Artificial Intelligence in the Software Development Life Cycle: A Case Study in a Mid-Size Software Organization
Liljavirta, Juuso (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-2025061122406
https://urn.fi/URN:NBN:fi:amk-2025061122406
Tiivistelmä
This thesis examines the utilization of artificial intelligence (AI) within different phases of the software development life cycle (SDLC) in a mid-size IT consulting firm. The objective is to understand the current use of AI, identify future opportunities and highlight obstacles that prevent effective integration into SDLC processes, focusing specifically on the phases of requirements, development, testing and maintenance.
The research was based on existing academic literature that discussed AI applications, including machine learning, natural language processing and AI agents within software development. Qualitative data was collected through semi-structured interviews with nine employees and two external AI specialists, offering practical examples and experiences regarding AI’s role and potential.
Findings showed that AI was primarily utilized during the development phase, particularly through coding assistants such as GitHub Copilot and conversational tools like ChatGPT. Developers reported benefits in productivity, reduced routine workload, and improved capability for identifying coding issues and optimization opportunities. In contrast, its application remained limited in the requirements engineering, testing, and maintenance phases, mainly due to the complexity of specifying business logic, insufficient data for robust models, and concerns over accuracy and traceability.
The study identified several future opportunities for AI across the SDLC. These included automating documentation, generating detailed test cases, and improving communication among project stakeholders. Interviewees also suggested advanced applications such as automated refactoring, predictive maintenance, and intelligent traceability mechanisms that could streamline processes and reduce human error. Agent-based solutions and retrieval-augmented tools were also discussed as promising developments for the future.
The study also revealed multiple obstacles to broader adoption. These included data security and privacy concerns, lack of training and knowledge, technical imitations, and uncertainties about trust and responsibility. Organizational and process-related challenges were also noted.
This thesis concludes with general recommendations for software development organizations aiming to utilize AI more effectively. These include a focus on gradual experimentation, better knowledge management, and preparing for agent-based collaboration in the future.
The research was based on existing academic literature that discussed AI applications, including machine learning, natural language processing and AI agents within software development. Qualitative data was collected through semi-structured interviews with nine employees and two external AI specialists, offering practical examples and experiences regarding AI’s role and potential.
Findings showed that AI was primarily utilized during the development phase, particularly through coding assistants such as GitHub Copilot and conversational tools like ChatGPT. Developers reported benefits in productivity, reduced routine workload, and improved capability for identifying coding issues and optimization opportunities. In contrast, its application remained limited in the requirements engineering, testing, and maintenance phases, mainly due to the complexity of specifying business logic, insufficient data for robust models, and concerns over accuracy and traceability.
The study identified several future opportunities for AI across the SDLC. These included automating documentation, generating detailed test cases, and improving communication among project stakeholders. Interviewees also suggested advanced applications such as automated refactoring, predictive maintenance, and intelligent traceability mechanisms that could streamline processes and reduce human error. Agent-based solutions and retrieval-augmented tools were also discussed as promising developments for the future.
The study also revealed multiple obstacles to broader adoption. These included data security and privacy concerns, lack of training and knowledge, technical imitations, and uncertainties about trust and responsibility. Organizational and process-related challenges were also noted.
This thesis concludes with general recommendations for software development organizations aiming to utilize AI more effectively. These include a focus on gradual experimentation, better knowledge management, and preparing for agent-based collaboration in the future.