RPA Debunked: Why Rules-Based Automation Isn’t Built To Last

Alexander Hagerup

Alexander Hagerup

Co-founder & CEO

November 9, 2022

5 min read

RPA Debunked: Why Rules-Based Automation Isn’t Built To Last

This article was originally published by Forbes, Oct 24, 2022, 10:00am EDT.


Rules-based automation—a broad set of technologies that use rules to perform high-volume, repeatable actions—can be found in all kinds of businesses today, freeing employees from monotonous tasks. The most sophisticated subset of rules-based automation is robotic process automation, or RPA, which uses software “robots” that automate tasks, such as extracting data or filling out forms.

The automation of highly repetitive tasks has resulted in significant labor efficiencies across industries. Indeed, RPA is still one of the best solutions for automating certain tasks.

However, while we should not discount its achievements, RPA has natural limitations, from its high cost to how easily it breaks down when pushed beyond its rigid rule set. In many use cases, RPA has been surpassed by the capabilities of true artificial intelligence (AI).

What does the future hold for RPA? It’s not going away anytime soon. However, as business leaders and technology innovators, we should only use RPA where the ROI is clear.

RPA Works—Until It Doesn’t

There is nothing inherently wrong with rules-based automation. I refer to RPA as a “horizontal” solution, meaning it can be “pretty good” at many things.

However, RPA is not the optimal choice for every “vertical” task. Once you get into the details and permutations of verticals, such as the accounts payable process, these rigid, rules-based solutions are mostly not up to the task.

RPA is riddled with unknowns. Before an RPA solution is implemented fully, it’s impossible to know how much it will cost, how much maintenance it will require, how often it will break down and what the ramifications of those breakdowns will be. An RPA solution may also require keeping a contractor on retainer to maintain the system. Taken together, these variables render an accurate prediction of ROI virtually impossible.

Stuck In A Rules-Based Rut

Despite the shortcomings inherent in RPA, many customers find it difficult to let it go. There are several reasons for this hesitancy.

First, RPA is a known and broadly used technology. There are countless solutions on the market—an entire ecosystem made up of well-known brands your partners are probably selling.

Also, once businesses have an RPA solution in place, they’re likely going to use it “horizontally” for multiple functions. Again, these solutions are expensive, and these companies might have also invested time and money in educating their employees to use them. When businesses are spending upward of several million dollars per year on RPA, it’s only natural that they would try to squeeze every last use out of it, whether those use cases are fit for RPA or not.

Shifting The Mindset

While some businesses looking for AI solutions have never implemented automation before, the majority have already used RPA in their back-office functions. These customers have hit a wall: They’ve squeezed out all the efficiencies they can from their RPA, and they’re starting to look for what’s next.

However, even future-looking customers can be skeptical about using new and emerging technologies such as AI. While many people are well-acquainted with AI solutions like home assistants in their personal lives, that comfort has not yet fully transferred to the professional sphere. There is also skepticism of AI among businesses that have been burned by solutions falsely selling themselves as being truly intelligent.

For those struggling to make the mental leap to AI, I have found that “seeing is believing.” To reduce the risk, companies should request proof of intelligence (learning capabilities) before they commit to purchasing an AI-powered technology. Once companies see true AI in action—and an accurate prediction of quick ROI—they understand and embrace the difference.

Looking specifically at invoice processing, the difference in using an AI-powered solution is that it can read any invoice with up to near-perfect accuracy, with no rules and no humans needed. In contrast, an RPA solution would require a human to program every type of invoice their company processes, so the rule knows where to look and what to do with the information it finds—if it can find it at all. As you can probably guess, the latter solution is not only still highly manual but also highly breakable.

Aim For The Verticals

Despite its flaws, RPA still has its place, for now. AI technology has progressed dramatically in recent years and has its sights on most, if not all, tasks handled by rules-based solutions today.

Business leaders should look at their verticals—specific tasks or business functions that AI could better handle—before falling back on less efficient, less effective horizontal RPA solutions. Where you can go vertical, you should.

And to my fellow tech visionaries: Let’s not allow RPA to overstay its welcome. Instead, let’s keep pushing the boundaries of true intelligence by developing AI solutions for all the challenges our business leaders face in the back office and beyond.

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?

Be a Finance Pioneer

Sign up to receive the latest news, updates and valuable accounting and finance resources conveniently in your inbox.