Considering OCR? Read This Before Making Any Decision

Blog - 12 Aug 2020

If you are currently looking at ways to improve your accounts payable process you may have come across Optical Character Recognition (OCR) technology. OCR is widely thought of as being the original invoice automation tool. However, with the development of more comprehensive automated accounts payable solutions, the limitations of OCR have become apparent. While OCR can be a very useful tool for accounts payable teams, OCR removes only a small fraction of manual processes that fully automated solutions handle.

You should be fully aware of the problems that exist within your AP process before making a decision on any tool or system.


How OCR Works

In the Accounts Payable world, OCR is a software that scans an invoice and converts the data into editable text, which can be used to populate fields in a database/AP system and create an electronic version of the invoice.

OCR can be trained to find the position of certain data points on an invoice through the use of intelligent filters and searching for commonly used keywords in that proximity.

All good OCR systems should be easy to implement and set up with no pre-training of vendor invoices required. This is particularly important if you have a large number of vendors. Its self-learning capabilities also mean that as invoices are processed the technology will improve its accuracy. OCR systems are also very flexible tools, capable of working with multiple languages and data inputs e.g. email/paper/electronic.

OCR is a useful tool for extracting invoice data, however, beyond that its capabilities can be limited. Before making a decision on OCR, you should assess each element of your accounts payable process and whether or not data extraction is your only problem.


5 Key Scenarios to Consider

If you are considering OCR, you should assess how long it currently takes your AP staff to key invoice data. Asking this question will unlock a number of AP scenarios that should help you decide on whether OCR will handle your issues or if a more comprehensive AP solution is needed.


Scenario #1 – PO Invoices

When processing a PO invoice and all data matches between the documents, it is often the case that very little manual keying of data is actually required. In this case, OCR will do little to improve your AP process. You should gather this information from your AP staff before making a decision. It may be the case that other processes are actually far more time-consuming. Automating the rest of the process may in fact be what is taking the time, and therefore OCR is not the answer.


Scenario #2 – PO Invoice Not Receipted or Containing Wrong Price

Within your current process, what currently happens when a PO invoice is received but it has not been receipted or contains the wrong price? Is the transaction canceled? Is the invoice put to one side or is the relevant person emailed to ask for approval or confirmation of delivery?

While OCR can capture certain data points, it will not know if the data is incorrect or if something has not been receipted. How would your AP team handle the situation if OCR just sent the invoice data across to your ERP/Finance system even if the price was wrong or the goods had not been delivered?


Scenario #3 – Partial Invoices

Do you currently receive invoices that relate to specific lines on a PO? Very often with partial invoices, users have to manually identify and select items on the PO to which the invoice relates in order to post for payment. OCR may struggle in such a scenario to link the invoice data to the correct PO line, meaning manual intervention is likely required.


Scenario #4 – Duplicate Invoices

How do you currently detect and handle duplicate invoices? If this is currently a problem it may continue to be with OCR as the technology does not possess the ability to detect duplicates and may post a duplicate to the ERP.


Scenario #5 – Validations/Posting

You should think about your current validation process and the minimum amount of data required to post an invoice to your ERP. Contemplate the implications if OCR were to send an invoice missing key data to your ERP. Does your ERP expect each line of the invoice to reference the specific order line number to which it relates? – As mentioned already, this will need to be handled manually, even with OCR in place.

If any of these scenarios pose potential issues within your AP process, OCR is not the solution to your problems and you should instead look at a fully automated accounts payable system. When exploring potential solutions, you should make sure to focus on the capabilities of AP solutions that address such issues.



6 Key Features of an Automated Accounts Payable Solution



As has been established, OCR is a good tool for capturing invoice data. When choosing an automated accounts payable system, look for one that includes OCR so any data capture problems that you have will be addressed. The OCR included should not require any pre-training of your vendor invoices and should be capable of handling multiple languages and data points and be able to learn from invoices and improve accuracy.



One of the areas that OCR falls short on is the validation of invoice data. An automated accounts payable system can perform a number of data validation checks for example, that all necessary invoice data is included for matching, that remittance details match those on record and that the invoice received is not a duplicate.



Due to the complexities often involved in matching invoices to POs and GRNs, OCR can struggle to link the data and manual matching will be required. A comprehensive automated accounts payable system will handle all matching scenarios. For example, AI could be used to identify and link invoice data to appropriate PO lines and GRNs in the case of partial invoices and where invoice data is incomplete. Automated matching engines within AP automation also handle a number of other tasks that facilitate matching, such as validating that prices and amounts are correct, escalation of invoices for receipting where required, identifying or accounting for any price variances both inside and outside of tolerances and also routing any invoice queries for resolution.


Non-PO Invoice Processing

If you currently process non-PO invoices, there will likely be a lot of manual work involved in assigning invoices to cost centers and approving invoices. OCR will do little in terms of improving this process, however, with the right automated accounts payable systems, you can assign non-PO invoices to cost centers through pre-configured rules. Default approval routes can also be assigned to such invoices.


ERP Integration

If posting invoice data directly from OCR to your ERP you run this risk of posting without the required data for payment, or posting duplicate invoices without validation. Automated accounts payable systems integrate with ERP systems to allow for full posting of invoice data following matching, with linking to POs and GRNs within the ERP.



For AP staff, one of the most stressful and time-consuming elements of the role is reporting. Often, much more so than data capture. An important feature of automated accounts payable systems is the availability of intuitive reporting dashboards. Available reports range from accrual reporting, operational reporting (outstanding invoices by user and supplier), straight-through processing reporting to payment practice & performance reporting. Such reports help AP staff to both improve productivity and optimize working capital for the business.

If you believe invoice data capture is the only issue within your AP process, you need to ensure you review fully your entire AP process to make sure that is definitely the case. You may discover that your issues run far deeper and in fact, a fully automated accounts payable solution is required.

If you are interested in learning more about how you can address your AP issues with a fully automated accounts payable solution, you can request a demo here. A member of our sales team will then be in touch.

Leave a Reply

Your email address will not be published. Required fields are marked *

Posted by

Robert Lynch

P2P Insights Analyst

Posted by

Robert Lynch

P2P Insights Analyst