nextAI - Line recognition

nextAI - Line recognition supports the automatic processing of invoice lines within AP Automation.

Where nextAI - Header recognition focuses on the header data of an invoice, line recognition identifies, interprets and structures the individual invoice lines.
With nextAI - Line recognition, lines such as items, descriptions, quantities, prices and costs are automatically identified and processed. This increases invoice automation, reduces manual actions and applies an additional check.

When do you use it?

Line recognition is used when processing invoices where detailed information at line level is important.

Typical situations:

• Processing invoices with multiple lines (for example, purchase order or expense invoices)
• Automatically coding invoice lines
• Processing invoices with multiple VAT rates
• Invoices with additional costs such as transport, packaging or surcharges
• Situations where insight into the cost allocation per line is required

How does it work?

Step 1: Submit the invoice
A PDF invoice is processed within AP Automation. nextAI analyses the document and identifies both header and line data.

Step 2: Perform line recognition
In addition to the invoice header data, nextAI identifies the individual invoice lines and recognises, among other things:

• Line descriptions
• Item numbers
• Quantities 
• Prices
• VAT percentages
• Additional costs
• References such as order numbers

Step 3: Structure and code
The recognised lines are automatically converted into structured data and linked to the correct fields in AP Automation. This supports automatic matching and further processing.

Step 4: Check and refine
The recognised lines can be checked and adjusted where necessary. Recognition can be further improved for future invoices using the 'Feedback' button and prompts.

Practical tips

• Use your own prompts to improve specific line structures or exceptions
• Check new or deviating invoice layouts particularly carefully during initial processing
• Apply line recognition in combination with header recognition for maximum automation

Common mistakes

• Expecting all line structures to be recognised fully automatically straight away without checking
• Incomplete invoices or incomplete orders for matching invoices with orders. For example, the item number may not be on the order, or information may be missing from the invoice to match the order to the invoice correctly without making assumptions.
• Not using your own prompts for recurring deviations

Frequently asked questions

What is the difference between header recognition and line recognition?
Header recognition focuses on general invoice data such as the invoice number and supplier. Line recognition processes the detailed lines of the invoice.

Is line recognition always fully automatic?
Line recognition automates most of the process, but checking and any necessary corrections remain important for exceptions.

Can line recognition be customised?
Yes, recognition can be refined and tailored to specific suppliers or invoice types using prompts and feedback.