| Keywords [eng] |
LLM-based agents, agentic AI, enterprise AI adoption, AI scaling, platform ecosystems, hyperscalers, business model, deeptech |
| Abstract [eng] |
This Master’s thesis investigates large language model (LLM)-based agent frameworks, which, despite rapid technical advancement, struggle to scale beyond pilot deployments in enterprise environments. While agentic systems demonstrate growing functional maturity, their adoption remains constrained by organizational, governance, and integration challenges. The study examines how agent frameworks are positioned within emerging platform ecosystems and identifies the factors that most strongly limit enterprise adoption and scaling. The research focuses on adoption constraints rather than technological performance, addressing the conditions under which agentic systems transition from proof-of-concept implementations to sustained organizational use. The research adopts a qualitative, exploratory design combining a systematic literature review with expert interviews involving creators, integrators, and adopters. Reflexive thematic analysis is used to integrate technical and organizational perspectives. The findings show that enterprise adoption is shaped less by agent capability than by data readiness, governance requirements, human oversight needs, and unclear value realization. Rather than progressing toward full autonomy, agentic systems are most adopted in configurations of supervised autonomy corresponding to Level 2 AI adoption. This dynamic is conceptualized as a maturity-adoption bottleneck, where organizational constraints limit the translation of technical capabilities into scalable deployment. |
