Skip to Content
Product Information

Intelligent Invoice Scanning

In this blog post, I will have a closer look at the ‘Intelligent Invoice Scanning’ feature, introduced in SAP Business ByDesign with release 2008. There are two key differentiators compared to the existing scanning solution:

  1. It is now sufficient to only upload PDF documents; a TXT document is not needed anymore.
  2. Scanning of the PDF is done using SAP’s Document Information Extraction Service (DOX), including an integrated OCR service as well as a machine-learning based model.

Supported languages are English, German, French, and Spanish. Please be aware, that the DOX service is currently not available in a Chinese data center, and therefore cannot be used in ByD tenants hosted in China.

In addition, I am pleased to mention that usage of the new scanning service is included in the normal SAP Business ByDesign licenses, no additional costs will be incurred!

This graphic shows the end-2-end process flow. It starts with uploading the PDF documents from your local file share using a new work center view ‘Invoice Scanning’. A mass upload of PDFs can also be done using SAP’s Collaboration Window, This can be downloaded from the Self-Service section in an SAP Business ByDesign system (Work Center ‘Home’).

Once uploaded, the PDFs are automatically sent to SAP’s Document Information Extraction Service,  and invoice header and item information is sent back to SAP Business ByDesign. Since supplier master data can be uploaded to DOX, the supplier ID can directly be mapped to the scanned supplier information, like name, address, or other contact information.

The scanned information is first stored in a staging object, where the results can be reviewed, corrected, and once fine, the supplier invoice can be created.

Let’s have a look at the steps in more detail.

To activate the solution in your tenant, you must set the ‘Do you want to enter supplier invoices via PDF scanning?’ question in scope. If you want to automate the creation of the invoice in addition, you must also select ‘Do you want automatic creation of invoices from scanned documents?’. This means, that automatically a supplier invoice is created from the staging object.

No further setup needs to be done to establish the connection of the SAP Business ByDesign tenant to the DOX service.

Next, the new view ‘Invoice Scanning’ of the ‘Supplier Invoicing’ work center needs to be assigned to the relevant users.

I recommend to upload the supplier master data to DOX before the first invoices are scanned. This can be done in the ‘Invoice Scanning’ view in the work center ‘Supplier Invoicing’. Currently, whenever new scanning-relevant supplier master data have been created or changed, this upload should be repeated.

The action ‘Synchronize Master Data’ launches a screen, where the synchronization can be started.

Now, the system is prepared to start the scanning process.

This can be done by the ‘Upload’ action in the ‘Invoice Scanning’ view. A screen is launched, where the company can be specified for which you want to upload the PDFs. Then, the corresponding PDF documents can be selected from a file share, and via ‘Start Process’ are handed over to DOX.

As mentioned above, a mass upload of PDF documents can best be done using SAP’s Collaboration Window. Please do the following configuration settings under Menu – Settings – Invoice Scanning:

 

Please flag both options and specify the ‘Observed’ folder, from where to upload the PDFs. Next, you need to copy your invoice PDF into the observed folder, from where they get automatically uploaded, if the Collaboration Window is open.

The company information is retrieved from the user logged in to the Collaboration Window.

Independent of the way you use for uploading the PDFs, they then show up in the ‘Invoice Scanning’ view:

The status indicates, if scanning has been completed. You might need to refresh the list after a while. Once scanning is complete, you can take a look at the scanned values using the the action ‘Review’, that opens the staging object:

The scanned values are visible in the read-only fields on the left side of the UI. In case some values have not been scanned correctly, or have not been recognized at all, the reviewed values section allows you to manually override them. In my example, the supplier was not recognized. Once you are fine with the result, the invoice can directly be created from this UI, or from the ‘Invoice Scanning’ work list.

The invoice, that got created, can of course be reviewed once more in the ‘New Scanned Invoice’ UI, where also additional data like accounting assignment can be added.

With that, the invoice scanning and creation process is complete. Please also check out the video.

 

7 Comments
You must be Logged on to comment or reply to a post.
    • Hi Andre,

      currently this in not supported, but I would ask you to create a customer influencing request for the same, so that we can see how many partners and customers vote for it.

      Best regards,

      Stefan

  • Hi Stefan!

    We have run a few tests to check the capabilities and make some experience with this new feature in our test tenant.

    The result is a mixed bag:

    1. The supplier synchronization did not work at first but was later fixed by the SAP support.
    2. Configuration and set up was a breeze, thanks for the blog & video.
    3. I loaded a handful of pdf files manually (prepared to match the existing PO, same supplier, no invoice/PO differences, straightforward invoice layout, GR complete, clear information with dates, field names identical to the mapping tables in fine tuning).
    4. There were numerous issues that made me doubt that this was an AI infused application. I am happy to show you the test set up.
    5. The interesting issue was that even after the “scanned values” were adjusted with the “reviewed values”, when creating the invoice, those values were disregarded by the process. Basically we they had to be entered again. Here I am not sure if this is wanted or an issue that should be reported.

    The classical process (with pdf and txt files) works much better – so far. But it needs the two files (prepared by OCR scanning manually or using an automated process).

    But it would be good to learn some details about the AI features so that the functionality is fully understood, and expectations set accordingly. Things like: How was this tested? What model was used? Are there statistics of success vs manual handling etc.

    Thanks

    -Jan-

    • Hi Jan,

      it would make sense to have a short meeting to discuss the next steps. What would be the best way to contact you?

      Best regards,

      Stefan