Skip to Content
Product Information
Author's profile photo Gloria Benitez Utrilla

Simplify your Procurement Process and Automate Accounts Payable with Invoice Object Recommendation (Part 3)

This blog post is part of theSAP AI Business Services introductory–, and product portfolio series (see further published parts of the series listed below under “More Information”). 

Manual invoice processing is extremely time-consuming and error-prone. Sudden increases in volume often result in expensive invoice matching errors and compliance issuesManual invoicing often requires at least 15 steps before the final posting is done and the invoice can be paid. Many companies process around 1000 or more invoices per day, taking up to at least 15 minutes of manual processing time per invoice. AP automation can save companies valuable time otherwise lost on repetitive and redundant invoicing processes. In fact, organizations using AP automation can process over 4 times as many invoices as those without automation, reducing the invoice processing time from 20 days down to only three days on average.  

Accounts Payable and Procurement are part of the same process and must work closely together to ensure efficiency, cost control and accuracy. If the procurement process allows uncontrolled, rogue spending, then invoices will appear as a surprise in the Accounts Payable department. Accounts Payable and Procurement are strategically connected to streamline the supply chain and find ways to cut costs without compromising quality every step of the way. The better the collaboration between Accounts Payable and Procurement is, the better the supplier relationships that can be developed, resulting in considerable cost savings in the future. The reality however is, that huge discrepancy between the two business areas often causes heavy financial losses.  

Accounts Payable automation plays a critical role in this process. Accounts Payable automation (AP Automation) refers to technology that is used to streamline and automate accounts payable processes, removing manual tasks and providing better visibility and control over important financial data. Thanks to AP Automation everyone in the company receive access to relevant supplier and payment data in real-time, hence employees can save money and improve interactions, such as offering discounts for early payments to suppliers. AP automation can save as much as $16 per invoice or more, depending on the size of your company. On average, organizations using AP automation save 19% on invoice processing. 

Accounts Payable Automation through Invoice Object Recommendation

Invoice Object Recommendation is an SAP AI Business Service supporting the accounts payable department’s service agents in their decision making, through the prediction of correct account and cost center assignments. By predicting the right values through the usage of machine learning technologies the solution simplifies the internal process, increases productivity as well as the quality of invoice processing.  

 Based on historical data the model recommends the top three General Ledger Accounts and Cost Center choices to any invoice without a previously assigned PO. This enhances financial reporting by minimizing roadblocks due to discrepancies in accounts/budgets, as well as reducing repetitive and redundant tasks, such as manually searching and selecting the correct General Ledger Account or Cost Center, as well as correcting wrongly assigned ones  

The Challenge of Finding the Correct General Ledger Account

Jane Doe is working as an account’s payables clerk at Company X. Her everyday job is to process invoices. Jane experiences an especially hard time in her daily job when the incoming invoices do not mention Purchase Order. 

In order to process an invoice without Purchase Order, Jane needs to extract the needed information from the invoice (e.g. Vendor, Invoice Date, Tax Code, the Total Amount to be paid) and then, matches them to the corresponding General Ledger Accounts and Cost Centers. Finding the right General Ledger Account is a highly repetitive & also cost intense task. Hence, a missing Purchase Order means for Jane, that she needs to spend a long time searching for the right cost centers and G/L accounts, losing valuable time searching the correct invoice details.   

We have seen that invoices without reference details such as a Purchase Order can constitute quite a problematic task to service agents since they need to invest a lot of time manually inserting the required information.  


Make your Invoice Processing Faster through Invoice Object Recommendation

Through the help of the Invoice Object Recommendation Service Jane has now the possibility to process invoices without a Purchase Order much faster. The information given in the invoice is filled out automatically in the ERP-mask. In the next step, Jane will get predictions (e.g. one to three top choices) with a certain probability/accuracy percentage of the correct General Ledger Account and/or Cost Center. These predictions are based in two different aspects: the inference data (all data which is found in the invoice) and on the other hand based on the training data, which the company uploaded based on their historical data gathered in the past ~6 months.  

Customers can now decide individually whether they want to automate the process entirely, such as based on a probability rate of 90% for example, or whether they want to leave the final decision to the employee who is provided with the predictions and can still control the final step and inputThe process is followed by a Machine Learning based check on the invoice to make sure the correct account has been assigned and charged. 



Invoice Object Recommendation service finally allows a higher accuracy throughout the Accounts Payable process with quantifiable resultsThrough correct assignment choices, the machine learning-based service prevents employees from spending additional effort on correcting the accounts after payment has been processed. It furthermore benefits companies by enhancing financial reporting by increasing the accuracy of financial reporting and minimizing roadblocks due to discrepancies in accounts/budgets 

Lastly, the solution helps to free up valuable time for Accounts-Payables-Clerks which can be rather used on more value-adding, strategic and fulfilling tasks.  

Technical Specifics – Deep Dive


The Invoice Object Recommendation Service only takes structured data into consideration. 
Information like Vendor, Price, and Tax can be directly seen on the invoice and thereforeit’s being extracted immediately. On the other hand, information like item text information, items or services specified in the invoice, won’t be extracted immediatelyWhatever information is not immediately extracted from the invoice, the Clerk is the responsible one to look up the necessary information (item text information) 

The service aims to us all those features (Vendor, Item Price, Tax Code, Company Code, Document Type, etc.) are used in a classification model to predict the correct General Ledger Account and Cost Center. 

Moreoverby using other services in addition, such as Document Information Extraction, the accuracy of the prediction can be improved. 



More Information:

Assigned Tags

      You must be Logged on to comment or reply to a post.
      Author's profile photo Avijit Dhar
      Avijit Dhar

      Hi Gloria,

      Wonderful blog. I just want to know what ML technique is in the service. Is this based on TensorFlow?

      Thanks and Regards

      Avijit Dhar

      Author's profile photo Gloria Benitez Utrilla
      Gloria Benitez Utrilla
      Blog Post Author

      Hi Avijit,

      Thank you so much for your positive feedback ?

      To answer your question, it is not a deep learning model - since it's a classification service we use a classification model. Hope that helps!

      Please let me know if I can answer any other questions ?

      All the best,

      Author's profile photo Lakshmi Sankergi
      Lakshmi Sankergi


      Would like to know the details on training dataset required for the service. Required fields are not mentioned in the help document.


      Lakshmi Sankergi




      Author's profile photo Gloria Benitez Utrilla
      Gloria Benitez Utrilla
      Blog Post Author

      Hi Lakshmi,

      Hope you are doing great.

      Apologies for the late reply,
      You can find the details in the help portal, for direct access please follow this link:

      Thank you and all the best,