Implementation of AI Automation for Ticket Classification
Automatic ticket classification is the first step toward an intelligent helpdesk. The system determines the type, subject, and priority of incoming tickets faster and more accurately than manual sorting.
Multi-dimensional Classification
Each ticket is classified simultaneously across multiple dimensions:
class TicketClassification(BaseModel):
# Topic
category: str # billing / technical / account / general
subcategory: str | None # first-level detail
# Priority
priority: Literal["P1","P2","P3","P4"]
urgency_indicators: list[str] # urgency signs from text
# Characteristics
sentiment: float # -1 to 1
customer_type: str # new / existing / churn_risk
language: str
# Action
recommended_team: str
auto_resolve_possible: bool # can be auto-closed
similar_tickets: list[str] # IDs of similar tickets with solutions
Training Sample from History
The best data source is historical tickets with labels marked by operators. Important: the quality of labeling in history is often low (operators rush). Before training — data cleaning: remove ambiguous examples, balance classes.
Volume for a good classifier: 500+ examples per class for BERT, 200+ for GPT few-shot.
Zero-shot Classification for New Categories
When a new category appears (new product, new request type) you don't need to retrain the model. GPT-4o with category description:
def classify_new_category(ticket: str, categories: list[CategoryDef]) -> Classification:
categories_text = "\n".join(
f"- {cat.name}: {cat.description}" for cat in categories
)
return llm.parse(f"Classify ticket by categories:\n{categories_text}\n\nTicket: {ticket}")
Auto-closure of Simple Tickets
Tickets like "Thank you!" after resolution, "Complaint withdrawal", system notifications — are automatically closed. Rule: auto_resolve_possible = True AND priority = P4 AND sentiment > 0 → close with auto-response.
Precision on auto-closure must be > 99%: incorrect closure of a real issue is unacceptable.