Machine learning is a powerful type of artificial intelligence (AI) that relies on algorithms and patterns in data to optimize and improve outcomes.
Dooap is already applying this game-changing technology, for example, predicting correct coding and approval workflows - based on what the solution has learned from past invoice data.
What new applications can we expect soon? One of the more fascinating improvements relates to 3-way matching, but it does require us to rethink the concept of matching.
3-way matching is an old, long-used method in AP automation. It allows us to automate invoice approvals by comparing three different documents – the purchase order, the goods receipt note, and the invoice – to validate the accuracy of all three. Usually, all of them consist of lines, and we talk about line-level matching. In conventional matching, invoice lines are manually compared against their ordered and received counterparts. Any discrepancy would fail the matching and require a time-consuming investigation. There are no grey scales in this philosophy; for instance, the product numbers used by the buyer and the vendor are different, the lines do not match. Because the rules are stringent, this process leads to unnecessary exceptions and can be very frustrating.
Artificial Intelligence helps us look at this more mercifully, out-of-the-box. The ultimate goals of AP automation are to reduce manual work, increase accuracy and just plain and simple – save money. So, let’s look at how we could do matching in a more light and intelligent manner:
The conventional way involves manually adding stagnant rules that block the matching mechanically, which leads to cumbersome exception handling. With line-level optical character recognition (OCR) and AI, we can allow intelligence to recognize a problem in the line-level items and only pinpoint the real exceptions that require action. This will help us reach a higher level of touchless AP automation and concentrate our human focus on action items that have a higher financial significance than typical non-matched line items.
Example 1:
- Conventional: 3-way matching is off due to a price increase on one item. The process will pause until the discrepancies are investigated, found, and fixed.
- With AI: The price difference is detected and automatically corrected using digitalized invoice data and AI. After signing off on the price change, the invoice is routed out for payment.
Example 2:
- Conventional: The vendor adds an unknown amount of freight charges on each invoice. The 3-way matching cannot pass and will need manual processing.
- With AI: Freight charge is detected and automatically populated when it is within tolerance. The invoice for payment is automatically approved.
And yes, the system-assisted approach to matching may not be acceptable in all businesses. Still, we’d like to challenge the norm. Harnessing AI to perform the laborious work and allowing the user to make the final call can provide the best of both worlds while retaining total control of the process. Think of what has been going on in the insurance industry for more than ten years already. Insurance claims are handled electronically by utilizing AI for fraud detection and are no longer handled by humans. Instead, the system alerts if it detects a high enough possibility of a fraudulent claim. This development has taken place because it no longer made business sense to go through all claims manually. AI brings this same common-sense approach to matching in Accounts Payable.
Learn more about the benefits of machine learning for AP invoice processing in Microsoft Dynamics 365:
