Agent-based chatbot in corporate environment

Abstract

The rapidly increasing interest in Artificial Intelligence (AI) has spurred major corporations to continuously innovate, particularly in the development of autonomous agents aimed at generating subscription-based revenue. The report explores the development and results of an autonomous chatbot specifically designed for Laerdal. This solution leverages transformer architectures and customized tools to address tasks within Laerdal, streamlining operations and enhancing communication efficiency among employees.

Extensive literature on existing Large Language Models (LLMs) and agent solutions was reviewed to support the development of a chatbot specifically tailored for Laerdal. This chatbot leverages internal data to address Laerdal-specific queries. The system operates in two versions: one utilizing the Language Agent Tree Search (LATS) architecture for complex query handling, and another streamlined version focusing solely on tool utilization without the LATS framework.

This self-reflective solution handles advanced user prompts and can significantly boost efficiency at Laerdal by performing tasks rapidly and cost-effectively. Features include fast retrieval of employee information, updated daily, to facilitate global internal communication and collaboration. Additionally, it can understand and generate images with live data from the internet using various tools in combination.

Overall the chatbot can serve as an internal tool designed to enhance productivity across Laerdal, demonstrating the practical application of autonomous agents in improving operational efficiency within a corporate setting.