Palkeet develops invoice automation with the help of Snowfox.AI artificial intelligence
The Finnish government shared services center for finance and HR (Palkeet) is a provider of group services in the administrative branch of the Ministry of Finance. Palkeet provides finance and HR services for central government agencies, departments and funds, as well as state-owned businesses and fully state-owned limited companies that provide services for the Finnish government, as well as individual government employees.
Palkeet had a public tender for an AI-based posting service in spring 2021. The winner of the tender was Aitomation Oy, that offered the Snowfox.AI AI service for Palkeet.
“Over 400 000 purchase invoices that could be automated by utilizing AI are posted manually in the financial services provided by Palkeet. The more manual labor the invoice workflow contains, the more you need people to handle the invoices which leads to a more expensive process.
The goal of the tender was to find a ready AI-based service, that could raise the automation level of the now manually handled purchase invoices to a high-as-possible level. Artificial intelligence was seen as a very necessary technology as it provides a change to raise the automation level of non-PO invoices significantly.” – Eija hartikainen, Chief of development, Palkeet
“It’s wonderful to see that a state-owned company invests in developing their processes through modern technology. The AI-based service we have developed fits very well to the current situation of Palkeet and their way of constantly developing their operations to be more fluent and efficient.” – Miikka Savolainen, COO, Snowfox.AI
How does the AI work?
The AI learns based on the historic purchase invoice data and how the invoices have been posted earlier. Based on this the AI automatically creates a model that can post the incoming invoices automatically.
The AI utilizes all the information on the invoice, also reading the image of the invoice when creating predictions for the invoice. Even if there is information missing from the invoice, the AI can still automate invoice processing efficiently, as it utilizes all the information and doesn’t get confused about e.g. a missing supplier number or reference number.
It is possible to modify the postings during the circulation, if the AI has for example predicted a wrong value for a certain accounting dimension. After the invoice circulation has been completed, the AI receives feedback from the final postings and uses this information for fine-tuning its models. This results in the AI constantly learning and being able to handle the invoices with a growing precision. All this happens fully automatically in the Snowfox.AI AI-service.
The quality of the AI-service was tested during the tender process
One part of the public tender process was testing the capabilities of the artificial intelligence. Palkeet sent historic purchase invoice data and their posting information to all participants in the tender process for training their AI models. After this Palkeet sent a new batch of purchase invoices, that the created AI model had to provide postings to. Participants in the tender process then sent the postings their AI had made to Palkeet, that analyzed their validity and scored the performance of the different AI-solutions.
At Snowfox.AI we liked a lot of the idea by Palkeet to take testing the capabilities of the AI into the tender process. This way they could validate the capabilities of the offered solutions and the scoring was not based only on promises.
“I’m really excited about our beginning collaboration with Palkeet. I’m especially proud because our AI-based service was the best in the tests during the tender process and our benefit-based pricing model appealed to Palkeet.” – Tuomas Haapassari, CGO, Snowfox.AI
The AI is connected to the existing invoice management system
The implementation process for getting the solution operational for Palkeet has already begun. The purchase invoice AI will be connected to the current, existing invoice management system of Palkeet.
The deployment of the Purchase invoice AI will be done in stages. The project is already on the way with preparations for a pilot period and the AI will be fully operational for all customers of Palkeet by March 2022.
When the AI becomes operational, the only thing that will change in the day-to-day operations is that the purchase invoices that now have to be posted manually, can be automated with AI.
Benefit-based pricing brings costs only from actualized automation
Snowfox’s DNA has the idea built into it, that the service must create more value than what the customer pays of it. Due to this our pricing is totally based on transaction costs of automated purchase invoices, that are significantly lower than manual labor.
The transaction costs materialize to the customer only if the AI has produced correct postings predictions. If the AI for some reason can’t predict the posting correctly, there are no costs to the customer. This way Palkeet only pays for delivered automation and no money is wasted on incorrect predictions.
“Our goal is to be the world’s best provider for financial administration AI services and Palkeet trusting our product is a great indication that we are on the right path.” – Ville-Veikko Laukkanen, CEO, Snowfox.AI
A real-time analytics service for the AI
During the deployment of Snowfox.AI, Palkeet will get credentials for a real-time analytics service, that enables monitoring the AI’s actions. The analytics shows the share of invoices that the AI has automated and how well the AI has been able to predict postings on them. The analytics also shows how confident the AI has been with each prediction and much more.
The analytics tool enables both Palkeet and Snowfox to see clearly the capabilities of the AI, as well as places for improvement. This is crucial, as the AI is constantly developed further, and the analytics service makes it possible to identify key areas for improvement.
“We’re excited to work towards purchase invoice automation for the government. We take much pride in carrying out significant development for the Finnish government.” – Miikka Savolainen, Snowfox.AI