AI for cold email: where it helps and where it hurts

AI is a good assistant for cold email and a poor replacement for thinking about your market. It can draft faster than you can, and it can also send a thousand confident, bland emails if you let it. The difference comes down to what you give it and whether you read what it gives back.

Where it helps

• First drafts. Give it the audience, problem and offer, and get five different angles in a minute.
• Shortening. "Cut this to 60 words and keep it plain" is one of its best uses.
• Variants for testing. Subject lines and openings to A/B test.
• Personal first lines, from real facts you give it, like a company's description.
• Reply drafts. A first answer to a reply that you check and send.
• Sorting replies. Separating interested replies from out-of-office messages and refusals.

Where it hurts

• Your offer. It can't tell you whether anyone wants what you sell. A weak offer in better prose is still a weak offer.
• Your list. It can't replace knowing who has the problem.
• Made-up facts. Ask for a personal line with nothing to go on, and it may invent a podcast they never did.
• Everyone sounding the same. Prospects get a lot of AI emails, and the phrasing is getting easy to spot.
• Your results. It may state fake numbers about your own work. Never send a figure you can't stand behind.

How to prompt it

A vague prompt gives you a vague email. Give it these every time:

• Who: the job title, industry and company size.
• Their problem, in the words they'd use.
• Your offer, and any real proof you have.
• The ask: one small next step.
• Rules: under 75 words, plain language, no "I hope this finds you well", don't invent facts.

There are 30 ready-made ones in the prompt library.

Read everything before it goes out

1. Is every fact true? Check names, numbers and claims.
2. Does it sound like you, out loud?
3. Could this email go to anyone? If so, it isn't specific enough.
4. Does it cover one problem and one ask?
5. Do the fields fill properly? The merge tag preview shows you.

Fully automatic sending

I wouldn't let AI write and send to cold prospects with nobody checking, at least not until you've read enough of its output to trust it on that exact task. A bad campaign costs you more than time: it can cost you your domains. Replies especially deserve a human, since that's a person who is actually talking to you.

AI doesn't fix deliverability

A perfect email in the spam folder gets no replies. Authentication, warm-up, volume and list quality still come first. Here's the order I check them in.

mailzy's AI writer drafts campaign emails from your brief, and the Unibox can draft a reply for you to check before it's sent.