Cold outreach that reads like you wrote it
AI researches each prospect individually and drafts a one-to-one email grounded in what their company actually does, at a volume no salesperson could match.
The problem
How it works by hand
Generic mail-merge outbound is dead. Buyers can smell a {{first_name}} template from the subject line and delete it on sight, and enough of them hit "report spam" that your domain reputation pays for it. The alternative, properly researching every prospect by hand, takes twenty minutes per email, so nobody does it past the first ten.
A worked example
What a working version looks like
The system takes your prospect list and researches each company one at a time: their website, what they sell, recent signals like a new office or a job ad. An LLM then drafts an email for that one prospect, referencing something true and specific about their business, in your voice and to your rules, no fake "I loved your recent post" filler. Drafts queue for your approval or send automatically once you trust the output, throttled and spread across sending accounts so deliverability holds. Replies route back into the same system, so a "tell me more" never gets lost under the out-of-offices. The research that took twenty minutes per prospect now runs while you sleep.
The exact tools change per business. The shape does not.
What it needs
Honest inputs, nothing exotic
- 01A prospect list, or we build one (see the prospect list use case)
- 02Sending inboxes warmed for cold email (we set these up)
- 03Two or three of your best past emails, so the AI learns your voice
- 04Your rules: who to never contact, what to never claim
The payoff
What you get back
Outbound that earns replies instead of spam complaints. Each email reads like a person spent twenty minutes on it, because the system effectively did. You spend your time on the conversations that come back, not on the drafting.
Do it yourself
How you would build this yourself
No course, no upsell. This is the order we would build it in, with the tools named, and a prompt to start from.
- 1
Start with twenty prospects in a spreadsheet, not two thousand. Name, company, website. The research is the product, volume comes later.
- 2
Collect the raw material per prospect: their homepage, about page and anything recent. A small script, which Claude Code will write for you, fetches and summarises each one.
- 3
Draft with rules, not vibes: give the model two of your best emails for voice and require one true, specific observation per draft. Ban flattery and filler outright.
- 4
Send low and slow from a separate warmed domain, never your main one. Smartlead or Instantly handle warm-up and throttling, and plain Gmail at twenty a day works to start.
- 5
Read every reply yourself at this volume and feed what worked back into the prompt. The prompt after fifty sends should be twice the prompt you started with.
Build me a personalised outbound drafting tool. Input: a CSV of prospects with name, company and website URL. For each row: 1. Fetch the company's homepage and about page and summarise what they actually do 2. Draft a three-sentence cold email that references one true, specific thing about their business. No flattery about blog posts, no "I hope this finds you well" 3. Write the draft and the research summary into new columns in the CSV for me to review My voice: I will paste two of my best emails below. My rules: never claim we have worked with a company we have not, and skip any company where the research comes back thin rather than guessing. Do not build any sending. I will review and send the first batches myself. Start with a plan and an estimate of the API cost per hundred prospects. My two best emails: [paste them here]
Copy it into Claude Code, fill the brackets, and it will plan the build with you before writing a line of code.
We would rather show you how than bill you. The whole ladder of free help, answers, guides and the weekly build-along, is on the do-it-yourself page.