Technology Blogs by SAP
Learn how to extend and personalize SAP applications. Follow the SAP technology blog for insights into SAP BTP, ABAP, SAP Analytics Cloud, SAP HANA, and more.
cancel
Showing results for 
Search instead for 
Did you mean: 
Priti_PM
Product and Topic Expert
Product and Topic Expert
463

Action AgentAction AgentWith the rise of AI and machine learning, one of the most exciting developments in this space is the emergence of Large Action Models (LAMs). Remember this video of Satya Nadela with Bloomberg where he compared iPhone’s vision by Steve Jobs with that of Rabbit’s R1 device which essentially utilizes LAM to do tasks for you e.g.; book a flight or narrate direction. Well, maybe Rabbit didn’t go big with R1 yet, but Action Models got a hype that it deserves.

Since then, there have been many movements around this space, recently Adept.ai - a San Francisco based startup that had received $350 Million funding to build “agents” which automate enterprise workflows; its executives were hired by Amazon along with tech licensing so that Amazon can boost up its AI portfolio.

We think, these models and agents together represent a significant leap forward in the way we approach process automation, particularly within the context of SAP and enterprise systems.

Now it’s time for Action!

What is a Large Action Model (LAM)?

A Large Action Model (LAM) is a sophisticated machine learning model designed to understand, predict, and automate complex tasks across various domains. Unlike traditional AI models that focus on specific tasks (e.g., natural language processing, image classification, sentiment analysis), LAMs are designed to predict sequences of (complex) actions that can be executed in a real-world system with minimal human intervention.

LAMs are trained on vast amounts of data, data in this case is not numbers, texts, but it encompasses diverse workflows and operational scenarios. This allows these models to generalize across different processes, making them highly adaptable to various enterprise applications.

What is an AI Agent?

AI Agents autonomously execute tasks by using tools and managing resources, adapting their actions based on the context. With built-in memory, they retain knowledge from past interactions to improve future performance and decision-making.

The Large Action Model provides a framework for an autonomous agent to select and execute actions based on its current state and the desired outcome. Thus, LAM functions like a brain for an agent in the planning phase.

PritiSAC_0-1725654596781.png

 

The Role of LAMs and Agents in SAP

SAP systems are the backbone of many large enterprises, managing everything from supply chain operations to human resources and financials. However, as you know best, ERP systems often involve complex, multi-step processes that require manual intervention, even with existing automation tools.

This is where LAMs powered Agents come into play. By utilizing these agents into SAP environments and processes, we believe enterprises can achieve a higher level of process automation and gain efficiency like never before. Here’s how:

  1. End-to-End Process Automation: LAMs can automate entire business processes that span multiple SAP modules. For example, in a procure-to-pay process, a LAM can handle everything from purchase requisition creation to invoice processing, reducing the need for human oversight.PritiSAC_1-1725654596784.png

     

  2. Adaptive Workflows: Traditional automation tools require predefined rules and conditions (e.g., RPAs). LAMs, on the other hand, can learn and adapt to changing business environments. They can modify workflows on the fly, ensuring that SAP processes remain efficient even as business needs evolve.
  3. Enhanced User Experience: With the integration of LAMs, the user experience within SAP systems can be significantly improved. LAMs can provide real-time assistance to users, guiding them through complex processes, or even taking over routine tasks entirely, allowing employees to focus on higher-value activities.

And in future, when SAP data is exposed to LAMs; then agents can not only execute process steps but could also help with  Predictive Analytics and Decision-Making: Here LAMs can predict potential issues or forecast values within SAP workflows before their occurrence. For instance, by analysing historical data, a LAM can predict number of days the order will be delayed or forecasts the price in the order without user needing to fill in.

Some examples in SAP where LAM’s can bring high value

  1. Intelligent Financial Close:
  • Financial closing is a critical process for any enterprise, often involving multiple SAP modules and significant manual effort. LAMs can automate the entire closing process, from reconciling accounts to generating financial statements, ensuring accuracy and speed.
  1. Supply Chain Optimization:
  • In supply chain management, LAMs can be used to automate demand forecasting, inventory management, and supplier coordination.  
  1. Human Capital Management (HCM):
  • LAMs can transform how HR processes are managed within SAP systems. For example, they can automate employee onboarding, performance reviews, and even handle complex payroll calculations, ensuring compliance with local regulations.
  1. Customer Relationship Management (CRM):
  • In the CRM space, LAMs can automate customer interactions, from initial contact to post-sale support. They can analyze customer data to provide personalized recommendations, automate follow-ups, and even predict customer churn, allowing for proactive engagement.

Conclusion

Large Action Models powered Agents represent the next frontier in process automation for SAP and the broader enterprise world. By leveraging this, businesses can achieve unprecedented levels of efficiency, accuracy, and agility in their operations. As we move forward, we are looking for input and feedback so that we can shape this innovation in context of SAP data and processes.

Leave your comments to share your thoughts!

  • We need your input on use-cases, workflows and processes in SAP that you think will benefit the most!
  • For usability - Would you see automation of process steps happening in background or in front so that you can review it?
  • Some brainstorming on what marks the successful execution of process execution by AI?
  • What happens in case of exception and edge cases when some specific condition deviates the current run of execution from the typical flow of process steps which LAM is not aware of? Say for example a threshold limit triggers a custom-built popup on UI that the model has not seen before.
  • How much input do you think is sufficient to give upfront that LAM is able to fulfil E2E execution of process flow without being stuck and waiting for user input. For example, mandatory fields in the UI that can not be pre-determined via a logic but can only be provided by end user.

So, are you ready to explore how SAP’s Large Action Models can revolutionize your SAP processes?  

Take ActionTake Action