Cold Outreach Email is a Claude AI skill — signal-researched, role-calibrated emails that open with something real.
Open your sent folder and read your last five cold emails out loud. Not skimming — actually reading, as if you're the person receiving them for the first time, with no context about who sent them or why. Notice where your attention starts to slide. Notice the sentence where you'd have already made the delete decision if this had arrived in your inbox from a stranger. That sentence is almost certainly in the first three lines, and the reason you're reaching for delete almost certainly has nothing to do with whether the writing is good.
Cold email gets deleted before the offer lands. The reply decision — or the delete decision — is made on the opening, and the opening fails not because of grammar or subject line optimisation or the length of the CTA, but because it signals something the recipient clocks immediately: this email wasn't written specifically for me. It was written for someone with my job title at a company of roughly my size. That distinction, felt in under two seconds, is the entire game.
There are five specific patterns that produce this response. They appear in template emails, AI-generated emails, and manually written emails alike — because they're not a function of who wrote the email, but of what the writer did or didn't do before writing it.
The Five Patterns That Get Cold Email Deleted
That gap is exactly what the Cold Outreach Email skill for Claude was built to close.
Why AI Makes All Five Patterns Worse
Every pattern above predates AI-assisted writing. They've been endemic to cold email for as long as cold email has existed. What AI does is scale them. A writer producing template cold email manually might send fifty emails a week. AI-assisted outreach at the same effort level produces five hundred. The patterns that were already getting ignored get ignored at ten times the volume, to ten times as many recipients, who all develop a sharper filter for exactly this kind of email because they're receiving it ten times as often.
The irony is that AI, used differently, is the most effective tool available for eliminating these patterns — because the research that removes them is exactly the kind of work AI accelerates. Finding the recent funding round, identifying the job posting that signals a strategic priority, noting the LinkedIn post where the prospect named a problem, cross-referencing the company's recent product announcements with the gap your offer closes — that research takes a human thirty to forty minutes per prospect. It takes a well-built Claude skill file a fraction of that, and the email it produces is built around what it finds rather than around a template that ignores it.
The problem with AI-written cold email isn't that it sounds like AI. It's that it sounds like no one looked.
The structural fix isn't better writing. It isn't a more compelling subject line or a shorter CTA or a different value proposition framing. It's changing what happens before the email is written — specifically, identifying a live signal in the prospect's world that gives the email a reason to exist today, and building the opening from that signal outward rather than from a template inward. Everything else about the email — the offer, the social proof, the ask — follows from that first sentence and inherits its credibility from it.
What a Signal Actually Looks Like
A signal is any datable, verifiable event in the prospect's world that creates a plausible need for what you offer. It doesn't have to be dramatic. A funding announcement is an obvious one, but it's also the one every other sender is using the same week. Less obvious signals are often more effective precisely because fewer people have written to them about it yet.
A job posting that implies a gap. A LinkedIn post where the prospect named a challenge. A recent product launch that creates a new integration need. A leadership hire that typically precedes a particular type of organisational change. A press mention that reveals an expansion into a new market. A company milestone that implies a specific operational pressure. Any of these gives the email a timestamp and a reason to exist today rather than six months ago.
The discipline is connecting the signal to the offer precisely, in one sentence, without requiring the recipient to make the logical leap themselves. "Saw the EMEA expansion announcement — companies at that stage typically discover that their RevOps stack wasn't built for multi-currency attribution, which is where we come in." That sentence has a signal, a plausible implication of the signal, and a connection to the offer — all without the recipient having to work to see the relevance. The email earns the next sentence by doing that work in the first one.
Most senders know signals matter. The obstacle is finding them consistently and writing from them at speed. Manual research per prospect doesn't scale past a small number of high-value targets. The Cold Outreach Email skill is built specifically around this step — researching the signal before generating a word of copy, so the email that comes out is anchored to something real rather than calibrated to a template that erases everything real about the prospect it's being sent to.
The delete rate on cold email is overwhelmingly a signal problem, not a writing problem. Fix the writing without fixing the signal, and you get a more polished version of something that still gets deleted in under two seconds. Fix the signal, and the writing becomes almost secondary — because the recipient is reading the second sentence before they've made a decision, which is the only thing that matters.
The next piece most people tackle from here is a sales page that closes without feeling pushy. If you're working across the full Founder workflow, the Founder bundle covers everything in one place.
Put this to work: the Cold Outreach Email skill for Claude turns everything above into one guided workflow you run in a normal Claude chat. Not ready to buy? Start with a free Claude skill and see how it works first.