When should we start using AI to automate the purchase invoice handling process?

The question in the title is often asked in customer meetings; when should we start using AI to automate our purchase invoice handling process?

Let’s say you’ve just implemented a new purchase to pay solution. When the changing your P2P system, the previous automation rules have been removed and you will start developing automation rules from the beginning. Your goal with the new system is to make the purchase invoice handling process as automatic as possible.

From experience, you know that building automation rules can take a long time if there are a lot of invoices to take care of. Especially when you have a thousand other tasks that require your attention too. As a result, rule-making can be delayed and the automation potential of the system is not reached as soon as you would have hoped.

Hence, your purchase invoice process remains very manual, possibly for a long time, meaning that purchase invoices are largely both circulated and posted manually. The price of one manually processed purchase invoice is at least € 8 and even over ten euros, depending on how efficient the process in use is.

“Invoice processing becomes very expensive if business managers or executives spend their precious time posting for invoices.”

What to do in a situation where building rule-based automation takes too much time?

Another modern way to automate the posting and circulation of purchase invoices is to utilize artificial intelligence in the process. Artificial intelligence is an easy way to automate as its implementation does not require manual rule building. Artificial intelligence creates automation rules itself by utilizing purchase invoice history and learns how to post and route your invoices automatically.

In addition, (at least) in the case of Snowfox.AI, the pricing of artificial intelligence is based on correctly posted and routed invoices. In practice, this means that we only charge if AI manages to automate the posting and routing of the invoice which would otherwise be posted and routed manually.

“The cost of a manually processed purchase invoice is at least eight euros. A purchase invoice processed using artificial intelligence costs tens of cents.”

Does artificial intelligence compete against rule-based automation?

The answer is simple: there is no competition between rule-based and artificial intelligence automation. Artificial intelligence and the system’s own automation work well together. Both technologies have the same goal to automate the manual work.

Why is there no competition? The logic works so that the system’s own automation always dominates. This means that when a purchase invoice arrives in the system, the system first checks if there is an automation rule for the invoice, and if not, the invoice ends up being automated by the AI. If an automation rule is found in the system, then the invoice will not be processed by ertificial intelligence at all.

In this way, the use of artificial intelligence is scaled as needed, depending on how well the system’s own automation works and is utilized. If the system’s rule based automation have reached the 30% automation level, the AI ​​will handle the remaining 70%. When the automation of the system reaches the level of 40%, 60% of the invoices, the remaining ones are processed by artificial intelligence.

So at what point is it beneficial to include artificial intelligence as part of the automation solution?

Let’s start with the fact that the most expensive option is always to handle purchase invoices manually. The cheapest option, on the other hand, would be the 100% level of automation achieved with the purchase invoice system. However, this is not realistic even on paper. There are several reasons for this, but for Non-PO purchase invoices, the very typical level of automation is from 10% to 30%. In summary, up to 90% of Non-PO purchase invoices are being processed manually.

The most realistic and sensible alternative is the automation achieved through the seamless cooperation of the purchase invoice system’s own automation and artificial intelligence. In practice, the system’s own automation is utilized as far as possible and the remaining invoices are directed to artificial intelligence.

However, as mentioned, building automation in a system takes time. On the other hand, artificial intelligence can be deployed very quickly, even in a couple of weeks, and it will produce value from day one. Therefore, artificial intelligence should be implemented as early as possible. This allows for a high level of automation quickly and a reduction in the cost of manual processing of purchase invoices.

Once artificial intelligence is implemented, it is beneficial to start taking advantage of the system’s own automation and make rules for frequent recurring purchase invoices. In this way, artificial intelligence can handle difficult purchase invoices and the system’s own automation can take care of recurring invoices.

As the system’s own automation increases, the need for artificial intelligence decreases. Each purchase invoice automated by the system reduces the cost of artificial intelligence and makes the process even more efficient.

However, you should not wait until you get the most out of the system and only then think about using artificial intelligence. In our experience, the potential of artificial intelligence should be unleashed as soon as possible to support the fast increase of the level of the automation.