Use case

Automated invoice processing
for Hungarian companies

In most companies an incoming invoice is still a PDF that someone opens, reads and retypes into another system. DocAI removes that step: it reads the invoice data, checks the partner against four authoritative registries, reconciles it with NAV data — and hands the approver a pre-qualified item instead of a raw file. Processing runs on the company's own server; the invoices never leave the organisation.

Request a demo How processing works

Starting point

Storage was never the bottleneck

At most companies the incoming invoice is already digital: it arrives by email, lands in a folder, maybe in a records management system. So the document exists — it just knows nothing about itself. For it to reach accounting, someone has to read it, collect eight to ten values from it, and then verify that the partner is a real, operating company at all.

This work is monotonous, which is exactly why people get it wrong: a mistyped bank account number, a missed payment deadline, an invoice recorded twice. DocAI automates that stretch — the reading and the checking — while approval stays where the responsibility is: with the user.

The flow

What happens to an invoice after it arrives?

Five steps, four of them automatic. The fifth deliberately is not.

Arrival — the invoice comes in from email, an upload or an invoicing-system link, downloaded automatically.
Type detection — the system recognises it as an invoice and routes it to the matching processing path.
Reading (OCR + KIE) — the text is recognised, then structured extraction fills the fields: header data and line items.
Verification — partner validation from four authoritative sources, NAV matching, duplicate filtering, own company data cross-check.
Human approval — the approver sees a pre-qualified item, not a raw PDF. The decision is theirs.

DocAI cost invoice register: payment method, due date, net, VAT and gross amounts, payment and bookkeeping status in one view — Cost invoice register. Incoming invoices line up as pre-qualified items — payment method, due date, net/VAT/gross and status in one view, with the expected VAT liability totalled on top. Partner names are masked in the screenshot.

Extracted data

What does the system read from an invoice?

Not a text answer, but field-level data — the kind that can go straight into accounting.

Identification: Invoice number, issue, fulfilment and payment due dates, payment method, currency
Amounts: Net, VAT and gross values — to the cent, without rounding drift
Parties: Own and partner company data, bank account number, contract reference
Line items: Description, quantity, unit price, VAT rate, line total
On international invoices: Beneficiary postal address (country / city / street), bank name, IBAN, SWIFT/BIC
Source of each value: Marked per field: NAV, AI, contract, bank reconciliation or manual entry — NAV data is locked

Verification

Four authoritative sources for every partner

Approval has an invisible first filter that nobody enjoys doing by hand. This is the part that runs automatically, on every incoming invoice.

The four sources of DocAI partner validation
Source	What it checks
NAV	Whether the tax number is valid and the VAT status active; company name, address and group taxation membership
Ministry of Justice (IM)	Validity, address, legal form; whether the bank account on the invoice matches the registered official one
VIES	Validity of EU partners' community VAT numbers
KOMA + public filings	Freedom from public debt and published financial data

The system also cross-checks your own company data against the document — resolving Hungarian address abbreviations, comparing order-independently. Results appear per source with colour coding and an aggregate reliability signal.

DocAI partner validation: NAV tax number check, VIES EU VAT, KOMA public debt and Ministry of Justice company data status on one screen, alongside register and financial data — Authoritative validation. Tax number, company register (IM), community VAT (VIES) and public debt (KOMA) — with a status per source and the invoice's bank account reconciled. The company name, identifiers, bank account numbers and addresses are masked in the screenshot.

Integrations

NAV, banking, currencies, ERP

NAV Online Invoice. Outgoing invoices can be submitted automatically or manually, with submission status and errors reported back. Incoming and outgoing data can also be queried from NAV: the system reconciles it with the OCR-extracted invoices, filters duplicates, and marks which values were synchronised from NAV.

Bank reconciliation. Bank statements are read for opening and closing balance, period and transactions, then bank items are matched automatically to open invoices — on an amount and date window, handling partially paid invoices too. Payment deadlines feed a daily cash-flow forecast.

Currencies and transfers. Multiple currencies, IBAN/SWIFT recognition, and transfer files in EDIFACT (HUF and EUR) and ISO 20022 (MBH XML) formats with structured beneficiary addresses. Recurring invoicing can be scheduled (monthly, quarterly, annually), and invoice data with attached PDFs can be forwarded in the background to an external accounting or ERP system (for example Oriana), with validation and a history log.

Evidence

What does this deliver in practice?

Gree Hungary

Five departments, 30–70 incoming documents a day. DocAI produces the payment package from the invoices through key information extraction, and the bank application imports it as a file — work that 1–2 people used to record by hand. Read the case study →

Measured accuracy

On the Hungarian invoice corpus the production model reaches 0.975 F1 at field level. The methodology and raw results are public: docai-evals. Why we measure this way →

The data stays in the house

Invoices are processed on the company's own infrastructure rather than at an external AI provider — not a minor detail for GDPR compliance. GDPR and the EU AI Act in practice →

FAQ

Invoice processing — frequently asked questions

How does automated invoice processing work in DocAI?

The incoming invoice arrives by email, upload or an invoicing-system link. DocAI recognises that it is an invoice, reads its text with OCR, then fills the fields through structured extraction (KIE): invoice number, dates, amounts, partner data, bank account number and the line items. In parallel, the partner checks and the NAV matching run. All that is left for the user is to review and approve — the decision stays in human hands throughout.

What data does the system extract from an invoice?

From the header: invoice number, issue, fulfilment and payment due dates, payment method, currency, net, VAT and gross amounts, own and partner company data, bank account number and contract reference. From the line items: description, quantity, unit price, VAT rate and line total. On international invoices the beneficiary's postal address, bank name, IBAN and SWIFT/BIC are recognised as well.

Does DocAI check the partners too?

Yes, on every incoming invoice, against four authoritative sources: the NAV taxpayer registry (valid tax number, active VAT status), Ministry of Justice company records (validity, address, legal form, and whether the bank account on the invoice matches the registered one), VIES (EU VAT number) and KOMA (freedom from public debt). Results are shown per source with colour coding and an aggregate reliability signal.

Does DocAI work with the NAV Online Invoice system?

Yes. Outgoing invoices can be submitted to the NAV Online Invoice system automatically or manually, with submission status and errors reported back. Incoming and outgoing invoice data can also be queried from NAV: the system reconciles it automatically with the OCR-extracted data, filters duplicates, and marks whether a value came from NAV or was entered manually. Data coming from NAV is locked.

Does it handle foreign-currency and international invoices?

Yes. The system handles multiple currencies (HUF, EUR, USD and others) to the cent, with a separate own currency for fulfilment when it differs from the one on the invoice. Bank transfer files are produced in EDIFACT (HUF and EUR) and ISO 20022 (MBH XML) formats, with structured beneficiary addresses.

See it on your own documents

In a 30-minute demo we walk through your own workflow — with your documents, on your terms. No obligation.

Request a demo

More use cases

One engine, several areas

The same processing engine serves every area — introducing one gives you the whole engine.

Data extraction from documents

Field-level KIE from invoices, contracts, payslips and bank statements — with confidence signals.

Contract management

Central contract portfolio, 30/60/90-day expiry monitoring, expected vs. actual billing.

Document chat

Ask your own archive in natural language and get source-referenced answers.

All use cases →

Contact

Get in touch

Fill out the form below and our team will reach out shortly to schedule a demo.

We only use your data to respond to your inquiry. We do not share it with third parties. Legal basis: your consent and pre-contractual steps (GDPR Art. 6(1)(a) and (b)). For details, see our privacy policy.

Automated invoice processing for Hungarian companies