Automation Testing Framework for Reliable Autonomous Agentic AI
Year: 2025
Type: conference-paper
Abstract: With the rise of autonomous systems (AS) and agentic artificial intelligence (AI), a heightened automation of testing processes is required to build, deploy, or repair reliable intelligent systems. For this purpose, we developed a framework allowing any intelligent tester, whatever human or agent, to automate test scripts on their developed applications and provide detailed results of the users’ automated test cases execution. This framework enables testers to create and replay automated test scripts quickly, so automation can be used in short development projects by unskilled (manual or agentic) testers. Furthermore, the implementation of this framework aims to solve the problem of performing secure regression testing on applications before each new application release candidate and every time an intelligent developer - human or artificial - makes changes to the programmes code, which can take up huge amounts of testing time in a software development project and over the life cycle of the product application. Thus, this automated testing framework reduces the time to carry out testing so the costs, while it increases quality testing and system reliability. The first prototype of this framework was developed successfully and tested thoroughly, and its deployment in real-world context showed promising results, paving the way for reliable agentic AI systems.
Authors: Grant Martin, Joanna Olszewska
Institution: University of the West of Scotland
Language: English
Subfield: Software
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4413755230