AI has become the cornerstone of purchase invoice automation

In this blog, I will delve into our customers’ experiences with purchase invoice AI and how the role of AI has grown in recent years.

You have probably already heard that artificial intelligence (AI) is easily integrated into the purchase invoice processing workflow today. The easiest way to do this is to acquire a ready-made purchase invoice AI service and connect it to your existing purchase invoice system through interfaces. AI can accurately allocate expenses and determine the approval chain for invoices with incredible precision, significantly elevating the level of automation.

Why Integrate AI into Purchase Invoice Automation?

Without AI, at least 80% of purchase invoices are manually processed and circulated without any automation. It’s commonly accepted that the purchase invoice process is highly manual, and building automation through traditional methods is labor-intensive and time-consuming.

AI is an unparalleled automation solution because it provides an easy way to achieve a genuinely high level of purchase invoice automation. AI learns the logic of automatically processing invoices by leveraging historical data, eliminating the need to manually create a single automation rule. With AI, automation levels can instantly reach 80%, and with continuous learning, even surpass 90%.

Customer Stories

The City of Helsinki has recently adopted Snowfox’s purchase invoice AI to automate hundreds of thousands of purchase invoices. Before implementing AI, automation was minimal, and the vast majority of purchase invoices had to be manually processed in terms of allocation and circulation. We will soon publish a separate article about Helsinki’s experiences.

Colliers has been using Snowfox’s purchase invoice AI for some time now. Tiina Sollo, the Head of the Purchase Invoice Unit, mentioned in their article:

“We have explored the use of AI in financial services more broadly, and the processing of purchase invoices was identified as an excellent process to utilize AI. It has been great to see results just a couple of weeks after piloting. Expense allocation predictions are already streamlining our accounts payable work, and we expect forecast accuracy to increase as AI continues to learn.”

In a recent article, ISS recommends all organizations jump on the AI development bandwagon now. Chief Information Officer Jenni Heinisuo writes:

“Since my student days, over my twenty-year career, we have talked about the possibilities of AI and the forthcoming leap in development. Now those promises are a reality. Fortunately, companies are not alone in their development projects; there are an unprecedented number of practical AI solutions available for leveraging and further customizing to fit the industry and company’s needs.”

She also mentioned that ISS has incorporated AI to bring intelligence to their purchase invoice process.

“For example, we have integrated intelligence into the automated purchase invoice processing, so instead of grinding through the same process repeatedly, the system learns and improves.”

You can read more about Kiinteistötahkola’s experiences on Snowfox’s website, but here are a couple of quotes from Olli-Pekka Kuha’s interview:

“We process nearly half a million invoices annually through the purchase invoice process, so the volume is significant. However, there is a lot of repetition in invoices, so we knew from the start that with smart solutions, we could make things much smoother.”

“After reading about Snowfox’s solution, I realized that this pre-packaged algorithm-based AI service would fit our needs perfectly. A similar product had not appeared on the market before, but I immediately saw its potential.”

“In the very first month, the accuracy of invoices in the automation process exceeded 80%, and currently, we are operating slightly above 90%.”                      

Atlas Copco’s global service center has decided to implement Snowfox for their purchase invoice automation. Michal Schwarz shared her expectations:

“Processing invoices without physical handling is a continuous process for us, and we have struggled with non-purchase order invoices quite a bit – it’s interesting to see what Snowfox AI will bring to the automation of purchase invoices for our global service center.”

PHM Group abandoned the use of a software robot when implementing AI because AI could handle the same tasks as the robot and much more. Ida Bäckström commented on the improvement in the quality of work due to AI:

“Work is more meaningful now because we don’t have to waste time on details and routine copy-paste work. What has disappeared is the work that raised stress levels and heart rates during breaks.”

Implementing AI can be a concern for some companies, and it may even involve some fear. PHM Group’s IT Service Manager, Hannu Kaskilahti, mentioned the implementation as follows:

“This was by no means an IT project, but rather the development of accounts payable operations. IT does help, but IT leading the way may not always support the progress of the process in the best possible way. It was really good that this was a joint project.”

Sarastia, Finland’s largest provider of support services to the public sector, handles a vast number of purchase invoices, making automation crucial to their work. Katja Somppi has seen AI bring significant automation to the processing of purchase invoices.

The implemented AI handles approximately 90,000 purchase invoices per year. AI has made a significant leap in automation, resulting in smoother purchase invoice processing and time savings for us.

The accuracy of nearly all selected predictable accounting dimensions is well above 90%, even close to 100%. This is an excellent level of performance for AI, indicating that purchase invoice data is of high quality and AI models work exceptionally well.

Mehiläinen, one of Snowfox’s first customers, recognized early on that the processing of purchase invoices from financial processes is ideally suited for AI. It’s great to see that they had such forward-thinking individuals who dared to embrace AI almost from the outset. Roope Rauhanlinna saw that AI could be well-suited for automating the processing of purchase invoices.

“The generic process and the inclusion of large data sets and repetitions argue for taking the processing of purchase invoices as the first automation target for AI.”

In addition to invoice allocation, Snowfox’s AI service handles the circulation of purchase invoices for Mehiläinen.

Our vendor landscape is fragmented, and the level of content in invoices varies. Therefore, we were unable to achieve the desired level of automated routing in the past. Now, with Snowfox.AI, 90% of invoices are automatically routed. This forecast accuracy has exceeded our expectations and significantly reduced the workload in accounts payable.

How Many Purchase Invoices Should You Receive Annually to Benefit from AI?

A rule of thumb for AI usage is that the more non-purchase order invoices you receive, the greater the need for AI. Also, the more complex your environment, the better it is for AI. Our largest customers receive several hundred thousand purchase invoices and operate in multiple countries, but even smaller volumes are sufficient to realize the benefits of AI.

For AI implementation, it is advisable to have volumes of over 20,000 purchase invoices per year. Even half of this can be enough, but we have found that when there are more invoices and consequently more manual processing, there is a genuine desire to eliminate it.

If you want quick results and a truly significant automation leap, I recommend at least exploring the possibilities of AI and requesting a quote for the costs of AI and an assessment of its benefits in your process.