AI document reading
Docs2AI reads your archive
Scanned documents and books become verified, searchable data in your own database, entirely on your premises.
See the accuracy per field before you decide.
Overview
Docs2AI reads your archive. Letters, forms and contracts become records with their fields, books become searchable full text and registers become rows, all saved to your own SQL Server database. It runs on one AI workstation at your site, with no cloud, so no data leaves the building.
Nothing is guessed in silence. A value is confirmed only when two independent AI readers agree on it; anything else is flagged, and your staff check it beside the page image and accept the right reading with one click. You define your own document types and fields, and an accuracy report per field, measured on your own pages, shows where you stand before go-live and after every change.
What it does
-
Reads any material
Letters, forms and contracts become records with fields; books become searchable full text; registers become rows; photos are described.
-
Arabic and English
Typed and handwritten, Gregorian and Hijri dates, with a full Arabic interface.
-
Never a silent guess
Two independent AI readers must agree before a value is confirmed; anything else is flagged for a person.
-
Review beside the page
Your staff check flagged fields next to the page image and accept the right reading with one click.
-
Your document types, your fields
Define types and fields in the interface, with no programming; every change is tested before it goes live.
-
On your premises
Runs on one AI workstation at your site and saves to your SQL Server: no cloud, and no data leaves the building.
-
Measured accuracy
An accuracy report per field on your own pages, before go-live and after every change.
How it works
-
Select a folder on the server
The whole folder tree is scanned first, so you know what is coming.
-
Each document gets its type
Letter, form, contract, book or register: the type decides which fields are read.
-
AI readers read every page
Independent AI readers read each page on the workstation at your site.
-
Agreed values confirmed, the rest flagged
A value is confirmed only when the readers agree; any other value is flagged for review.
-
Staff review; results saved to your database
Your staff check the flagged fields beside the page, and the results are saved to your SQL Server.
Measured results
Accuracy measured on completed archive projects of ours, against independently checked answer keys.
Product screens
From the design prototype, with sample data. Select a screen to enlarge it.
-
DashboardPages read, speed, queue and review backlog at a glance. -
New importSelect a folder on the server; the whole tree is scanned first. -
ReviewFlagged fields beside the page image, with every reader's value. -
DocumentsArabic and English full-text search, with filters and exports. -
BooksBooks in full text: contents, page and recognised text side by side. -
Document typesDefine your own document types and fields, with no programming. -
AccuracyAccuracy per field, measured on your own pages.
Ask your archive
A chat in Arabic and English that answers questions from your documents, with every answer linked to the page it came from.
Request a pilot on your own documents
See the accuracy per field before you decide.