Build a Multi-Agent LLM System for Automated Customer Support (LangGraph or CrewAI)

We run a mid-sized SaaS product (project management tools) and currently handle support tickets manually through Zendesk. We want to build an AI agent system that can triage incoming tickets, answer common questions using our knowledge base, and escalate complex issues to a human with a summarized context. We're looking for someone experienced with multi-agent orchestration (LangGraph, CrewAI, or similar) who can design a system with: A router agent that classifies incoming tickets by intent and urgency A RAG-based agent that answers FAQ-type questions using our existing help docs (~300 articles) An escalation agent that summarizes context and hands off to a human agent with all relevant history Integration with Zendesk API for reading/writing tickets We have a working knowledge base already exported as markdown files. We do NOT have any existing agent infrastructure - this would be built from scratch.