This thesis explores how risk assurance consultants in a Big Four professional services firm adopt and experience large language models (LLMs) in their daily work. Using a qualitative approach with semi-structured interviews, this study identifies the technological, organizational, and environmental factors that drive or hinder adoption. Additionally, this study investigates the consequences of GenAI on work practices and experiences of consultants. Findings predominantly reveal that consultants embrace LLMs mainly for efficiency, ideation, and drafting tasks, but remain quite cautious with client-facing or high-stakes work due to concerns about data security, quality, and accountability. Adoption is shaped by peer influence, leadership advocacy, and ease of use of the technology but constrained by transparency gaps and unclear governance. Governance ambiguity often leads to rejection of GenAI tools, which was particularly found to be the case in high stakes work where mistakes carry significant consequences both for the professional and the firm.
The research shows that LLMs are transforming consulting workflows from “creator to curator” ways of working; consultants increasingly start with AI-generated drafts and focus on refinement and contextualization. While productivity and output quality improve, workload remains high as cognitive effort of consultants and managers appear to shift to reviewing and verifying outputs. This redefines professional identity and expertise, emphasizing judgment, prompting skill, and ethical evaluation. Ultimately, the study concludes that generative AI is not replacing consultants but augmenting their capabilities. Successful integration requires trust, training, and responsible governance, marking the beginning of a new phase in knowledge- intensive work where human insight and AI collaboration coexist.