Virtual Assistant
Customer-service conversational agents deployed for retail brands including Natura and Avon, resolving customer questions and issues to cut down on human-agent interaction.
Problem
Customer-support volume across multiple retail brands doesn't scale with human agents alone — most incoming questions are repetitive, but customers still expect an immediate, accurate answer.
Approach
Built brand-specific conversational agents trained on each brand's product and policy knowledge, handling common questions and issue resolution end-to-end, with handoff to a human agent reserved for cases the bot can't resolve confidently.
Architecture
Brand-specific knowledge base → conversational agent (Dialogflow/LangChain) → FastAPI backend for business-logic integration → confidence-based handoff to human agents.
Results
- Deployed across multiple retail brands, including Natura and Avon.
- Reduced human-agent interaction for common customer questions and issues.
Future work
Extending the same architecture to additional brands and further reducing handoff rate through better confidence calibration.