Evaluate Weigh the pros and cons of technologies, products and projects you are considering.

What are the best uses for RPA and AI in ERP?

Listen to this podcast

IFS VP Bob De Caux discusses the push for intelligent process automation in ERP, the scalability limitations of RPA and how to tell software-based intelligence from plain old IT.

AI and robotic process automation are taking over more and more tasks from ERP users. RPA mimics human behavior, recording users as they enter data, execute commands and move documents across applications.

The machine-learning AI in ERP scans information for patterns, "learns" what to expect, makes decisions and even tries to predict the future.

Both technologies are enabling a new kind of ERP that is more automated, responsive and easier to use than its often frustratingly obstinate ancestors.

But just how intelligent, really, is the AI in ERP? Where does robotic process automation (RPA) belong? And where do the two intersect?

Bob De Caux ponders such issues in his role as vice president of AI and RPA at IFS, the Sweden-based ERP vendor. He joined me and Brian McKenna, business applications editor at ComputerWeekly, for a wide-ranging discussion about the respective roles of RPA and AI in ERP and where the technology is headed.

Defining intelligence

The use of AI in ERP has taken off in the past five years, and with it, the frequency of AI buzzwords being applied to features that owe more to traditional IT with its binary logic and decision trees.

It would help to define the terms -- intelligence for a start.

"If we think of hard-coded rules, they're set in stone, they don't change, they don't adapt," De Caux said. Instead, to be considered intelligent, software must be able to "adapt and learn, and to find deep complex patterns in data itself without them needing to be hard coded."

Bob De Caux, VP of AI and RPA at IFSBob De Caux

Intelligence also manifests in how software represents knowledge, similar to the way the brain works. "It could be more complex hierarchies, it could be relationships with different strengths," he said.

It doesn't help that vendors use the term AI too loosely in fields such as healthcare, when what's really providing many of the advances is conventional data analytics and IT. McKenna discussed the issue in a recent blog post on AI washing.

De Caux agreed that overinflating the term isn't helpful. "It's raised expectations of what AI can deliver, and certainly what it can deliver without a good amount of work to set it up."

'Robotic' is right in the name

RPA, in contrast, often uses AI but isn't itself a type of AI. "It's effectively trying to replicate how a human interacts with a system, in terms of the clicks they're making and the processes they're following." Doing so requires solving a lot of complex problems that employ AI in things like computer vision, understanding the buttons that appear on a screen and handling uncertainty about users' intentions.

"There are lots of uses for AI within RPA, but is it solving a problem that artificial intelligence should really be applied to? Not really," he said. "It's really trying to automate a workflow," which can often be done more effectively with system integration tools such as APIs.

RPA is a good way of replicating what a human does, but it doesn't scale well or adapt to change and can end up replicating an inefficient system. Often, it makes more sense to use machine learning to automate and optimize the workflow.

De Caux said many of the initial applications of artificial intelligence in the enterprise are focused on getting value from data, a theme he expounded on in a post on the IFS blog. But it can't just be any data. The data set must be large enough to train intelligent algorithms and exhibit patterns for the algorithms to identify, he wrote.

The next stage will be intelligent process automation that solves more complex problems than the automation of repetitive, time-consuming tasks that is the focus of the current generation of RPA and AI in ERP. De Caux likens this emerging intelligent automating of business processes in the system of record -- ERP -- to the intelligence in industrial automation.

De Caux also explained:

  • the tasks AI is best suited for;
  • where it fits in major ERP functions such as human capital management and CRM;
  • the importance of "explainable" AI; and
  • what's next from IFS in employing AI in ERP.

To hear the discussion, click on the podcast link above.

Data Management
Business Analytics
Content Management