Automatic creditor invoice management with RPA

1. Before all, understand your process

Before automate anything, you need to understand your process by defining each step that is required, each actions you or your colleagues needs to do. A good way to achieve that, is to draw a workflow and/or a decision tree. You also have tools like Task Mining or Process Mining that help you to collect what happens in the process. Often, we do the analysis for the client to reduce their workload.

For today’s process, we have:

• Input variation:
  • 3 communication channels where invoices arrive: By mail, E-mail and E-Bill (directly in e-banking)
  • A list of providers with each his own invoice format (known in advance!)
• A tool to manage the accounting and payments of the company where the steps are standardized
  • A decision flow exists to know which account is concerned by the invoice
• A validation step of the payments entered in the system

Each point defined above is broken down into steps and all of them are documented.  By being precise, you can define safeguarding rules (reference number must be 27 digits, for provider A the total amount must be below 5’000CHF…)

2. When automating, set a realistic scope

Well, my robot won’t open the mail. No arms, no mail we say… But maybe I could scan or take a picture of the mail and have some image processing and a robot assist me. Reducing my workload.

The idea is to simplify your life, to reduce your administrative workload, to generate time. The overall process might be very complex with a lot of exceptions, but by breaking it down into sub-processes, you will see that some of these sub-processes are straightforward and make sense to automate. So, maybe start with just 2 or 3 of them with some invoice formats that have a big volume. Divide and conquer! Keep in mind that with RPA, it is not required (and not necessarily wise depending on the process) to automate 100% of the process. In several days, you would already have a big chunk of work automated. With that additional time you now have, you can expand the automation to new cases.

Start small, start smart.

3. Demonstration of automatic creditor invoice processing by RPA

In the following demonstration, we will see an UiPath RPA assistant processing an invoice for telephone costs. The invoice is received by email and the assistant gets it in the background, reads it and extracts the values of interest. We don’t see any of that happening (expect in log messages). The next step is to input all that in our tool and also to prepare a payment order. This part is visible because the assistant interacts with the tool, here a WebApp named Bexio. Once it is finished with one invoice, it proceeds to the next until their are no more.

Now the employee only needs to do 1 click on the tool to validate and send the payment order to the bank (yes, that click is also automatable). A lot of time is gained and we removed the risk of mistyping. Have a look: 

4. Conclusion and what is required for such a robot

In this video, there was no artificial intelligence. The robot/assistant knows exactly what to put and where for each invoice thanks to a decision tree. Textual example:

• Is it an invoice from Salt/Lidl ?
  • Yes, then put it on 6570 IT costs.
• Is the invoice from Groupe Mutuel?
  • Yes. Is it the invoice for accident insurance?
    • Yes, then put it on account 5730
    • No. Is it for Daily Sickness insurance?
      • Yes, then 5740.

If we wish to change a rule of that decision tree, it takes only several minutes. Like changing the account 6570 to 6510 (communication costs).

To automate this process, it only took 2-3 days to implement, test, stabilize and put live. That is because everything was clear and known in advance and we did not have any deviations in the process. The implementation of a virtual RPA robot requires only a few days of work by an external expert and a few hours of work by the customer. It is of course possible to adapt the robot to your needs and to work with your own applications (SAP, Oracle, Abaccus, Microsoft…)! We can even add OCR or make it communicate with an AI.

Do you wish to discover more? Do you want your own automatic creditor invoice management with RPA? See our other articles or contact us!