Capture PDF documents in SAP Business One automatically

An honest inventory before anything gets built:

• E-invoices and electronic documents: Via the Electronic Document Import, B1 processes structured formats depending on your localization — no AI required.
• Document Information Extraction (SAP BTP): SAP offers its own document recognition integration for incoming invoices. It is a cloud service on the SAP Business Technology Platform, licensed separately, focused on A/P invoices (PDF/scan → draft in B1).

If you only want to automate incoming invoices and have no issue with documents flowing into the SAP cloud, you should know about this option and evaluate it.

Three requirements the standard doesn't cover:

1. Sales orders: Customer purchase orders that arrive as PDFs and need to be created in B1 as sales orders — with line-item matching against the item master and customer item number mapping. Among mid-sized companies, this is the most common and most expensive data entry case, and it's exactly where B1 offers no standard path.
2. Strict on-premise requirements: When documents — terms, prices, customer relationships — must never leave the building, any cloud extraction is off the table. The kitun pipeline runs entirely inside your own network, on your own hardware.
3. Every document type, one system: Order confirmations, delivery notes, credit notes, and Swiss documents (no e-invoicing mandate, QR-bills without line-item data) — one pipeline, one operating model, instead of one tool per document type.

1. Intake: Documents arrive via email mailbox, folder, or scan. Duplicates are detected before any processing (file hash).
2. Extraction — locally: Layout parsing and AI extraction run on-premise with open-weights models. Template-free: new senders and changed layouts with zero setup.
3. Validation: Deterministic checking of every line item (quantity × price = total), matching against the item and business partner master in B1, every value grounded in the original text. Uncertain documents go to a review screen instead of entering the system unchecked.
4. Creation in B1 — through the official interfaces: Validated documents are created via the Service Layer (REST/OData) or the DI API — as a sales order, purchasing document, or goods receipt. All B1 business rules and validations apply, exactly as with manual entry. No direct database access — the pipeline fully respects SAP's support guidelines.

The result: documents appear in B1 as if entered by hand — just minutes after arrival, arithmetically verified and linked to the source document.

Extraction and validation run on your own hardware (a GPU workstation or compact AI appliance handles mid-sized volumes). Open-weights models instead of cloud APIs: no third-country transfers, no tangle of data processing agreements, no per-document pricing — and the model remains a swappable component with no vendor lock-in. More on the on-premise architecture →

• Pilot (2–3 weeks): ~10 real documents, gold set, end-to-end through to a draft document in B1.
• Validation: ~100 documents incl. unknown senders; metrics and threshold calibration.
• Production: Go-live with review queue; corrections continuously improve the pipeline.

Effort on your side: provide sample documents, one contact person for master data questions, Service Layer access (test system). No changes to your existing B1 customizations.