OpenAI Email Reply Automation
Budget / SalaryHourly project
TypeFreelance project
LocationRemote
Posted34 minutes ago
I run a support-heavy inbox and want to offload the first draft of every customer reply to an AI workflow. The idea is simple: when a new message lands in my email (Gmail/Outlook—whichever is easier for you to hook up), an n8n or Make scenario should pull the thread, feed the relevant context to OpenAI, and return a well-structured draft. Nothing is ever sent automatically; the draft just pops back into the same inbox, clearly flagged for my review so I can hit send, tweak, or discard.
Here is what I need built and handed over:
• A working n8n or Make scenario that watches the chosen mailbox, collects the incoming message—including previous conversation history when available—and passes it to OpenAI (GPT-4 or 3.5-turbo, whichever produces the best balance of speed and cost).
• Prompt engineering that ensures the draft: a) answers the customer’s question directly, b) keeps the tone warm and professional, and c) leaves placeholders when information is missing so I can fill them in quickly.
• A simple mechanism (label, custom header, or folder move) that drops the AI-generated reply back into my inbox for approval while leaving the original thread intact.
• Clear README or loom walk-through showing how to change prompts, swap email accounts, or adjust the trigger interval.
Acceptance criteria:
1. New customer emails trigger the automation within five minutes.
2. Draft replies contain no hallucinated data and follow my style guide (I’ll share examples).
3. I can edit and send directly from my normal email client with no extra clicks beyond opening the draft.
If you’ve wired OpenAI into n8n or Make before, this should be straightforward. I’m happy to test in real time and give rapid feedback so we can iterate quickly.
Here is what I need built and handed over:
• A working n8n or Make scenario that watches the chosen mailbox, collects the incoming message—including previous conversation history when available—and passes it to OpenAI (GPT-4 or 3.5-turbo, whichever produces the best balance of speed and cost).
• Prompt engineering that ensures the draft: a) answers the customer’s question directly, b) keeps the tone warm and professional, and c) leaves placeholders when information is missing so I can fill them in quickly.
• A simple mechanism (label, custom header, or folder move) that drops the AI-generated reply back into my inbox for approval while leaving the original thread intact.
• Clear README or loom walk-through showing how to change prompts, swap email accounts, or adjust the trigger interval.
Acceptance criteria:
1. New customer emails trigger the automation within five minutes.
2. Draft replies contain no hallucinated data and follow my style guide (I’ll share examples).
3. I can edit and send directly from my normal email client with no extra clicks beyond opening the draft.
If you’ve wired OpenAI into n8n or Make before, this should be straightforward. I’m happy to test in real time and give rapid feedback so we can iterate quickly.
Apply on Freelancer →
Project sourced from Freelancer.com. Applications happen directly on the original platform — we never collect your data.