Neil Robertson Mar 5, 2018 12:04:00 PM 9 min read

The implications of AI and ML in finance

AI and machine learning deliver the largest returns in areas where there is a high level of repetitive manual tasks, which makes the finance department the ideal profile for automation. To consider the implications of AI and machine learning on the accounts payable role, let’s start with the common process for paper invoices.

The accounts payable process always starts with opening the post and separating and collating the paper invoices. The next step is usually to check for the ever-growing number of emailed PDF invoices, print them out and add them to the pile to start accounting for them. The general ledger coding is then added, followed by separating all the invoices into piles after deciding who needs to approve it, which may include multiple approvers for higher value invoices. Best practice requires the maintenance of an “invoice register”, recording the invoice details of both who the invoice has been sent to and when – so you know where it should be and also for the month end accruals of unapproved invoices. This is followed by the physical delivering the invoices to the approvers.

Then there are both the internal and supplier queries to deal with. Either to chase the approval, raise a query on the invoice or manage suppliers chasing payment. As invoices get misplaced, there is the additional time to request duplicates from the supplier and the resulting risk of processing and paying the same invoice twice. Once the approved invoice is returned, the invoice register is updated, and the invoice is finally posted to the accounting software, enabling Finance to see the costs and cash requirement information within the accounting software.

The final task of the process is to then file the paper invoice for audit purposes, as very few people ever bother to try and find one after it is filed away. Then there is the task of the month-end accruals. Toting up all of the invoices listed in the invoice register that did not get approved in time for the cut-off date for the monthly accounts.

When looking at productivity gains, it is a common mistake to ignore all the work that is required to take the invoice through the entire process and only focus on the time taken to enter the finally approved invoice into the accounting software.

Productivity gains

The real productivity gains are in the removal of all the manual activities required to be able to post the invoice to the accounting software. The implication of AI and machine learning is that most of the above will simply disappear and for most businesses, it will save between 70% - 90% of the entire process. What follows is an explanation of the process when using automated invoice capture and approval software, using real-time integrations with your existing accounting software, whether it is cloud or locally deployed within your office.

The first step is to remove all the paper. You will have experienced the very rapid growth of suppliers sending their invoices via email as a PDF and by asking all your suppliers to use this methodology, you will be able to get up to 90% of all invoices being received this way. All email PDF invoices are automatically forwarded to be electronically processed, extracting all of the header and line information, validating the values to ensure 100% accuracy as well as undertaking anti-fraud steps such as validation of the supplier bank account details, VAT numbers etc. All paper invoices are simply collated and then scanned to deliver a PDF which is also emailed for processing. Every invoice received is captured and automatically updates the application invoice register to confirm it has been received. This information is updated after processing (usually in less than one hour), capturing all of the invoice information at both header and line level.

Using machine learning, the application enables the “defaulting” of the general ledger coding at invoice header level, but also at line level for those invoices that have multiple GL analysis codes. For most business, this will be between 60% and 80% of all invoices, depending on the granularity of the chart of accounts. The same is true for selecting and “defaulting” the approval process, as most suppliers invariably are approved by the same people, depending on the value of the invoice. Every invoice that can be defaulted is now received, digitised, coded and delivered directly to the first approver without any touch by the business. On final approval, it is posted directly to the accounting software (with a full audit trail) and there is no filing. Month end accruals are maintained in real time, also removing this task. This leaves the exceptions. These are the invoices that require the knowledge of the AP staff to correctly code them and select the appropriate approval process, plus those invoices that are placed under query.

Automation also removes most invoice queries from internal management, as well as the suppliers. Internal management can now see all their supplier information in real-time. How much they are spending and the exact status of every invoice within the approval process, plus all historic invoices, including whether or not they have been paid. If they do need to talk to Finance, everyone is looking at the same information. Account payable can also see the exact status of every invoice to be able to answer supplier credit control calls, including the ability to simply “nudge” the approver to keep the wheels turning.

The bottom line is that the productivity gains for almost every business will be between 70% - 90% over a manual process. What is often ignored is the value this also brings to the CEO or MD and the directors or senior management that purchased the goods and services from those suppliers.

For the CEO or MD, not only can they see the exact state of play for every supplier, they also have access to the real-time cash requirements report which includes every invoice from the moment it is processed, using the standard supplier payment terms in the accounting software to reflect the likely payment date. This can also include the estimated VAT payment, plus weekly and monthly PAYE/NI payments to deliver the full picture. The directors or senior management can now manage their suppliers, expedite approvals within minutes of their arrival from any device, including their smartphone, and more effectively manage (or delegate) invoice queries.

The challenge of disruptive technologies is the resistance to change as we all are guilty of maintaining the status quo because it is easy to do, usually quoting the phrase “if it isn’t broke, don’t fix it”. However, when the productivity gains reach a 70% - 90% threshold, delivering far more robust processes and far better management information to everyone that needs it, history shows that the change will take place. Invoice capture and approval automation is no different, except that when the CEO, MD and the board of directors find out what is possible, maintaining the status quo is extremely difficult to justify. When your accountants also recommend it, change becomes inevitable.

New call-to-action


Neil Robertson

Neil has a 39-year track record of building successful direct and channel global software businesses predominately in the financial software market place. Neil Robertson is Executive Chairman of Compleat Software. A 39-year veteran of the financial software marketplace, Neil has a long track record of building disruptive start-ups into successful businesses, including his time as CEO EMEA of Great Plains where he built the business outside of the USA from 1995 - 2001. Compleat is no exception and perhaps the most disruptive of them all.
Book a Demo
Contact Us
Become a Partner