidden multipliers. Reducto plans to release a lower-cost r-1 mini model and automatic page-level routing later in the year, while offering up to $5,000 in credits and migration assistance for users evaluating it against other document parsers.