Pivotal transformation: From assisted driving to fully autonomous finance

Mauricio Peña, Chief Safety Officer for Waymo, one of the most well-known companies developing technology for driverless vehicles, recently wrote an op-ed on CNBC about the difference between “autonomous” and “self-driving” cars. In particular, he discussed the disconnect between the promise of autonomous driving and the danger of mixing distinctly different levels of automation under the same or similar labels. This confusion can lead to dangerous situations, hamper progress and innovation, and sow fear and distrust. The key to pivotal transformation may be as simple as the words we use.

In the context of driving, there are vastly different capabilities and implications between these technologies, so using the correct terms is important. At one end of the spectrum, there are low-level autonomous systems that can handle parts of the driving task but still require the full attention of a licensed driver, such as cruise control. Peña argues that these should be called “advanced driver-assistance systems” (ADAS) to ensure drivers remain safely aware at the wheel. At the other end are technologies that can consistently operate a vehicle on their own with no driver at all. These are the true “self-driving” systems, none of which yet exist at the consumer level.

Experts in autonomous driving technology have defined a set of levels to distinguish between degrees of assistive technology, from fully manual to fully self-driving systems. As a company focused on bringing cutting-edge AI to finance teams worldwide, we have found there’s also a need to provide clarity around the different stages of assistive and autonomous invoice processing technologies. To that end, we developed the Autonomous Index for Finance.

This index breaks down the manual-to-automated transformation journey, allowing you to assess where you currently are on that journey and determine the next, appropriate step. In this way, our finance customers are able to plot a path to successful digital transformation.

Autonomous Index for Finance

At Level 0, AP teams have no automated technological assistance. It's all you, manually reading and entering invoices into your accounting or enterprise resource planning (ERP) system. You're directing, you're matching, and you're doing all the necessary checks and approvals. The AP system may offer options like defaulting payment terms, for example, by providing last year's accounting entries.

At Level 1, we start to see the introduction of automation tools. These can include optical character recognition (OCR) to extract values from images, robotic process automation (RPA) to enter the invoice information, supplier portals and networks, or electronic data interchange (EDI) and XML integration to digitize invoices and receipts.

Moving upward through Levels 2 and 3, these technologies provide increasing amounts of assistance. But finance operators must still review the invoice information to make sure it has been correctly processed, make changes when necessary, and monitor for timeliness and completeness. The challenge is in providing anywhere over 60% assistance, because there are so many exceptions and variations, and the automations are fundamentally rules-based. The number of new suppliers coming into your supply base every year, the number of changing formats, and so on make it impossible to automate every permutation, variation, and nuance. As a consequence, the remaining manual invoices present a real burden:

• High processing and storage costs, ranging from $12 - $33 per invoice
• Manual review or handholding through at least a few steps in the process
• Long cycle times leading to roughly one-third paid from 10 days to over 3 weeks late
• Duplicate waste of up to 3.5% of payments that should have never been issued 
• 5 years or more of heavy investment in electronic invoicing for a bare majority of supplier adoption

Level 4: Making the leap from assistance to autonomy

Level 4 is the highest level of autonomous technology currently available. By applying AI, the system is capable of autonomously extracting, reading, understanding and making process decisions. It can perform all the tasks, including monitoring and correcting itself. It can make decisions based on your AP policies in all aspects of invoice entry, matching, accounting approvals, and compliance. On top of that, the AI knows where it can and cannot make a decision, so it can proactively engage the AP operator only when needed, so that nothing falls through the cracks.

The practical difference between Levels 1, 2, and 3 assistance on the one hand, and Level 4 autonomy on the other, is quite dramatic. Let’s look again at the metrics above to see how moving from assisted to autonomous processing can make all the difference.

Processing cost

With assistive technology, manual invoices cost $12 to $33 each to process. Autonomous processing can drive that cost down to $2 to $3 each, on par with world-class invoice processing metrics for electronic invoicing. That’s an improvement of over 80%, and in some cases over 90%.

Manual review

With autonomous invoice processing, manual review becomes a thing of the past for the majority of invoices that required it before. Autonomous systems offer a guarantee that the invoice is processed correctly. They don’t pass it through unless they’re sure.

Cycle times

When invoices are processed manually, steps are sequential and depend on getting through every processor’s personal queue, which is why the time to process is so high. An autonomous system can scale as much as needed and work through every step almost instantly, so the whole process shrinks to under a day.

Duplicates

In an assistive-technology invoice processing world, auditing for duplicates before payment is both challenging and unwise because it further slows an already lumbering process. With autonomous invoice processing, auditing can be done automatically, in line with the rest of the process, without lengthening cycle time at all. What’s more, the same AI that understands an invoice can also understand expense receipts and reports, and identify cross-system duplicates that were never before detectable.

Transformation timeline 

Instead of slowly forcing change on reluctant suppliers, an autonomous invoice processing system uses AI to read, extract, and understand invoice data, and make the right coding, accounting, and policy decisions for you. It’s really smart when you first set it up, like hiring the best and the brightest finance team around. And like the best and the brightest, it’s always learning and improving. Within just a few months, and with hardly any effort on your part, you receive incredible results. 

Just as there are important considerations around autonomous driving, it’s important to remember the ultimate end goal of finance transformation — a Level 5, fully autonomous system that doesn’t require any intervention, ever — and acknowledge that the technology is not yet ready to deliver that. But you can be sure that the industry will keep pushing the envelope to get closer and closer to making that dream a reality. And we will do our part to make finance AI as safe and reliable as the best driving technologies available.

Get more information on how to supercharge your finance team, click here.