Automating Resolution of IT Support with Chatbots
summary
Freshservice’s Virtual Agent is an ITSM chatbot integrated with Slack and Teams that classifies 32 intents and extracts six entities to resolve tickets, show tickets, surface KB articles, or trigger automated workflows (demoed by provisioning GitHub access). Models use a Rasa pipeline, a fine-tuned BERT-based classifier for service-item matching, Elasticsearch plus semantic similarity for long articles, TensorFlow/TF-Serving, model storage on S3/EFS and serving via ECS. Team runs ~100 experiments across 165 metrics, trains on ~10k rows, sees peaks ~500 req/min, and logs feedback for continuous retraining.
AI-written summary · about
chapters (6)
- 0:00Introduction to virtual agent
- 1:25Demo of virtual agent
- 5:06Virtual agent architecture
- 7:33Data science aspects
- 15:45Model lifecycle and monitoring
- 20:03Q&A session
related talks
34mAutomating SRE Operations with Multi-Agent AI
19mAutomating 90% of support tickets: How we built Vercel's support agent
13mExploring AI-Powered Chatbots and Their Transformative Role
32mSREBot—More Than a Chatbot—An Intelligent Bot to Crush Mitigation Time
19mThe Future of AI
24mSREBot More than a Chatbot...
47mBuilding Intelligent Enterprise-Grade Help Desk Bots using Conversational AI
56mBots - the next UI revolution