Navigating the Trade-offs of Generative AI in Consulting: Balancing Efficiency Gains with Epistemic Risks

Generative AI (GenAI) increasingly supports management consulting by accelerating preparatory and intermediate knowledge work such as summarization, drafting, structuring, and document analysis. However, faster production of fluent AI-shaped material does not automatically create defensible consulting output. This thesis examines how consultants integrate GenAI into everyday consulting practice to realize efficiency gains, and how they validate, safeguard, and professionally re-own AI-shaped material under conditions of epistemic risk and work-system pressure. Based on 25 semi-structured interviews with consultants across different seniority levels, firm types, and consulting contexts, and supported by two contextual governance documents, the study uses a Gioia-oriented inductive-abductive methodology. The analysis develops seven aggregate dimensions that inform a process-tension model of GenAI-assisted consulting work. The findings show that consultants use GenAI primarily in governed, task-bounded, and provisional ways. GenAI accelerates orientation, summarization, first-pass drafting, extraction, meeting support, translation, and routine communication. Yet AI-shaped material becomes professionally usable only when it passes through tool and data boundaries, remains task-bounded, is validated and source-grounded, is contextualized for the client problem, and is re-owned by consultants as defensible reasoning. Where these conditions cannot be met, material is reworked, caveated, escalated, withheld, or discarded. The thesis contributes to research on human-AI knowledge work and professional judgment by reframing responsible GenAI use in consulting as bounded acceleration under epistemic accountability rather than as adoption, automation, or productivity improvement alone.