AI integration: agentic customer support triage
Agentic LLM workflow that reads incoming tickets, gathers context from CRM and order systems, and drafts a reply for human approval. Not a chatbot.
The brief
A small support team was drowning in repetitive tickets. They did not want a public chatbot. They wanted faster drafts behind the scenes.
What I built
- Agent that pulls customer history, latest order status and known issues before drafting.
- Strict tool boundaries: read CRM and orders, never write without human approval.
- Full audit log of which sources informed each draft.
Outcome
First-response time roughly halved. Quality scores up because drafts now cite the right order numbers on first send.