AI Agents—Battle of the Giants 

This topic is not new to me or my IT colleagues in the SAP sector. However, we were all surprised by the detailed description and analysis in a leading business magazine. What is this all about?

Robotic Process Automation (RPA) was highly promoted in the SAP community many years ago by Professor August-Wilhelm Scheer, among others. However, many of these concepts were too rigid and lacked agility. The idea was sound: many business processes in traditional organizational structures and workflows can be automated.

The challenge lies in drawing the line between organizational standards and permissible exceptions. Ultimately, RPA works best with a human supervisor. However, having a supervisor for monotonous, repetitive processes is not sustainable, so RPA never achieved widespread adoption.

Combining RPA with GenAI (generative AI), possibly paired with a company-specific LLM (large language model), could provide better services. See the research work of Aleph Alpha for more information.

Since we've been focusing on AI at SAP Basis, I've been sharing entertaining stories from the GenAI Hub of the SAP Business Technology Platform with my wife. She counters with a mild smile and tells me about the problems with our mopping and vacuuming robot. It has its own base station where it fetches fresh mopping water and deposits the dirt it has vacuumed up. The robot is equipped with LiDAR, a laser light sensor that helps it navigate our house. It can't climb stairs, but it also doesn't fall. The challenges lie elsewhere, my wife explained. For the robot to clean effectively, all moving objects must be removed: chairs from the dining table, trash cans from the chest of drawers, and shoes from the closet. Once the floor is largely free of unwanted objects, the cleaning process can begin. The robot can recognize these objects and move around them, but it can't move chairs to the side to clean under the table.

This means the success of our cleaning robot depends on preparatory work and a certain level of supervision. See also RPA. AI agents in the ERP environment work more agilely and prudently. This concept is well-described in The Economist's text, and now, every B2B software provider sells corresponding AI agents, so their functionality should be common knowledge. However, this is where the challenging thought experiment of the report, "AI agents are turning Salesforce and SAP into rivals," begins.

What happens if two AI agents meet? Who has priority? There are no general rules for this class of software; there are no rules of the road. This means the Salesforce AI agent could take on ERP tasks, while the SAP AI agent is dedicated to CRM. However, which system leads? The back office (ERP) or the front office (CRM)? The scenario from the Economist text could be expanded if the AI agents from SAP, Salesforce, Workday, UiPath, and ServiceNow met at the intersection of all business processes. Who will have priority?

In theory, GenAI with AI agents could become a battle of everyone against everyone else, like in the Wild West. The strongest will prevail, and the future ERP will be ruled by AI agents from Salesforce and Workday. Meanwhile, UiPath, ServiceNow, and SAP will provide the necessary support services. My IT department is experimenting and learning. Who will ultimately claim AI leadership is completely open and unpredictable at this point. You can read the Economist text here.