accounting academy weekly ledger accounting academy weekly ledger

Human in the loop: How the Back Office Can Embrace AI and Empathy

Ralph Perdomo

Artificial intelligence, the simpler, catch-all term for machine and deep learning algorithms, is all around us.

From setting prices and reducing wait times for an Uber, to suggesting the next music act we’re sure to like.

AI even goes so far as being the perfect camera—one that replaces the shutter button with intelligence that knows precisely when to capture a moment in time.

While we may be eager—willing even—to let AI handle more of these types of tasks, there are still certain things only a human knows best.

robot getting approval from human
Created by Ryan Mason

It’s been challenging for AI to get a foothold in certain back office operations—chiefly accounts payable. Perhaps because these operations happen closer to the everyday flow of money and businesses aren’t entirely comfortable to cede this control to a nebulous algorithm.

After all, however intelligent an algorithm curating a playlist based off your music interests is entirely different from one curating invoices to pay suppliers. An error in the former can be resolved by clicking on to the next track. An error in the latter, however, may result in late or missed payments, penalty fees, or even a damaged supplier relationship.

But if a recent study conducted by Stanford computer scientists is any indication, the accuracy of AI (this study focused on the underlying AI tech powering speech recognition) is rapidly improving:

“...speech recognition is now about three times as fast, on average, as typing on a cell phone. The error rate, once 8.5%, has dropped to 4.9%. What’s striking is that this substantial improvement has come not over the past 10 years but just since the summer of 2016.”

Simply put, what the AI of today can do is much greater than what it could do a few short months back. But this reality hasn’t quite caught up with the back office, where a chasm still exists between AI’s capabilities and what it’s entrusted to do.

Here’s how back offices can slowly warm up to the idea of AI and make the (inevitable) adoption and proliferation of this technology much more palatable.

Human in the loop

Meet the human in the loop.

Human in the loop (or HITL) is often used to train artificial intelligent models. With this approach, a human can keep things (read, the artificial intelligence) from falling off the rails, providing supervision, and to train, tune, and test a particular algorithm.

But its usage isn’t only limited to training AI.

Take Clara Labs’ approach to solving the complicated task of automating scheduling between multiple parties and how HITL comes in—

“When the Clara AI has a high degree of confidence in its proposed response, it can send the email without bothering a human. But in all other cases, the AI sends the text in question to a [remote assistant].”

And rather than using those remote assistants to better train and begin to slowly replace themselves with more robust AI tech, Clara Labs’ quickly learned that sometimes a human’s touch is best. Or as Jason Laska, Research and Development for Clara’s machine learning program put it, “sometimes you really need a person to do it.”

Through the lens of an accounts payable department, adopting HITL practices (similar to Clara Labs’) might look a little like this:

Consider a smaller, mom and pop-type of supplier. Their average invoice amount may be a drop in the bucket compared to your revenue, and their inability to process a card payment makes paying by check slightly irksome, but the livelihood of this supplier may very well rest entirely on your organization. Based on its small size and payment method, an AI may de-prioritize this supplier—perhaps even waiting until the last day of payment term to send out the payment.

Any AP clerk could catch, flag, and expedite payment for this supplier; a kind of way of putting a face to the PO. What may be a seemingly trivial thing is just another way that technology, used correctly alongside caring people, can better the relationships between organization and suppliers—regardless of size.

Embracing the preface to human in the loop is how organizations can begin to acclimate itself with the future of AI and, all the while, gain an empathetic edge. It’s just another way how technology is empowering organizations in the marketplace.


Find out why payment automation is really about cutting costs and improving cash flow. [Watch the webinar]