Receive. An invoice enters the project from an attachment or Drive.
Step 1 of 3 · ReceiveAn extraction project (1 of 3)
Preview taken in a demonstration workspace. All data shown is fictitious.
Documents arrive. Data follows.
Define the fields to extract, receive documents, and review results beside the source. IDP relies on ODIN, Devana’s document extraction engine.
- Supplier invoices
- HR documents
- Technical documents
Features
From the document received to validated data, step by step.
1The project
One project per type of document.
A project holds the fields to extract, the extraction mode and what follows extraction. You set it up once, for every document of the same type.
Four templates to start from
Expense reports, supplier invoices, HR documents, bank statements: each template brings its fields, its instructions and its export. A custom project starts from a blank page.
Twelve field types
Text, amount, date, address, choice list, table of lines: every field has its type, and its description guides the AI to the right value.
Three extraction modes
Full AI for varied documents, zones for a fixed layout, technical drawing for your industrial plans.
Zones drawn on the page
On a reference document, draw one rectangle per field, page by page: the model reads these zones on every new document.
A preset for drawings
Technical drawing mode creates the fields of the trade from the start: reference, material, dimensions and tolerances, threads, surface treatments, standards.
Shared with your team
Open the project to people or to teams, from viewer to administrator.
2The sources
Four ways in. One project.
A document joins the project from your computer, from Drive, from a mailbox or from a deposit page, then follows the same path.
From your computer
Drop your PDFs, Word documents, Excel workbooks and images, several at a time: an identical file already in the project is set aside.
From Drive
Select files already stored in Drive: they join the project and stay where they are.
From a mailbox
A mail rule or a workflow sends the attachments of a mailbox to the project, filtered by type, size or file name.
From a deposit page
A link, password-protected if you wish, opens a page where your colleagues and suppliers drop their documents, with their name and your own fields.
3The reading
ODIN reads. The model extracts.
ODIN, Devana’s document extraction engine, reads the document. The project’s model then finds the value of each field, with its source.
ODIN reads the document
Native or scanned PDFs, Office documents, images: ODIN extracts the text page by page, and the document’s status follows each step live.
The model you choose
Each project picks its model among those of Suite 366 and those of your organisation, and takes your extraction instructions.
Every value, with its source
For each field, extraction returns the value, its confidence score, the page and the passage where it was found.
Automatic extraction
Extraction starts as soon as reading is done. You run it again on one document, a selection or the whole project after changing the fields.
Project tracking
Completion rate, average confidence, spread of scores and documents at each step: the project dashboard follows the progress.
Your documents, on your premises
IDP is deployed with Suite 366, on your servers or on a DIWY box. Source documents and extracted data are stored in your infrastructure.
4The review
AI proposes. You validate.
Every field carries a confidence score. Below the threshold you set, the document waits for a person to review it, beside its source.
A confidence threshold
You set the project’s threshold: above it, the document is done; below it, it moves to “To review”.
The document beside the fields
The review screen shows the source on the left and the fields on the right. A click on a field shows where its value comes from in the document.
Correct, then validate
Edit a value, add a row to a table, validate: the document is marked “Validated”, and Ctrl + → opens the next one.
Duplicates flagged
After extraction, a document whose values match those of another is marked “Possible duplicate”, with the name of the similar document.
5The outputs
Data goes where it is needed.
Table, export files, CRM item, workflow, accounting: each project chooses its outputs.
A table of every document
The Table tab sets one row per document and one column per field. A double-click corrects a value in place.
JSON, CSV, Excel
Extracted data downloads as JSON, CSV or an Excel workbook, ready for your spreadsheet or your business tool.
A ZIP sorted for you
An export profile files documents into folders and renames them from the extracted fields: year, supplier, amount. A CSV summary comes with the archive.
An item in the CRM
Each extracted document creates an item in the board you choose: a name built from the fields, columns filled in, source file attached.
Your workflows take over
The created item triggers the board’s workflows: notify a team, hand an analysis to an agent or a PersonIA, ask for a validation, call an external tool.
Accounting reconciliation
Connected to your accounting software, the validated invoice is sent to it, and AI suggests the matching bank transaction.
6Your uses
Three ways to use it.
Supplier invoices
Extract suppliers, amounts and references.
HR documents
Structure information from submitted files.
Technical documents
Configure extraction for your drawings.
Runs on your premises.
Where this application runs, and the technologies that come with it.
- On-premises
- On your servers, with a configuration sized for you.
- DIWY
- Available on DIWY, depending on your configuration. Discover DIWY
- Requirements
- Suite 366 module with ODIN and configured text or vision models. Fields, extraction mode and outputs are set per project.
- Works with
In Suite 366.
This module is part of Suite 366. It works with the other modules, in the same workspace.