The IAron Agent project aims to revolutionize administrative tasks within WEG Energy's PPC department by automating complex processes within the SAP environment. Developed by a systems developer within the PPC team, IAron leverages a custom-built library that interacts with the SAP GUI, moving beyond traditional SAP Scripting by intelligently identifying and manipulating on-screen elements. This innovative approach allows for streamlined automation development and execution.
Trained using Google AI Studio's Gemini 1.5 Flash model, IAron boasts advanced capabilities beyond simple scripting. It can interpret and execute multi-step procedures, compose emails using WEG's standard template, analyze and summarize code from multiple files (including Python and VBA scripts within Excel spreadsheets), and even autonomously perform tasks based on the interpretation of internal documentation.
Before the development of IAron Agent, the PPC department faced several key challenges related to administrative tasks within the SAP environment:
Manual Data Extraction and Manipulation: Employees spent significant time manually extracting data from SAP reports and making mass changes to Production Orders and Project information. This was a tedious, time-consuming process prone to human error.
Cumbersome SAP Scripting: Automating tasks using traditional SAP Scripting proved complex and inflexible. Identifying and interacting with specific UI elements required extensive coding and was often broken by even minor changes to the SAP GUI.
Limited Automation Capabilities: Existing automation solutions were limited in scope and often required significant developer intervention. This created a bottleneck for automation requests and slowed down process improvements.
Lack of Process Standardization: Inconsistent procedures and documentation made it difficult to automate tasks effectively. Ambiguities and a lack of clarity hindered the development of robust and reliable automated solutions.
Dependence on IT: The PPC department relied heavily on IT resources for even minor automation requests, further delaying implementation and increasing the workload on the IT team.
These challenges resulted in reduced efficiency, increased risk of errors, and limited agility in adapting to changing business needs. The IAron Agent project was initiated to directly address these pain points and empower the PPC department to automate their processes more effectively.
For those who don't know, SAP Windows looks like:
The IAron Agent project tackled the challenges of SAP automation head-on through several key developments:
Custom SAP Interaction Library: A new library was created to simplify interaction with the SAP GUI. Instead of relying on fragile, element-ID-based scripting, this library intelligently searches and interacts with UI elements based on their type and properties (e.g., buttons, input fields, grids). This makes scripts more robust and resilient to changes in the SAP interface.
AI-Powered Automation: Leveraging Google AI Studio's Gemini 1.5 Flash model, an AI agent was trained to understand and utilize the custom SAP interaction library. This enables the AI to generate automation scripts, interpret and execute multi-step procedures, and even understand and summarize existing code.
Multi-Task and Prompt Decomposition: The AI agent possesses advanced prompt processing capabilities, allowing it to break down complex user requests into smaller, manageable sub-tasks. This enables the execution of sophisticated automation workflows with minimal user input.
Extended Capabilities: Beyond SAP automation, IAron can interact with other applications and file types. It can compose emails, analyze code from multiple files (including Python and VBA scripts within Excel), and perform tasks based on information extracted from internal documentation.
Structured Procedure Format (In Development): A standardized format for documenting procedures is currently being developed. This will enable the AI agent to autonomously interpret and execute complex tasks based on clearly defined steps, further reducing the need for manual intervention.
How it works:
A user interacts with IAron through natural language prompts. The AI agent interprets the request, breaks it down into sub-tasks, and uses the custom library to interact with the SAP system. It can extract data, manipulate information, execute transactions, and generate reports, all while following predefined procedures. The AI's ability to understand code and documentation further enhances its automation capabilities, allowing it to adapt to new tasks and processes quickly. The ongoing development of the structured procedure format will enable even greater autonomy and efficiency.
The IAron Agent's development followed a structured approach, incorporating several key stages and technologies:
pywinauto module to interact with the SAP GUI. Instead of relying on fragile element IDs, the library dynamically identifies and interacts with UI elements based on their properties (e.g., control type, name, position). This approach makes scripts more resilient to changes in the SAP interface and significantly reduces development time.Here's how SAP script automations are typically written:
Here's how SAP automations are written with the developed library:
AI Model Selection and Training: Google AI Studio's Gemini 1.5 Flash model was selected for its powerful generative capabilities. A comprehensive training dataset was created, consisting of numerous input-output examples demonstrating various SAP automation tasks and interactions with the custom library. This dataset was carefully curated to cover a wide range of functionalities, including data extraction, input manipulation, navigation, and report generation.
Prompt Engineering and Multi-tasking: Significant effort was dedicated to refining the AI's ability to understand and respond to natural language prompts. The model was trained to decompose complex requests into smaller, manageable sub-tasks, enabling it to execute sophisticated automation workflows with minimal user input. This involved careful design of the training data and iterative refinement of the prompting strategies.
Integration of Extended Capabilities: To expand IAron's functionality beyond SAP, additional capabilities were integrated. This included leveraging existing libraries for email generation, code analysis (Python and VBA), and file manipulation. The AI was trained to recognize and utilize these capabilities based on user prompts.
Structured Procedure Format Development (Ongoing): A key ongoing effort is the development of a standardized format for documenting procedures. This structured format, likely using a markup language like XML or JSON, will enable the AI to parse and execute complex procedures autonomously, further reducing the need for human intervention. This involves careful consideration of the types of procedures to be automated and the necessary information to be captured within the structured format.
Testing and Refinement: Throughout the development process, rigorous testing was conducted to ensure the reliability and accuracy of IAron. Feedback from testing was used to refine the training data, improve the custom library, and enhance the AI's ability to handle various scenarios and edge cases. This iterative process is crucial for ensuring the robustness and effectiveness of the final solution.
The IAron Agent project has yielded several promising results:
Increased Automation Efficiency: IAron significantly reduces the time and effort required to automate tasks within SAP. The custom library and AI-driven scripting simplify the development process and allow for faster implementation of automation solutions.
Improved Accuracy and Reduced Errors: By automating manual processes, IAron minimizes the risk of human error, leading to more accurate data entry and manipulation within SAP.
Enhanced Productivity: Automating repetitive tasks frees up PPC department employees to focus on more strategic activities, boosting overall productivity.
Greater Agility: IAron's ability to adapt to changing processes and interpret new procedures allows the PPC department to respond more quickly to evolving business needs.
Reduced Dependence on IT: By empowering PPC employees to develop and implement their own automation solutions, IAron reduces the burden on IT resources.
Improved Process Standardization: The development of a structured procedure format promotes greater consistency and clarity in documenting processes, facilitating more effective automation.
While still under development, early testing and implementation of IAron have demonstrated its potential to transform the way administrative tasks are handled within the PPC department. As the project progresses and the structured procedure format is refined, further improvements in efficiency, accuracy, and productivity are expected. Furthermore, the project's success can serve as a model for other departments within WEG Energia looking to leverage AI for process automation.
The IAron Agent project represents a significant step forward in leveraging AI to automate complex processes within the SAP environment. By combining a custom-built SAP interaction library with the power of a generative AI model, IAron empowers WEG Energy's PPC department to streamline administrative tasks, reduce errors, and improve overall efficiency. While still in its developmental stages, the project has already demonstrated promising results and highlights the potential of AI to transform business processes. The ongoing development of a structured procedure format will further enhance IAron's autonomy and broaden its applicability. Looking ahead, the IAron Agent project not only promises to revolutionize internal operations within the PPC department but also serves as a valuable proof-of-concept for wider AI adoption within WEG Energy, paving the way for a future where intelligent automation plays a central role in driving efficiency and innovation.
This project supports multiple languages, such as Portuguese, English, French, and German.
To learn more about the IAron Agent and connect with the developer, please visit Robert Aron Zimmermann's LinkedIn profile: https://www.linkedin.com/in/robert-aron-zimmermann-1806ba272/
There are no datasets linked
There are no models linked
There are no datasets linked
There are no models linked