Using a predictive database for AI-enabled RPA

Photo by Lukas
author

A lot has been written about the importance of Robotic Process Automation (RPA) and why it is necessary. And even more, how combining both AI and RPA is combining the best of two worlds.

However, making the robotic processes more intelligent with AI still seems to pose some challenges. The misperception is that making your RPA AI-enabled is difficult, time-consuming and expensive. And that each process needs to be well-defined and have a narrow problem to solve.

While it is true that the problem should be defined properly, the process does not need to be narrow or automated completely. It does not have to solve the whole business process from A to Z, but can be chopped down or solve a part of the process while using these technologies in collaboration with human work.

This is where Aito comes in. Aito’s predictive database is a generic, affordable and lightweight solution, easy to integrate with your existing robots. It allows you to make the most of the robotic processes more intelligent.

Think of

  • Predictive project expenses
  • Sales recommendations
  • Job application management
  • Contract management
  • Predictive production planning
  • Classifying & managing e-mails

Don’t take our word for it – let’s dig deeper in one simple, but common, use case. Together with Sisua Digital, a disruptive company specialized in Robotic Process Automation, we set up a POC (proof of concept) for invoice posting automation.

The process

256 invoices were exported from the financial program Netvisor and converted to JSON for upload to Aito. The robot opens the invoices one by one from Netvisor and reads the necessary fields: supplier, account number and purchase amount. After this, the robot asks Aito to predict these fields in the account assignment spreadsheet. Aito also predicts the accuracy rates and will only fill in the fields if the accuracy rate is high enough. If it is not, the user will be urged to fill it in the necessary fields manually. Aito will learn along the way so the accuracy will be improved all the time. An accuracy of 77% was reached with this small data set. The robot continues the process until all invoices are processed.

Phase 1: Prepare data, define results and train the algorithm.
Phase 1: Prepare data, define results and train the algorithm.

Shown activities are limited to the steps involved in the invoice posting. Other steps can be automated normally with rule-based automation.

Phase 2: starting the pilot production.
Phase 2: starting the pilot production.

To see everything in action, watch our 90 sec demo. For this demo, you see the invoices processed one by one. However, in real-time, this process would be hidden for the user and only the end-results would be shown. The whole process of handling the 256 invoicing and putting them in the spreadsheet took 1 minute.

When all the purchase invoices are processed, the robot could move on to e.g. approving the invoices for payment and send a confirmation email about the handling to each recipient.

Want to make your own robotic process automation more intelligent as well? Book your demo today!

Back to blog list

New integration! Aito Instant Predictions app is now available from Airtable Marketplace.