AI document understanding platform that reads layout, tables and handwriting

Document understanding is AI that reads a document's text and layout to work out what each value means, and returns its fields and tables as structured data. It is also the name of a software category, not only of UiPath's product. Docsumo is an intelligent document processing (IDP) platform, and its AI document understanding works on its own, with no RPA suite. Where OCR returns characters, document understanding returns meaning: which document each page belongs to, which value is the business name or the amount due, and which rows make one table. Docsumo splits mixed files into their documents, extracts fields and tables with a confidence score on each, reads handwriting, and sends values it isn't sure about to a person. The field also goes by intelligent document understanding, and close names are document intelligence and document AI. There's more in our guide.

In layers. Docsumo splits a mixed upload into its documents and labels each one with its type. OCR reads the printed and handwritten text, and LLM-based models structure it into named fields by what each value means, not where it sits, joining tables that run across pages. Every field and every label gets a confidence score, and values below your threshold go to a reviewer. On the Enterprise plan, the documents in a file are also checked against each other. Results go to your systems as JSON by API or webhook, or as an Excel export. Auto-classification and splitting are on the Business plan.

No. UiPath Document Understanding is UiPath's product, part of its automation platform. UiPath now sells its document AI as UiPath IXP, and Document Understanding runs as a service reached through IXP. Docsumo is a separate product from a separate company: an intelligent document processing (IDP) platform whose document understanding works on its own, with no RPA suite. UiPath robots can still use it: they send files to Docsumo's API and take the results back by webhook. See Docsumo vs UiPath.

No. Files come in by email, upload or API, and results go out as JSON by API or webhook, or as an Excel export, so there's no robot to build. Any automation tool that can call an API can use Docsumo, including UiPath and Automation Anywhere robots, a workflow tool or your own code.

OCR turns a page image into text. Document understanding reads that text with the page's layout to work out what each value means and which document it belongs to. Intelligent document processing (IDP) is the workflow around that reading: intake, checks, review and delivery to your systems. Docsumo is an intelligent document processing (IDP) platform, and document understanding is the reading inside it, so one product covers both. See intelligent document processing and OCR software.

No. Docsumo's LLM-based models read new document types without templates or training data, and find each field by what it means rather than where it sits, so a statement from a new bank or an invoice from a new supplier doesn't need a template. You can start on your own documents on day one. You name the fields you want, and accuracy improves as reviewers correct values.

Yes. When a table runs across pages, such as the transactions on a five-page bank statement, Docsumo joins it into one table and maps the headers onto every page's rows. On bank statements it also checks each transaction against the running balance. Each table comes back as rows by API, or in an Excel export.

Yes. Docsumo reads handwritten text as well as print, such as a filled-in form, a signed date or a handwritten amount, from scans and phone photos. It doesn't quote one accuracy figure for handwriting, since that depends on the writing, so test samples of your own. Values it isn't sure about go to a reviewer with the source line highlighted.

Docsumo reads most printed languages, including Latin, Cyrillic, Greek, Arabic, Hebrew, Chinese, Japanese, Korean and Indic scripts, and it classifies documents in them too. Handwriting support varies by language, so test samples in yours.

Docsumo reaches 99% field-level accuracy across 250+ document types, and 95%+ of documents are processed straight through, with no manual review. Field-level means each extracted field counts as right or wrong, which is stricter than counting characters. Every field and every document label carries a confidence score, and values below the threshold you set go to a reviewer before they reach your systems.

Start a free trial on your own documents: 14 days, up to 1,000 pages and 10 users, with the API and webhooks. Or book a demo, bring a real file, and we'll run it live. Paid plans are quoted, with up to 10% off annual plans; see pricing.