GEA (Graph Explorer Agent): A Business Automation Service That Supports Productivity Improvement and Faster Business Process Transformation Without Custom Development

Amid labor shortages and rising labor costs, organizations need to improve the efficiency of routine tasks. However, conventional robotic process automation (RPA) and similar approaches require time and cost for custom development for each process, while many routine tasks in business applications that do not support API integration remain manual.
GEA (Graph Explorer Agent), developed in-house by ABeam Consulting, uses AI agents to explore business applications and organize screen layouts, operating procedures, and the relationships among screen transitions as structured information called an application graph. GEA then uses this information to generate executable workflows, enabling the "automation of automation" without custom development and supporting continuous business process transformation and productivity improvement.

GEA uses AI agents to operate business applications in a browser and explore screens (nodes), input fields, and operations and transitions (edges) to build an "application graph." An application graph is reusable structured information that maps the overall structure of an application and its operational flows. GEA uses this information to generate executable workflows and automate routine tasks. Once an application graph has been built, workflows for the same application can be produced easily and deployed across a wide range of tasks. Because GEA works with browser-based applications, it requires no API integration, modification of business applications, or scripting for individual operations. This "automation of automation" enables rapid deployment without custom development. Application graphs and workflows can also be updated easily when screens or procedures change, reducing the maintenance burden.

AI agents explore the screens, input fields, operations, and branching transitions of business applications and structure them as an "application graph," with screens represented as nodes and operations and transitions as edges. GEA captures the as-is state of the entire application and automatically generates workflows based on business objectives. Conventional RPA and script development require requirements definition, scenario creation, and testing to be repeated manually for each process, taking several weeks to several months and requiring specialized personnel for implementation. With GEA, AI agents handle this development process, allowing personnel to focus on reviewing and adjusting the generated procedures. If screen layouts or business procedures change, the affected areas can be explored again to update the application graph and workflows, eliminating the need to rebuild them from scratch. This streamlines automation implementation and maintenance, enabling operational teams to realize benefits sooner.

GEA reproduces the actions people perform on browser-based screens, including data entry, selection, and button clicks. Even when APIs are not publicly available, or when modifying core systems, legacy systems, or SaaS platforms would involve time, cost, and risk, GEA can automate tasks without API integration or modifications to business applications. Traditionally, system integration has required additional development, modifications, and connector configuration, and waiting for the information systems department to respond has often created a bottleneck. GEA enables organizations to begin implementation quickly without changing existing assets or interrupting current operations. For example, routine data entry and inquiry tasks in business applications without publicly available APIs can be automated solely through screen operations. Eliminating the approval process and development time required for application modifications accelerates the path from planning to operation. Low-risk implementation without modifications and broad applicability across applications enable organizations to realize the benefits of automation without waiting for lengthy development projects.

GEA uses generative AI to explore business applications and develop workflows, but not to execute completed workflows. During operations, predefined workflows are executed mechanically, reliably reproducing the same procedures each time without variation between runs. There is also no delay from queries to generative AI, enabling fast and stable processing of high-volume, repetitive, and scheduled tasks. Because no generative AI usage fees are incurred for each execution, running costs are less likely to rise as processing volumes increase, making costs easier to forecast. Separating exploration and development from execution combines ease of implementation with reliability, speed, and low cost during operations. End users can continue to process repetitive tasks quickly and affordably with consistent quality.