Automate AI Customer Email Responses

Our customer support inbox has become a bottleneck, so I want to replace the current manual triage process with an AI-driven system that reads incoming messages and sends accurate, friendly replies without human intervention unless escalation is required.

Here is what I need:
• A workflow that pulls each new support email from our existing mailbox or help-desk (we currently use Outlook and Zendesk) and feeds it into an AI model.
• Intelligent classification of the request so the system understands the intent, sentiment, and urgency.
• Automatic drafting and sending of the full reply, matching our brand voice and support policies.
• A confidence threshold/flag to route uncertain cases back to a human agent.
• Logging of every interaction for audit and future training improvements, ideally inside our CRM.

Acceptance criteria:
1. At least 90 % of routine support questions are handled end-to-end by the AI with a customer satisfaction score equal to or better than our current human baseline.
2. Deployment runs on our cloud environment (Azure or AWS—whichever you recommend) with clear documentation so my team can maintain it.
3. A short hand-over session plus written SOPs covering model fine-tuning, template edits, and fallback workflows.

If you have experience integrating GPT-style LLMs, vector databases, or tools such as Zapier, Make.com, or LangChain to automate email pipelines, I’d love to see examples. Let me know the tech stack you propose and an estimated timeline, and we can get started right away.