When a document is uploaded, Candis tries to fill in its fields automatically. You can recognize this by the green or orange dot next to the value.
Various sources are used for reading, these can be seen in the tooltip.
For example, the invoice number of this document was read from the document. Other fields, such as the IBAN, can be stored at the business partner if this was not found on the invoice.
The following data sources are used for reading fields:
Document
Transaction
Credit card
Contact
Cost center/cost object
Field suggestions
Data source: Document
In this case, the suggested data was read from the document itself. This can happen in different ways:
Optical Character Recognition (OCR):
The process involves capturing letters, words and numbers in image files, such as scanned documents, and converting them into text. In some cases, recognition can be subject to errors. Example: Whether a character is an “O” or a “0” (zero) can only be recognized by the context.
The certainty (confidence) of a value is shown in orange and green:
Orange: The field is classified as unsafe. These fields should always be checked.
Green: The field is classified as safe. These fields can be skipped.
Fields change from orange to green over time. If documents from a supplier are often read out and not corrected by you, they will be displayed in green in future, even though the data source is the document.
Candis Template:
If the text recognition is inaccurate for certain documents, Candis Support can create a so-called template for certain incoming invoices/suppliers. How can OCR recognition be improved?
E-invoices:
An electronic invoice, or e-invoice for short, is a machine-readable document that provides its data in a structured form. The data does not have to be laboriously read by a text recognition system, but is always correctly available.
Candis supports the two currently common types of e-invoices: ZUGFeRD and X-Rechnung.
Data source: Transaction
If invoices are linked to transactions, the transaction data is displayed in the fields instead of the invoice data.
Candis only exports transaction data with document image and therefore certain invoice data cannot be recorded or transaction data cannot be changed.
Data that is used from the transaction is:
Document type
Invoice date
Delivery date
Due date
Invoice correction
Total amount
Currency
Data source: Credit card
The credit card data source is displayed if booking rules have been added to the credit card.
Data source: Contact
The following fields can be proposed via the business partner:
Create bank transfer
Payment term (cash discount date, cash discount rate and cash discount amount)
Due date
Candis uses the following data to recognize the business partner of a document itself:
Company name
VAT ID
Tax number
IBAN
If the recognition of the contact is incorrect, please message an “administrator” or “accountant” so that they can check the data assigned to the contact. It is often the case, for example, that the company's own IBAN, tax number or VAT ID have been stored for the business partner.
Data source: Cost center/cost object
If the cost center or cost object is read from a document, the corresponding approver is also added automatically.
The prerequisite is that an approver has been assigned to the corresponding cost center or cost object. You can also add several people. These are all added to the 1st approval step. This means that only one of these persons has to approve.
If this is not (or no longer) correct, please contact the responsible team member with the Administrator role. Only the administrator can change the settings and add/remove a new user.
Data source: Field suggestion
The data from this field has been suggested based on past entries. There are field suggestions for the following fields:
General ledger account
Tax code
Cost center
Cost object
Approval (workflow; approver)
If a cost center has been assigned to a supplier's invoice for the first time, Candis will suggest the same cost center for the next invoice from the same supplier. For suppliers with more than 20 processed invoices, the last six months history is used as a basis for the suggestions.
Data source: AI Field Suggestion
The data from these fields is suggested by an AI model based on similar past entries. Unlike simple rules, the AI retrieves the most similar invoices from your history to understand the context and generate a suggestion. There are AI field suggestions for the following fields:
General Ledger Account: Suggested based on the invoice amount, content, contact, tax rate and origin country (e.g., EU / third country)
Cost Center / Object: Suggested based on the invoice address, content, recipient, project or customer relations and the (CC) card holder
Tax Code: Suggested based on tax rates and amounts, type of service, VAT-IDs (Organization and Contact), country of contact and references to tax exemptions (e.g. reverse charge)
Posting Text: The posting text is based on the contact, invoice content and invoice date (available in the Max Plan)
We identify the most similar invoices to the incoming document and uses their assigned data (the Context) to predict the correct field values for the new invoice.