Solution · myfactory
Capture PDF documents in myfactory automatically
myfactory's e-invoice intake handles XRechnung and ZUGFeRD — but customer purchase orders, free-form PDF invoices, and delivery notes are still typed in by hand. The kitun Document Pipeline closes exactly this gap: template-free extraction with local AI, deterministic validation of every line item, creation in myfactory via the official SOAP-API.
- Template-free
- New senders with zero setup
- On-premise
- Open-weights models on your hardware
- Zero undetected errors
- An operating target, not a marketing claim
- No per-document pricing
- Fixed costs instead of volume tiers
- Your code
- Repository handover, no lock-in
What myfactory can already do
myfactory ships with a clean standard process for structured electronic invoices:
- E-invoice intake: Incoming XRechnung and ZUGFeRD invoices are retrieved from a monitored mailbox, checked in the preview area, and imported directly; posting templates let the system suggest the booking. For e-invoices from German suppliers, that's the right path — no AI needed.
- Interface ecosystem: Shop integrations transfer orders from e-commerce channels into sales documents automatically — proof that creating documents via interfaces is a well-established pattern in myfactory.
The gap — and what the pipeline is built for
- Customer purchase orders as PDFs: The biggest data entry burden in industrial SMEs. B2B customers order via PDFs generated by their own ERP systems — and myfactory offers no automated capture path for them. The pipeline turns them into validated sales documents.
- Free-form PDF invoices: Foreign suppliers (especially Switzerland — no e-invoicing mandate there), transition cases, and special documents arrive without embedded XML. The e-invoice intake doesn't apply here.
- Delivery notes and order confirmations: For line-by-line matching against the purchase order — the forgotten documents.
- Local processing: myfactory is a cloud ERP — document extraction doesn't have to be. The pipeline runs on-premise; only the validated, structured document goes to the cloud.
How the pipeline works with myfactory
- Intake: Email mailbox or folder; duplicate detection by file hash before anything is processed.
- Extraction — locally: Layout parsing + open-weights models on your own hardware; template-free for any sender layout.
- Validation: Per-line arithmetic, matching against item and customer master data, every value grounded in the original text. Uncertain documents → review screen instead of silent errors.
- Creation in myfactory — through the official interface: Validated documents are created via the bidirectional SOAP-API (web service access configured in administration, function permissions per API user). myfactory's business logic and permissions apply exactly as with manual entry — nothing bypasses the platform.
Hybrid by design: true e-invoices keep flowing through the native e-invoice intake — the pipeline handles only the documents that have no structured path. No duplicate process, no competition with the standard.
On-premise extraction: sovereignty despite a cloud ERP
Business documents contain terms, prices, and customer relationships. The pipeline keeps the sensitive phase — reading and validating the raw documents — entirely inside your own network: open-weights models, your own hardware, no extraction cloud, no per-document pricing. How that works technically →
Process and effort
- Pilot (2–3 weeks): ~10 real documents, gold set, end-to-end through to a draft document in myfactory (test tenant).
- Validation: ~100 documents incl. unknown senders, threshold calibration.
- Production: Review queue active, corrections continuously improve the pipeline.
Effort on your side: sample documents, web service access to the test tenant, one contact person for master data questions. Your existing myfactory setup and partner customizations remain untouched.
Frequently asked questions
Does the pipeline replace the myfactory e-invoice intake?
No — it complements it. Structured e-invoices belong in the standard process; the pipeline takes over everything that arrives unstructured: purchase orders, free-form PDFs, delivery notes, foreign documents.
Which document types can be created in myfactory?
The focus is on sales documents from customer purchase orders, plus purchasing and goods receipt documents. The exact document flow (document types, number series, workflows) is worked out in the pilot on the test tenant.
We're based in Switzerland — what about QR-bills?
The QR-bill structures payment data, not line items — it doesn’t help with capturing documents into the ERP. The pipeline reads Swiss documents including the usual formats (CHF, apostrophe thousands separators) and hands validated data over to myfactory. Background: e-invoicing and Switzerland
How reliable is the capture?
The benchmark isn’t “99% accuracy” — it’s zero undetected errors among automatically created documents: arithmetic checks, master data matching, and source grounding route every doubtful document to the review screen.
What does it cost?
One-time project cost plus fixed operating costs — no fee per document. A first business case with your own document volumes in the intro call; calculation logic openly documented.
20 minutes, one honest assessment
Bring two or three typical documents — in the intro call, we'll tell you what can be automated, what it costs, and whether it pays off. If it doesn't, we'll tell you that too.