/Cortex/
A fully autonomous customer support system that handles inbound tickets from five channels through six specialist AI agents — with zero human triage.
Noah's SaaS support team was overwhelmed across email, WhatsApp, live chat, portal, and social. We deployed Cortex — a multi-agent system orchestrated by OpenClaw, with Notion as the live knowledge base. Six specialist agents classify, retrieve, draft, QA, escalate, and analyse every ticket — auto-resolving issues and logging outcomes back to Notion in a self-improving loop.
/Project Goals/
Replace manual ticket triage across five support channels with an autonomous agent swarm that resolves, logs, and learns continuously.
The support team was manually routing tickets and copying responses from a Notion knowledge base. The system needed to handle classification, knowledge retrieval, response drafting, QA checks, escalation decisions, and analytics — all in sequence, per ticket, across five channels simultaneously. Every resolved ticket needed to feed back into the system to improve future responses.
/Results/
Support ticket volume handled autonomously hit 4,000 per month with a 92% accuracy rate from day one.
The system went live handling 4,000+ monthly conversations with 92% answer accuracy measured via human sampling. Average response time improved from four hours to under ten seconds. Support ticket volume sent to human agents dropped by 45% in the first month. The self-improving feedback loop meant accuracy continued rising — hitting 96% by week 8.
OpenClawMulti-Agent AINotion APIVector DatabaseRAGCustomer Support AutomationClaude