Home > AI Solutions > Artificial Intelligence > White Papers > Rethinking Hierarchical Text Classification: Insights from Multi-Agent Experiments with Small Language Models > Document purpose
This paper presents a Proof-of-Concept (POC) demonstrating how SLMs can substantially improve support case classification. The system the paper describes aims to:
Through multiple experiments and prototype development, the team explored several agentic workflows, addressing some limitations of traditional classification methods and exploring new possibilities for adaptive and ICC systems. Despite the hype surrounding these technologies, the outcomes were not entirely positive. This paper focuses primarily on sharing lessons learned from the experiments, including latency challenges with SLMs and consistency issues in complex business use-cases. The paper provides detailed descriptions of the experimental workflows and proposes an agent-based architecture to tackle this classification challenge, providing insights into the practical application of these advanced NLP techniques in real-world support scenarios.