Automatically classifying, prioritising and routing incoming support tickets is a genuinely useful project — it has a clear success metric (faster resolution, correct routing) and forces real engineering around classification and evaluation.
The pipeline
- Classify the ticket by category (billing, technical, general) and urgency, using a prompt or a small fine-tuned classifier.
- Route to the appropriate team or queue based on the classification.
- Draft a suggested first response for the agent to review and send, rather than sending automatically — keep a human in the loop for anything customer-facing.
How to evaluate it honestly
Compare classification accuracy against how tickets were actually routed historically (if that data exists), and track how often agents edit or reject the suggested draft response — a high rejection rate signals the drafting step needs work, not just the classification step.
See fine-tuning for classification vs generation.
— Pranjul Rathour, GenAI Engineer from Kanpur, India. Open to GenAI roles, hackathon judging, mentorship sessions and guest talks at any campus: pranjulrathour41@gmail.com.