Getting Turkish e-invoices into your ERP automatically

What is the difference between e-Fatura and e-Arşiv?

e-Fatura is an electronic invoice whose data format and standard are set by Türkiye’s Revenue Administration (GİB), transmitted between seller and buyer through GİB’s central platform. It is issued only between taxpayers registered for e-Fatura. GİB adopted UBL-TR, an XML-based format built on the open UBL standard.

e-Arşiv is the electronic invoice issued to buyers who are not registered for e-Fatura, including people who are not taxpayers. According to GİB, it is delivered to the buyer on paper or electronically, at the buyer’s request.

For automation, the difference is simple. An e-Fatura reaches you as structured XML. An e-Arşiv invoice often arrives as a PDF attached to an email. In the first case the data is ready; in the second you must extract it from the document first. If you still receive paper and scanned invoices, one flow ends up handling four kinds of source.

How do you collect incoming invoice data?

There are three ways to receive e-Fatura. If you use the GİB Portal, invoices appear there and must be downloaded. If you use a private integrator (özel entegratör), most offer an API or folder export you can pull incoming invoices from. Direct integration with GİB makes sense only for large companies that have built the technical infrastructure for it.

The cleanest automation source is UBL-TR XML pulled from the integrator’s API. The XML carries the supplier’s tax number, invoice lines, unit price, VAT rate and totals in separate fields, so no AI is needed to read it. e-Arşiv and PDF invoices usually land in an inbox, so the flow starts by watching that inbox and picking up attachments.

The table summarises the method by source type.

Source type	How it is collected	How it is parsed	How it is written to the ERP	Human approval
e-Fatura (UBL-TR XML)	Integrator API or GİB Portal download	XML fields read directly	API, import file or staging table	Only for unmatched lines
e-Arşiv (PDF, sometimes with XML)	Inbox monitoring	Read XML if present, otherwise document analysis	API, import file or staging table	For low-confidence fields
PDF invoice (including foreign suppliers)	Email or supplier portal	Document analysis and field extraction with a language model	API, import file or RPA	Yes, at least at first
Scanned paper invoice	Scanner folder	OCR, then document analysis	API, import file or RPA	Yes

Should you write to Logo, Mikro or Netsis through an API or RPA?

An API or official import route always comes first. Logo, Mikro and Netsis, the ERPs common in Türkiye, may offer web services, import files or a staging table in the database, depending on version and licence. Ask your ERP partner or vendor which route is open to you; we do not describe product-specific APIs here.

RPA means a software robot using the ERP screen the way a person would. It works when there is no programmatic way in. But when a screen changes the robot breaks, debugging is hard, and every version upgrade needs testing. Treat RPA as a bridge, not a permanent fix. We explain how we make that call on our RPA consulting page.

The layer that runs the flow in between is usually a workflow automation tool such as n8n or Make. It picks up the invoice, checks it, matches it, writes it to the ERP and alerts the right person when something fails.

Where does AI help with PDF and paper invoices?

Without structured XML, data has to be extracted from the document. Classic OCR reads text but does not know which number is the VAT total. Document analysis tools and language models are better at finding fields across suppliers’ different layouts. Platforms such as UiPath Document Understanding organise this as a flow of classification, field extraction and human validation.

Keep the language model’s role narrow: read the document, extract fields into a fixed schema and flag fields it is unsure of. Decisions such as choosing the accounting code or correcting an amount are made by rules and human approval. More on this in our document analysis and LLM integration pages.

One caution: invoices can contain personal data. Sending them to an AI service abroad can raise transfer questions under GDPR and KVKK, Türkiye’s data protection law. Choose the model and hosting together with your legal counsel.

How do you set up matching, approval and error checks?

Reading the invoice is half the job. The other half is linking it to the right order, delivery note and supplier account. A typical flow looks like this:

1. Supplier match. The seller’s tax number is matched to the supplier record in the ERP. If none exists, the flow stops and raises a request to create one.
2. Order and delivery note match. Invoice lines are compared with open purchase orders and e-İrsaliye (electronic delivery note) records on quantity and unit price.
3. Tolerance rule. If the difference is below your threshold, the record moves on automatically; above it, it goes to an approval queue.
4. Approval. The owner sees the invoice and the difference on one screen and approves or rejects it.

Checks that prevent errors and duplicate postings are part of the flow too:

• Uniqueness check. Has this invoice number and ETTN (the universal unique identifier in UBL-TR XML) already been processed?
• Totals check. Do line totals, VAT and the grand total agree?
• Date and period check. Does the invoice fall into a closed accounting period?
• Currency check. For invoices in foreign currency, are the rate and date applied by the same rule as the ERP’s rate table?
• Alerts. When the flow stops on an invoice, the owner gets a short note on what happened and what to do; errors never pile up silently.
• Audit trail. For every invoice, keep its source, who approved it and which ERP record it became.

Where should you start?

1. Count the last three months of incoming invoices by source type: e-Fatura XML, e-Arşiv, PDF, paper.
2. Find out which routes your ERP offers for posting records from outside.
3. Start with the largest group; for most companies that is e-Fatura XML, which needs no AI.
4. Write matching and tolerance rules down together with the finance team.
5. Approve every record at first, then widen the share that flows through automatically as rules settle.

The load is heaviest in manufacturing and construction and building materials companies with many suppliers, and in SMEs with small finance teams.

At Corrplus we begin with discovery: mapping invoice sources, the ERP’s entry routes and approval rules. In design we set down the method (API, import or RPA) and where AI is needed, and we build in short cycles. More on our invoice processing, workflow automation and n8n and Make automation pages.