“Artificial intelligence has really upped our automation level. For us, this means smoother handling of purchase invoices, as well as time saved.”
- Katja Somppi, liiketoimintapäällikkö, Sarastia
+ 100 milj.
Sarastia is Finland's largest public sector financial and personnel management support service partner and a producer of modern municipal services.
Sarastia is Finland's largest company providing support services to the public sector. It is best known for its financial and personnel administration service operations, which include accounting and payroll services for municipalities and municipal associations as well as for the future welfare areas.
The invoice volumes are huge in the public sector. Public institutions are looking to optimize their processes by improving ways of working and through information technology systems. This article focuses on Sarastia’s service-producing unit that works with the Social and Health Services in Kymenlaakso (Kymsote) and its observations on improving the processing of purchase invoices.
Kymsote receives approximately 150,000 purchase invoices annually. Automation has been actively built into Sarastia’s purchase invoice system, but the organization has had to face the limits of rule-based automation. Before the introduction of artificial intelligence, Sarastia had about 90,000 invoices that had to be manually processed in addition to those invoices it handled through rule-based automation. Such a large number of manually processed invoices required a lot of manual work from Sarastia and created many challenges for invoice processing.
"We started using the Snowfox.AI purchase invoicing AI through the public procurement process by following the Procurement Act. We wanted to increase the degree of automation of the purchase invoice processing process, and artificial intelligence was seen as an interesting possibility for this", says Katja Somppi from Sarastia.
Snowfox's purchase invoice AI was introduced as additional functionality to the current purchase invoice system. There was no need to change the system, as AI is designed to work together with the existing purchase invoice system.
"We process more than 100,000 purchase invoices annually. Some of these have been automated with the built-in automation of the purchase invoice system. AI now processes approximately 90,000 purchase invoices per year. AI has allowed us to take a leap forward with our automation levels. Thai results in smoother processing of purchase invoices and in time savings", Somppi explains.
During the implementation phase, AI learned from Sarastia's historic purchase invoice data how the invoices had previously been posted, checked, and routed. Based on this information, AI created an AI model that automatically posts and routes purchase invoices Kymsote receives.
AI model uses all the information from the purchase invoice and the image of the invoice to predict the posting and routing information. Even if individual information is missing from the invoice, AI can still automate it effectively. AI utilizes all possible information from the purchase invoice and is not startled if the invoice is missing information, like a supplier number or any reference information.
It is possible to modify the accounting and reviewing information manually during the posting and routing process if AI has made an incorrect prediction for one of the dimensions. After the approval round, AI receives feedback about the final invoicing. Based on this, AI checks whether the predictions are correct and uses the feedback to learn and improve the model.
In this way, AI learns and develops from each purchase invoice it handles and can process invoices better and better.
AI started processing invoices at a moderate level in Sarastia, and it has developed significantly during its use. This development is characteristic of artificial intelligence because it receives feedback on how it processes each purchase invoice.
The prediction accuracy of almost all the selected dimensions is over 90%, even close to 100%. This is an excellent level for AI and shows that the purchase invoice data is high-quality.
Sarastia's goal in processing purchase invoices is to reach the highest possible automation level. This brings a positive challenge to the development of Snowfox and helps to develop the service.
In connection with the implementation of the purchase invoice AI, Sarastia also has access to real-time analytics, which it uses to monitor the development of AI. The analytics show how the AI automates the purchase invoices and how well the AI has succeeded in posting and routing. In addition, the analytics show how confident or uncertain the AI is for each prediction.
Analytics help both Sarastia and Snowfox to see the capabilities and development areas of artificial intelligence. This is very important, because the operation of artificial intelligence is constantly being improved, and the identification of development areas is possible through analytics.
"We are very excited about the cooperation with Sarastia and the good predictive accuracy of artificial intelligence. The cooperation continues in excellent spirits and we enjoy working with Sarastia's professionals."
– Miikka Savolainen, Snowfox.AI
Just leave us a quick message and we can contact you to talk about what benefits our other clients are getting from AI automation.
Or are interested in implementing Snowfox.AI, but you’d like to know more first? No problem! Just contact us and we together we can find out how we can help you automate your purchase invoicing process with AI.