Ticket Whisperer
AI-driven ticket management system for ServiceNow that automates categorization, routing, and resolution
Status: completed · 2024-09-30
Overview
Ticket Whisperer slashes manual ticket-handling time by parsing incoming ServiceNow tickets with a fine-tuned Azure OpenAI model, classifying intent & urgency, then routing them via ServiceNow API – all running in a Kubernetes cluster with GitLab CI/CD and Terraform-managed infra.
Technologies
ServiceNow, Python, Machine Learning, REST APIs, Natural Language Processing
- Resolution Time
- -70%
- Classification Accuracy
- 95%+
Ticket Whisperer – AI-Powered Ticket Management
Built at Petco - I can't give too much away on this one unfortunately
Ticket Whisperer is a production-grade Python service that plugs directly into ServiceNow and uses a fine-tuned Azure OpenAI model to parse, classify and route IT tickets without human intervention.
Key Features
- Automatic Categorization & Routing – GPT-4-class model analyses ticket text, maps to department rules and dispatches via ServiceNow REST API.
- Real-time Performance Metrics – Prometheus & Grafana dashboards track throughput, accuracy and latency.
- Kubernetes Native – Helm charts for zero-downtime rolling upgrades; autoscaling based on queue depth.
- Infrastructure as Code – Terraform provisions Azure resources, networking and Key Vault secrets.
- CI/CD – GitLab pipeline builds, tests and deploys containers to all environments.
Architecture
graph TD
A[ServiceNow Webhook] -->|JSON Ticket| B[Ticket Whisperer API]
B --> C[Azure OpenAI Inference]
C --> D[Classification Engine]
D -->|PATCH Ticket| E[ServiceNow REST API]
B --> F[Prometheus Exporter]
F --> G[Grafana Dashboard]
Tech Stack
- Python (FastAPI, Pydantic)
- Azure OpenAI for LLM inference
- Kubernetes on AKS – HPA & VPA
- Terraform for IaC
- Prometheus / Grafana – monitoring
- GitLab CI/CD – build & deploy
Metrics (snapshot)
| Metric | Value |
|---|---|
| Mean Response Time | < 300 ms |
| Classification Accuracy | 95 %+ |
| Tickets auto-resolved | 70 % |