Team using LangGraph for customer support bot
“80% reduction in query resolution time; automated 70% repetitive tasks; equivalent to 700 FTE; handles 2.5M conversations for 85M users”
startup
Why they built it
To scale customer support across global markets amid rising expectations and overload from millions of daily conversations/transactions
What worked
LangGraph's controllability for routing/low latency; LangSmith for test-driven dev, tracing, evaluations, meta-prompting which reduced token costs and improved reliability
What broke or was painful
Escalation overload and multi-departmental coordination challenges at massive scale (2.5M daily transactions)
The result
80% reduction in query resolution time; automated 70% repetitive tasks; equivalent to 700 FTE; handles 2.5M conversations for 85M users