You’ve probably seen this opener before: “I hope this message finds you well.”
That’s the sentence that kills most AI cold emails. Not because it’s offensive — because it’s a signal. It tells the recipient that the sender didn’t actually think about them when writing this. They thought about writing an email, and this is what the beginning of an email sounds like.
The rest of the typical AI draft isn’t better. The value proposition is “I believe there could be a mutually beneficial opportunity here.” The ask is “I’d love to connect and explore potential synergies.” Nobody wrote those sentences. They appeared because they were the average of every cold email the model had ever seen.
Try fixing this cold email
Classic AI cold email. Generic opener, vague value prop, no reason to reply.
Cold emails are the worst place to let AI do the whole job, because the one thing that makes a cold email work is that it sounds like a specific person wrote it to a specific person. The recipient has to feel that you know something about them that you didn’t just look up in thirty seconds.
AI doesn’t know that. It knows what cold emails sound like in aggregate. And so it produces something that pattern-matches to “professional outreach” and pattern-matches to nothing else.
What to replace first
The opener is the highest-leverage fix. “I hope this message finds you well” and “I came across your work” are filler that costs you the first five seconds of attention. Replace them with the thing you actually know.
“Cut the first two sentences and start with the specific thing I noticed about their product last week — the thing I actually wanted to mention.”
The value proposition almost always needs a real example. “We help SaaS companies improve retention” is something twenty other companies could say. The version that gets a reply is the one that names a situation close to theirs.
“Replace the second paragraph — say what we did for a team that was in the same situation she’s probably in right now, and say it specifically.”
The ask usually needs to get smaller. “Let’s explore synergies” is a big ask dressed up in vague language. A real ask names exactly what you want and takes thirty minutes to honor.
“Change the ask — don’t say ‘explore synergies’, say I want to show her the one thing we built that I think applies to what she’s doing.”
The real problem
The problem isn’t that the AI wrote a bad email. The AI did what it always does — it produced a structurally correct version of the thing you asked for.
The problem is that a cold email is a document that only works if it has you in it. Your specific knowledge of the person. Your specific reason for reaching out now. Your specific ask that only makes sense given both of those things.
None of that is in the prompt you gave the AI. So none of it made it into the draft. And then you made small edits to word choice and sent it anyway, which means you sent an email that sounded like it was written by someone who didn’t know the recipient and didn’t have a real reason to write.
That’s what the recipient felt. They just didn’t know why.
The draft is a skeleton. The opener, the structure, the closing line — that’s the scaffolding. What needs to go in are the three or four sentences that only you can write, because they’re about this specific person, your specific connection to their work, and the thing you actually want to happen. You have to put those in yourself. The AI was never going to have them.