Semantic workgroup identification pipeline using dense embeddings and vector similarity search to automatically route incoming support tickets. Unresolved tickets are encoded into 768-dimensional vectors via BAAI/bge-base-en-v1.5, then matched against a pre-indexed FAISS vector store containing 25,000+ validated workgroup labels from six months of ground truth data. FAISS k-NN cosine distance calculation produces confidence scores evaluated against a 0.75 threshold, triggering either automated routing to the correct workgroup or manual triage for human review. This architecture reduces manual ticket classification overhead while maintaining quality through confidence-gated fallback, enabling support teams to scale without proportional staffing increases.