Archives and historical records
Search large collections of scanned and handwritten records to find material for closer examination.
Overview
Cogni AI reads document layouts and contents to turn scanned, handwritten and business records into structured data. Teams can use the extracted information for search, review and operational tasks. The product entered commercial service in China in November 2025. In September 2026 we announced its use in a highway-engineering archive project in Ningxia, where it extracts and structures catalog information, compares catalog data with source documents and checks scanned page numbers for missing, duplicated or out-of-sequence pages.
Names, dates, amounts and tables can appear in different places on a page. Cogni AI is designed to identify and extract these elements without a separate fixed template for every layout. Documents can be processed in batches to organize information across larger collections.
Field-level confidence scores and consistency checks help teams identify information that needs closer examination. Extracted results remain part of a review process.
Recognize document types and layouts.
Pull out the relevant information.
Standardize fields and flag inconsistencies.
Check results against the source.
Search large collections of scanned and handwritten records to find material for closer examination.
Organize details from institutional files, company records and supporting documents. Prepare reviewed data for use in search, reporting and other business systems.
An intended use: extracting fields from statements and financial tables and flagging inconsistencies for review.
We are developing private deployment options for organizations that need document processing within their own data and access controls.
Planned options include institutional intranets, customer-owned servers and company-provided appliances.
We discuss the infrastructure, access requirements and review process with each institution when planning deployment.
Launched in China in November 2025. Read the launch announcement
Which documents take your team time to process, and what do you need from them?