AI Document Digitization (Universal OCR)

This is a build-to-order module, not an instant download. Typical lead time is 2–4 weeks depending on how many document types and target Odoo models are in scope. The flow is: you share representative document samples and tell us which Odoo models the data should land in; we build and configure the extraction + field-mapping, validate it against your real samples, then install it on your Odoo instance and train your team. You'll get a validation report showing accuracy before handover.

Every engagement includes a defined post-go-live support and bug-fix window with agreed response times, plus a documented path for keeping the module working across Odoo point releases. Because you receive the full commented source, your own team (or ECOSIRE on an extended support agreement) can add new document types or adjust mappings later. We can also scope ongoing maintenance and Odoo 17→18→19 upgrade support separately.

It works on both Community and Enterprise, and supports Odoo 17, 18 and 19. Unlike Odoo's built-in Invoice/Document Digitization — which is an Enterprise, IAP-metered, invoice-focused feature — this module is a self-contained addon you own. The AI extraction runs through the configuration we build, not a locked Enterprise-only credit meter, and it covers any document type, not just invoices.

No. That's the core difference from most apps.odoo.com OCR modules. The extraction is template-free: an AI model classifies the document and locates the relevant fields regardless of layout, so a new supplier's invoice or a different ID format works without you authoring a per-vendor template. You configure mappings per document type, not per layout, and low-confidence extractions go to a human-review queue before they commit.

Extracted fields are mapped onto the Odoo model and fields you choose via the mapping designer, and can trigger base-automation actions (create a purchase.order, populate a custom model, tag a partner, start an approval). Every field carries a confidence score; anything below your threshold lands in a review queue for a human to confirm or correct before the record is posted. The original file is stored as an ir.attachment on the record with a side-by-side QWeb preview, plus a full audit trail of what the AI proposed and what was changed — so nothing hits your data silently.

Yes. Intake works three ways out of the box: a watched folder (network/SFTP) swept by an ir.cron scheduled action, a monitored mailbox that digitizes emailed attachments on a poll cycle, and direct scanner output dropped into the watch folder. For deeper integration, we can expose XML-RPC / JSON-RPC endpoints so an external DMS, scanning appliance or third-party app pushes documents into the pipeline programmatically.