AI Purchase Order OCR Digitization

We build a dedicated Odoo module that closes that gap. A new document intake model (extending mail.thread so every source file, confidence score, and correction is chatter-logged) accepts uploads, drag-and-drop, or an aliased inbox address. Each document runs through an OCR + AI extraction step: text-layer PDFs are parsed directly, scans and photos go through OCR, and a large-language-model extraction pass returns structured JSON — vendor identity, document number and date, currency, and a line table of description, quantity, unit price, and UoM. The extraction result is persisted with per-field confidence so nothing is silently trusted. A resolver then matches the payload to Odoo: vendor against res.partner (including product.supplierinfo vendor references and prior aliases), each line against product.product via internal reference, barcode, supplier product code, or fuzzy description, and the parsed unit against uom.uom with category validation so a Dozen-vs-Unit mismatch is flagged rather than miscosted.

Technically this is a real Odoo app, not a bolt-on script. It ships with a proper __manifest__.py (declaring purchase, stock, and mail dependencies), ORM models with @api.depends computes for confidence roll-ups and match status, and backend views built in the current framework for your version — XML list/form views on 17.0, with OWL components where interactive review helps. Security is enforced through ir.model.access.csv plus record rules so buyers only see their own company's intake queue in a multi-company setup. A review screen lets a buyer see the original document beside the extracted lines, fix any low-confidence field, and confirm — which creates the draft purchase.order through the standard ORM so all of Odoo's downstream logic (approvals, taxes, analytic accounting, stock) applies unchanged. Bulk processing is handled by a queued automated action so a batch of documents can be OCR'd without blocking the UI, and every extraction is available over the standard XML-RPC/JSON-RPC API for headless feeding from a scanner or a mailbox.

Because this is build-to-order, you are not downloading a fixed app and hoping it fits. We start with a short scoping call to confirm your document mix (native PDF vs scan vs photo), your OCR/AI provider preference (a hosted vision model, or a self-hosted OCR engine if data residency requires it), your product-matching keys, and your Odoo edition and version. We then build against your exact 17.0, 18.0, or 19.0 instance, validate on a staging copy with your real documents, and hand over installable source and a git repository. Typical delivery is 2 to 4 weeks from confirmed scope, followed by a post-go-live support window.