Machine learning and invoice automation
One thing’s certain – technology will continue to shape how payments are demanded and collected in the years ahead. Paper invoices will become increasingly rare as more businesses move to digital processes. Bills will be generated immediately, with payments accelerated as vital functions are integrated into the invoice itself. Billing techniques will become more refined, customisable and intuitive for both business owner and customer.
But an even greater shift is beginning to take place. Machine learning allows invoicing systems to analyse and digitise invoices, extracting relevant information to speed up and automate payment processing. Plus, blockchains, the building blocks which power cryptocurrencies, are increasingly being put to use in invoice reconciliation, allowing for fast, transparent, secure processing of payments.
Machine learning software
Machine learning software is already available that scans keywords and numbers on incoming invoices, recognising company names, invoice numbers and money owed. Relevant information is automatically generated, removing the need for manual data entry; invoices are matched to purchase orders, and the individual giving final approval is informed and can confirm payment in a single click. One provider of this type of software, French firm AODocs, estimates that companies can save 15 minutes of employee time per invoice using this digitised approach, which would translate to a staggering 375 hours a month saved for a firm processing 1,500 invoices monthly.
Because these smart analysis systems extract structured data from sets of documents, the need for human intervention is almost entirely removed. Not only does this improve the speed, efficiency and accuracy of invoice management, the fact that captured information is automatically filed into databases makes business analytics much easier. It can also aid further software development in the future.
However, systems with machine learning capabilities often struggle with invoices that don’t have a set format. Accuracy can be high when scrutinising text and numeric data, but less reliable when trying to make sense of graphics, tables or visual features, such as logos. This is improving all the time as technology develops. But, a world where full automation of invoice processing is commonplace is a little way off yet.‹ View table of contents Next page ›