AI Outreach Message Generator

Why Most Cold Emails Get Ignored

Not because they are cold. Because they are obviously about the sender.

A busy inbox is triaged in about a second per message, on the subject line and the first visible line. In that second the reader is answering one question: is this about me, or about you? Generic templates, unearned familiarity, a paragraph of company background, and "just checking in" all answer it the wrong way.

You will see a specific number quoted for how many emails a professional receives daily. The figures vary widely by role and source and are rarely worth repeating. What matters is not the volume but the triage: your message is being sorted, not read, until it earns otherwise.

A Worked Rewrite

The version that gets sent constantly:

Subject: Quick question

Hi Sarah,

I hope this email finds you well! My name is Bilal and I'm the founder of an AI automation agency. We help businesses like yours streamline their operations and save time with cutting-edge AI solutions.

We've worked with many companies in your space and delivered incredible results. Would you be open to a quick 15-minute call this week to discuss how we could help?

Best regards

Three sentences about the sender before anything about the recipient. No specific observation. An ask that costs the reader fifteen minutes to satisfy the sender's curiosity.

The same email, rewritten:

Subject: Your booking form

Hi Sarah,

Your booking page asks for eight fields before someone can request a viewing. On a phone that is a long way to scroll, and it is usually where a chunk of enquiries drop off.

Most agencies I work with cut that to three and let the rest come from the follow-up call. It normally takes a day to change and the difference shows up in the enquiry count within a fortnight.

Worth a look at your numbers? Happy to send the two changes I would make, no call needed.

What changed. The subject names something of theirs, not a vague hook. The first line is an observation only someone who looked could make, which is the entire personalisation job. The claim is specific and modest rather than sweeping. And the ask is small — a reply, or nothing, rather than a diary commitment.

That last point is the one people resist and it matters most. "Worth a look?" gets a reply. "Do you have fifteen minutes Tuesday?" gets ignored, because agreeing to a meeting with a stranger is a much larger decision than answering a question.

The Three Things That Decide Whether You Get a Reply

Did you look? One genuine, specific observation beats any amount of merge-field personalisation. "I saw you're hiring three support staff" is personalisation. "Hi {{first_name}}, I love what {{company}} is doing" is not, and everyone can tell.

Is the ask proportionate? Match the size of the request to the size of the relationship, which at this point is zero. A question, a link, a one-line answer.

Is it short? Under 120 words is a reasonable ceiling for a first message. Length reads as a demand on the reader's time before they have decided you deserve any.

What You Can Generate Here

• Personalised first touches from a prospect's LinkedIn summary or company description
• Pain-point angles — pick the problem you solve and the email leads with it
• Tone variants — professional, conversational or direct, so you can test what your market responds to
• Follow-up sequences — a three-message drip rather than a single touch, since most replies come after the first

The frameworks underneath are the standard ones: PAS (problem, agitate, solution) for a clearly felt pain, AIDA where you need to build interest before the ask, and the forwardable email, written so your contact can pass it to the actual decision-maker without editing it.

Before You Send at Volume

Not legal advice, and the rules differ by jurisdiction — but these apply to almost everyone reading this.

In the UK and EU, business-to-business email to a corporate subscriber sits under GDPR and PECR. You need a lawful basis, usually legitimate interests, and that requires the balancing test to be done and recorded rather than assumed. The ICO's direct marketing guidance sets out what that means in practice. Individual sole traders and partnerships are treated as individuals, not businesses. Every message needs an easy opt-out and a real identity behind it.

In the US, CAN-SPAM requires a valid physical postal address, an honest subject line and a working unsubscribe honoured within ten working days. The FTC publishes a compliance guide for business that lists all seven requirements and the penalties.

LinkedIn's terms prohibit automated messaging and profile scraping. Tools that automate connection requests and InMail put the account at risk, and account restriction tends to arrive without warning. Automate the research and the drafting; send the messages yourself.

Frequently Asked Questions

Does AI-written outreach get flagged as spam?

Spam filters score sending reputation, authentication and engagement far more than prose style. A message written by a model from a properly authenticated domain with real replies coming back lands fine. What gets you filtered is volume from a cold domain, no SPF, DKIM or DMARC, and a low reply rate. Warm the domain before you scale, and fix authentication first.

How many follow-ups should a sequence have?

Fewer than most sequence templates suggest, and each one should add something. Three or four touches spread over a fortnight, where every message contains a new observation or a different angle, works better than seven that increasingly say "just bumping this". A follow-up with nothing new in it tells the reader you were never really writing to them.

Should I personalise every email by hand?

Personalise the first line by hand and template the rest. The observation is what earns the reply, and it is the one part a model cannot invent for you because it requires actually looking. Generating the body, the angle and the follow-ups is where automation genuinely saves time.

What reply rate should I expect?

It varies so much by market, list quality and offer that any figure quoted as a benchmark is close to meaningless. The useful comparison is against your own previous sends on the same list. Change one variable at a time — subject line, first line, the ask — and measure against your own baseline rather than someone else's.

Related Reading and Tools

How to Automate Lead Generation covers how outbound sequencing fits into a full pipeline, and AI Follow-Up System for Sales covers the multi-touch logic that turns one good email into a working sequence. The Prompt Library and Tester is useful for tuning the generation prompt itself, and the Website Data Extractor for gathering the observation that makes the first line work.

Automating Outreach at Scale

Writing one good email is easy. Writing five hundred, tracking who replied, suppressing anyone already in the CRM, and making sure a reply reaches a human before the next automated message goes out — that is the part that needs building. Let's talk, or take the work on through Upwork.