How AP Automation uses AI
When you submit an AP supplier invoice by email or upload,
- Correctly identify the supplier.
- Extract details such as names, dates, addresses, and line items.
- Code dimension details, such as the GL account, location, and department.
- Create a draft transaction using this information, that's ready for you to review and post.
- Detect and flag duplicates, so you don't overpay.
To recognize supplier documents, know how to extract information from them, and make improvements to AI predictions, AI uses machine learning (ML) technology.
How ML works
ML requires a lot of data to learn how to process incoming AP supplier invoices accurately. This data is used to build a model that helps AI know how to do the following:
- Recognize a wide variety of AP supplier invoice layouts.
- Understand the different ways in which names, addresses, dates, and line items can be presented.
- Match AP supplier invoices to the right supplier records.
- Code GL accounts and dimensions for transactions from a particular supplier.
ML relies on collective activity on the Sage Network to learn how to read AP supplier invoices in the most efficient way. Rather than learn only from the hundreds or thousands of AP supplier invoices submitted to your company, ML has the opportunity to learn from millions of AP supplier invoices submitted to many companies, with all privacy and safety laws applied on the collective data. By accessing this collective data,
In addition to collective activity, ML also learns from the changes and corrections that are unique to your company. Each time you review a draft transaction and post corrections, this information is sent back to the Sage Network to update the ML model. Your company-specific data helps
How soon will I see improvements?
The more you use
The Sage Network is updated with your changes and corrections every 24 hours. This means that when you correct a supplier match, this information is available to AI on the following day. If you upload more documents from the supplier before the network is updated, AI will not know about your correction when it creates the draft transactions.
ML also looks for consistent changes to make sure that it's not learning from one-off changes that you only want to apply to a single transaction. When you make the same correction multiple times, ML knows to add this information to the ML model and apply it to future transactions. This means you might need to make some corrections several times before AI applies the information to new transactions.
Your data privacy
All customer data shared with the Sage Network is done so while maintaining safety and privacy. Your personal details and specifics around supplier data are never shared with another customer. The data is stored securely in the United States and processed in the United Kingdom and European Union.