Posti boosts purchase invoice automation with Aito

The RPA team of Posti has moved to the next level by adding machine learning into their toolbox. Using Aito has allowed Posti to utilize machine learning independently and the first Aito implementation is now live in production, tirelessly churning through thousands of purchase invoices and saving their finance team countless of hours worth of mechanical work.

Automation in the Finnish logistics giant

As the largest postal service operator in Finland, Posti delivers hundreds of thousands of parcels and envelopes daily around the country. This naturally involves much more behind the scenes than just the logistics though, and automation has been at the core of Posti for years.

To streamline its operations, Posti’s RPA Center of Excellence has been automating countless business processes within the company. With Aito now in their toolbox, the RPA team can conquer even complex processes without having to go through the data science pipeline. No custom code or scripts are needed either since interaction with Aito happens directly through standard UiPath activities.

Handling purchase invoices

By using Aito and in collaboration with Sisua Digital, Posti has automated their purchase invoice handling processing. Processing the 3000 new invoices every month requires several cognitive decisions, making automation impossible with traditional RPA methods. Before storing the final version in their invoice management system, each digital invoice needs to be:

1. Assigned to the right reviewer
2. Allocated to the right cost center
3. Tagged for the right category
4. Corrected in case of missing information

Visualizing the problem with some mock data, this is how a new invoice looks like entering the system. A correct value needs to be selected for the empty fields.

Posti has automated this process by creating a software robot with UiPath and having it interact with Aito. The robot reads the information in each new invoice and asks Aito to predict the right values.

Launching the robot

To test the performance of the new automation, it was first operated in parallel with the manual process for one month and all the predictions of Aito were recorded for evaluation. During the simulation, Aito’s predictions fulfilled the accuracy requirement of 95% set by the project steering group, giving it the green light for production. The first version of the robot includes a few of types of purchase invoices in one legal company. Later during this the fall, more companies and invoice types will be added.

During its first couple of months in production, more than 7,000 purchase invoices have been processed automatically. The new automation has reduced the workload on the Accounts Payable team at Posti while boosting the speed of invoice review. But even more importantly, the RPA team of Posti now has the ability to replicate the technology for countless other business processes. After the initial learning curve, they found Aito very straightforward to use with standard UiPath functionality.