How AI is shaping AP automation
By Shannon Kreps, Vice President Product Marketing at Medius
The impact of the pandemic has driven many organisations to accelerate their digital strategies to transform business practices and remain viable. To mitigate fiscal risks, business leaders are increasingly focusing their attention on financial digital transformation, exploring technology such as machine learning and artificial intelligence (AI) to automate workflows, increase efficiency and improve cash flow visibility. Yet, when it comes to accounts payable (AP), finance professionals are still relying on manual processes to handle supplier invoices, with less than a third of companies employing AP automation and just 11% using AI. In an era where digital is synonymous with survival, this is leaving businesses open to unnecessary risk.
Behind the curve
For organisations who have not implemented AP automation solutions to handle tasks such as invoice data capture, matching and coding, AP clerks often find themselves spending valuable time on labour intensive work. Aside from entering important information into the system, resources are often wasted locating misplaced invoices or wading through heaps of paperwork. And that’s not all. According to the Accounts Payable Association, this is having a ripple effect on the entire organisation with 56% experiencing issues with cash flow forecasting and over three quarters of businesses admitting to paying their suppliers late.
With manual tasks causing many headaches for finance, it comes as no surprise that this is negatively influencing employee morale and productivity. Labour-intensive, repetitive tasks can make AP clerks more prone to mistakes, and common errors such as incorrect PO numbers, line items not matching POs, missing VAT numbers and invoices that have been addressed to the wrong department occur all too frequently.
Learning from mistakes
Deploying AP automation enables both small and large companies to digitise and streamline the process, reducing the amount of tedious, manual tasks carried out by employees and supporting organisational growth. Automation will help AP teams take ownership of the process, stay on top of invoices, and facilitate approvals when required. With processes and tools in place to control spend and eliminate errors such as duplicate payments, finance will have more oversight ensuring suppliers are paid on time. However, automation is just the first step towards AP efficiency.
With technology constantly evolving, going forwards there will be changes to the way AP operates and a solution that can accommodate AI and machine learning will be highly advantageous. Machine learning, for example, is ‘the science of getting computers to learn and act like humans do and improve their learning overtime in an autonomous fashion’. This is achieved by feeding machines data and information from observations and real-world interactions. Essentially, this means that the historic data entered into the system can then be used by the machine and learnt from, offering predictions and insights at a rapid rate. To avoid masses of information that is limited in value, organisations can manipulate the system by setting rules in order to receive relevant data outputs. AP software with machine learning capabilities will use a combination of data and rules to adapt, improving in efficiency as it learns what the user wants.
AI in practice
When it comes to applying AI and machine learning to accounts payable, there are two prominent areas where it can make a significant difference – invoice data capturing and coding invoices.
Invoice data capturing is the entry of the invoice details into the system. The early stages of the capture process involve using learning templates to obtain information, yet this only worked for structured documents and did not accommodate for changes to invoice layout. This later developed to extraction based on titles, logos, positions and formats, enabling for data to be captured from unstructured documents, with the automated tool capable of detecting moves and shifts to the invoice. Now, with the introduction of AI, everything is out of the box. Machine learning has the capabilities to read the invoice and make intelligent decisions based on data it has seen before. For example, if a new supplier is onboarded, it can autonomously process invoices with limited human intervention. This not only helps save on costs due to reductions in the manual workload, but it also increases turnaround times allowing employees to focus on more value-added work. Additionally, with an built-in validation system, the risks of invoice processing errors are minimised.
Automatic invoice coding is also a process ripe for AI. With invoices arriving in many different forms such as email, mail or fax, AP teams without an automated tool can spend countless hours on making sure invoice data is correctly coded; keying information into the accounting systems; and collecting approvals from heads of departments. While automation has helped streamline the process by using templates with default codes to match invoices, machine learning can go one step further by learning from historic data to figure out which code should be on a specific invoice. Using its learning capabilities, it can also notice relationships and patterns between different items similar to how a person in the system can make a connection. For example, it can detect and group items such as office suppliers, property purchase tax, audit fees etc. To provide reassurance, the AP team will still have the ability to accept or reject an invoice code. Again, as the system continues to learn from the changes made, it will adapt and increase in accuracy.
Just like any other automated process, there is still work to be done for AP to trust the data that is being managed within the system. The next step in AP is for machine learning to encourage more confidence in ‘touchless’ data capturing and invoice coding, making daily life easier, saving time and improving efficiencies for the organisation.