In order to improve knowledge transfer, increase consistency, and decrease onboarding effort, this thesis explores the ways in which Large Language Models (LLMs) can be effectively included into the onboarding procedure for new employees at a telecom contact center. The study is organized into three stages: first, an analysis of the present onboarding procedure; second, the creation of a conceptual framework for the integration of LLM; and finally, the execution of realistic test scenarios to assess the suggested solution.

The awareness phase revealed many important issues with the current onboarding strategy, including a heavy dependence on human mentors, unequal knowledge transfer, and trouble expanding training initiatives. Technical system introduction and product knowledge training were two of the six onboarding elements that were the emphasis of the focused idea for LLM-supported onboarding that was proposed during the recommendation phase.

A Junior system engineer was put through two realistic onboarding situations using the Empolis knowledge platform. The findings show that LLM-assisted access to internal documentation promoted just-in-time learning, boosted user confidence, and allowed autonomous job completion. With complete task completion and maximum confidence scores, participants gave their onboarding experience high ratings in every category. The thesis comes to the conclusion that LLMs have a great deal of promise to improve onboarding efficiency without taking the place of human mentorship when used with organized, explainable frameworks like Retrieval-Augmented Generation (RAG). By operationalizing LLMs in a real-world onboarding scenario, this study advances applied AI research and provides a scalable model for knowledge-intensive settings. Future study ideas include cross-industry validation, comparative performance analysis, and more extensive testing across onboarding stages.