(function(d,u,ac){var s=d.createElement('script');s.type='text/javascript';s.src='https://a.omappapi.com/app/js/api.min.js';s.async=true;s.dataset.user=u;s.dataset.account=ac;d.getElementsByTagName('head')[0].appendChild(s);})(document,249849,266871);

Accounts Payable - How to show artificial intelligence who’s boss

Raphael Bres
Chief Product Officer at Tradeshift
This article was originally published as a guest post on Spendmatters.com

AI in Accounts Payable

From the fiction of ‘WOPR’ in the 1980s film "War Games"to the terrifyingly real almost-apocalypse of 1983’s Soviet nuclear false alarm incident, humanity has plenty of reasons to be fearful of the consequences of putting computers in charge. Perhaps that explains our wariness of ceding control to artificial intelligence (AI) in a host of much more run-of-the-mill processes.

But this reluctance to hand over the keys — of a self-driving car, a business or even missile silo — betrays a fundamental misunderstanding of how AI is designed to work. Too often AI is presented as it is in dystopian fiction, as an overlord, when it’s actually more like a workhorseand a highly effective one.

Change your mindset, forget about relinquishing complete control, and you’ll soon discover that AI can give you a competitive advantage in areas of your business you barely gave a moment’s thought to before, like the sometimes unglamorous world of invoice processing.

The cost of doing business

One might be forgiven for not getting overly excited about the intricacies of e-invoicing, but that’s precisely the point. Processing invoices is a time-consuming and repetitive task. It’s also notoriously prone to mistakes and human error that require manual review and reprocessing for reasons ranging from a mismatch between invoice and purchase order (PO), to errors with invoice coding. Accounts Payable (AP) teams currently spend around a quarter of their time chasing and correcting errors. Even the most dyed-in-the-wool data entry clerk will tell you there are more valuable things they could be doing with their time.

The bottlenecks these heavily manual processes create can also lead to problems further down the line. Experts estimate that it costs an average of $11 to and takes eight days to process every invoice. Not only is that a major cost-center for large businesses, it can become a major source of friction with suppliers when processing delays translate into late payments.

Many enterprises might see this as the unavoidable cost of doing business — others, however, might dream of automating the process entirely. Neither approach is right. Businesses need to find the perfect balance between human and machine, ensuring that each complements the other so automation doesn’t come at the expense of accuracy.

Keep control of artificial intelligence

It’s often said you should always aim to employ people smarter than yourself — but that doesn’t mean making them managing director on their first day. It’s the same with artificial intelligence. The technology is fiendishly clever, but it’s like someone with a raw talent: It needs to be nurtured before it can be let loose on something sensitive like invoice processing.

So start small and hold its hand while you teach the technology how to recognize common errors. A perfect “robotic” task for AI is performing line matching between invoice and purchase order, or entering the right coding. These are easy ones to teach because you’ll have plenty of “good” examples from which the AI can learn. For the first month or two, you’ll need a human operative sitting by and reviewing to ensure that the AI is getting things right. But after three months with zero errors, the technology can clearly be trusted to do that task.

Of course, all this is predicated on having zero tolerance for errors. But many businesses may prefer to trust their AI a little more and to use its initiative — for want of a better word — within carefully calibrated parameters. Unfortunately, almost every AI for accounts payable is a case of “on or off,” but that’s changing. The latest technologies enable operatives to set different tolerances, so you can decide whether the AI operates conservatively — always querying and flagging items it’s unsure of — or given relatively free rein, or somewhere in between.

An employee like no other

The key with AI is to remain in control. Treat the technology like any other worker: Give it clearly-defined tasks that are within its capabilities, and make sure that the consequences of any mistake aren’t catastrophic. Follow this simple approach, and you’ll start to unlock a huge range of benefits. Automation can lower the average cost to process an invoice by 80% and in some cases we’ve seen AI reduce manual interventions by as much as 99.5%.

But the benefits don’t stop there. Automation enables businesses to reimagine long-ingrained yet inefficient business processes. In fact, using context and clues gathered from historical data, it’s quite possible we could soon do away with the approvals process entirely on certain recurring or low-value transactions from a trusted vendor. It’s also a powerful weapon against fraud, able to analyze vast quantities of data from breaches around the world and identify instances that would normally not be found by the human eye, to avert them in the future. And it can have a direct impact on cash flow by helping to identify and reward the partners who pay early.

These are just some of the tactical benefits that properly harnessed AI can bring. Its greatest value, though, is to free up workers to focus on more valuable, skillful and enjoyable work. The lesson is clear: When it comes to AI, show it who’s boss. It’s the one employee with an insatiable appetite for whatever work and mass data you throw at it.

Learn more about Tradeshift

Get a demo now

Talk to us