Reducto

SAN FRANCISCO – October 30, 2026 -- Reducto has released Reducto Office, a free plugin that lets AI agents edit existing DOCX, PPTX and XLSX files inside ChatGPT and Codex. The plugin is available now, with usage limits.

Users install the plugin, connect a Reducto account and bring their own files into an AI conversation. Agents apply targeted changes, which appear in a live editor where users can review and refine them.

Headless document engine lets agents edit structured files directly

Office runs on a document engine that agents operate headlessly, rather than through a graphical interface built for people. Reducto says reliable editing has to preserve structure: nested tables, reviewer comments and tracked changes in contracts, and slide layouts, fonts and charts in presentations. Output stays in standard Word, PowerPoint and Excel formats that users can keep editing and share.

San Francisco – September 20, 2026 -- Reducto, an agentic document platform, has opened a New York City office to anchor its East Coast enterprise expansion, the company said.

Reducto processes billions of pages after raising more than $108 million

The company has grown to process billions of pages for large enterprises and AI teams since disclosing more than $108 million in total funding last year from investors including Andreessen Horowitz and First Round Capital. Its customer base includes Harvey, Scale AI, Vanta, Toast, and unnamed Fortune 10 enterprises.

New model r-1 targets unified document parsing

Reducto recently released r-1, described as the first in a new generation of unified parsing models, as it works to extract structured data from complex documents for production AI workflows.

San Francisco – September 07, 2026 -- Reducto has introduced r-1, a document parsing model that processes complex files for 1 cent per page while cutting parsing errors by up to 20% compared with existing tools.

r-1 replaces multi-tool pipelines with a single model priced up to 6 times lower

Organizations handling dense tables, handwriting, scans, and non-standard layouts have typically relied on combinations of tools and post-processing layers to reach usable accuracy. Reducto says r-1 consolidates layout detection, reading order, tables, formatting, grounding, and granular citations into one model, eliminating the need to route files across multiple providers.