intermediate / 25 minutes
AI Cold Email Workflow
Build a first outreach draft that names what is observed, what is unknown and why the message may be relevant.
Result: A concise outreach draft with a genuine reason to contact the person, a narrow ask and no invented personal detail.
Jump to prompt ↓5 steps to a reviewable draft
- 01
Set a legitimate reason to write
State the specific public signal or referral that justifies contact, then discard a message if no such reason exists.
Result: A clear outreach premise or a decision not to send.
- 02
Record observed details
Keep source text, URLs and dates separate from guesses about the recipient's priorities.
Result: A short source record with unknowns visible.
- 03
Draft one first-touch email
Ask for a brief email that cites only the supplied context and makes one modest request.
Result: A bounded email draft with no pretend familiarity.
Your prompt
Draft a cold email using only this permitted research: {{source_notes}}. Recipient role: {{recipient_role}}. Reason to write: {{reason}}. Offer or question: {{ask}}. Write a subject line and a plain-text email under 120 words. If the source does not support personalization, use a general role-based opening. Do not invent achievements, tools, priorities, mutual contacts or urgency.Example inputRecipient leads operations; a public job post mentions weekly reporting. Ask whether manual reporting is still a live issue.
Example outputSubject: Weekly reporting at [Company]. I saw the operations opening mention weekly reporting. Is reducing that work on your team's list this quarter?
- 04
Check relevance and consent boundaries
Confirm that the source is permitted, the claim is supported and the ask fits the recipient's role.
Result: A message the sender can defend line by line.
- 05
Send or stop
Personalize the final details manually, choose the approved sending channel or decide that no email should go out.
Result: A reviewed first-touch message or a documented no-send decision.
What a finished draft should contain
A short email that makes one evidence-backed connection and leaves room for the recipient to decline.
What this process cannot decide
- Public research can be stale or misleading
- AI cannot determine lawful contact grounds or sender reputation.
