The automation of generative AI to enhance software testing
Soranummi, Masi; Pasanen, Roosa (2025)
2025
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2025121034299
https://urn.fi/URN:NBN:fi:amk-2025121034299
Tiivistelmä
Recent advancements of Generative AI technologies such as large language models (LLMs), have created new possibilities in automating and assisting with tasks that require heavy manual labour in software development and testing. The client of this thesis, ABB Oy, Distribution Solutions, was interested in improving their own software testing processes by using LLMs. The objective of this thesis was to automatize the use of generative AI to increase the efficiency of the calculation of coverage percentages in software testing.
The purpose of this thesis was to develop a tool, which consecutively prompts a generative AI-model based on user-provided prompts and materials. It explores the capabilities of large language models and uses those, along with the required steps to calculate test coverage, to create this tool. The tool aims to automate parts of this calculation with its two different modes of operation, both designed to assist in different tasks in the coverage percentage calculation process. One mode is designed to prompt a set of materials, which can be text-based files or entries in a JSON-array, along with instructions from the user’s prompt. The other is designed to compare two sets of materials against one another, intended to be used for processes such as mapping test cases to features.
Large language models were discovered to have potential uses in automating aspects of software testing, such as calculating test coverage. The resulting tool takes advantage of these discoveries and fulfils client needs by being able to automate parts of the calculation process with minimal manual labour. However, while the tool is functional, it will need further development in adding the client’s preferred model to reach full satisfaction of their needs.
The purpose of this thesis was to develop a tool, which consecutively prompts a generative AI-model based on user-provided prompts and materials. It explores the capabilities of large language models and uses those, along with the required steps to calculate test coverage, to create this tool. The tool aims to automate parts of this calculation with its two different modes of operation, both designed to assist in different tasks in the coverage percentage calculation process. One mode is designed to prompt a set of materials, which can be text-based files or entries in a JSON-array, along with instructions from the user’s prompt. The other is designed to compare two sets of materials against one another, intended to be used for processes such as mapping test cases to features.
Large language models were discovered to have potential uses in automating aspects of software testing, such as calculating test coverage. The resulting tool takes advantage of these discoveries and fulfils client needs by being able to automate parts of the calculation process with minimal manual labour. However, while the tool is functional, it will need further development in adding the client’s preferred model to reach full satisfaction of their needs.