n8n AI Support Automation -- 2
Budget / SalaryHourly project
TypeFreelance project
LocationRemote
Posted10 hours ago
I need a complete n8n workflow that lets two large-language models—Claude and Google Gemini—jointly run my customer support. The agents must monitor both Email and my embedded Website chat, draft helpful replies in real time, and send them automatically whenever confidence is high.
Key flow
• Incoming message arrives via IMAP or the chat widget’s webhook.
• The text is routed through a prompt chain that lets Claude propose an answer and Gemini cross-check it.
• If both models agree above a confidence threshold, the reply is sent back through the same channel.
• Whenever either model flags low confidence, the full conversation (including suggested answer and metadata) is pushed to a human teammate through a dedicated Slack channel I will provide.
What I expect to receive
• A ready-to-import n8n workflow file.
• API key fields clearly separated so I can drop in my own credentials for Anthropic and Google AI Studio.
• Clear, commented nodes showing where prompt templates live so I can tweak tone or policy wording later.
• Brief README or loom video walking me through setup, environment variables, and how to adjust confidence thresholds.
Acceptance test
If I create a sample email asking an unusual product question and the models decline with a “Need human” flag that lands in Slack, the job is done. Likewise, a common FAQ sent by chat should return a correct, well-formatted answer within ten seconds.
Please keep the solution entirely inside n8n (no external servers) and stick to native or HTTP Request nodes only—no custom TypeScript modules this time.
Key flow
• Incoming message arrives via IMAP or the chat widget’s webhook.
• The text is routed through a prompt chain that lets Claude propose an answer and Gemini cross-check it.
• If both models agree above a confidence threshold, the reply is sent back through the same channel.
• Whenever either model flags low confidence, the full conversation (including suggested answer and metadata) is pushed to a human teammate through a dedicated Slack channel I will provide.
What I expect to receive
• A ready-to-import n8n workflow file.
• API key fields clearly separated so I can drop in my own credentials for Anthropic and Google AI Studio.
• Clear, commented nodes showing where prompt templates live so I can tweak tone or policy wording later.
• Brief README or loom video walking me through setup, environment variables, and how to adjust confidence thresholds.
Acceptance test
If I create a sample email asking an unusual product question and the models decline with a “Need human” flag that lands in Slack, the job is done. Likewise, a common FAQ sent by chat should return a correct, well-formatted answer within ten seconds.
Please keep the solution entirely inside n8n (no external servers) and stick to native or HTTP Request nodes only—no custom TypeScript modules this time.
Apply on Freelancer →
Project sourced from Freelancer.com. Applications happen directly on the original platform — we never collect your data.