Services AI Audit Guides Marketplace Blog Contact
Method

Personalisation at scale. Merge fields do not count.

Inserting a company name into an otherwise identical template is not personalisation. It is a template with a hole in it, and recipients recognise the shape immediately — often faster than they would have spotted a message that made no attempt at all. Fake personalisation is worse than none, because it reveals that you tried to seem attentive without being it.

What genuine personalisation requires

One sentence that could only have been written to that specific company, demonstrating that someone looked. That is the whole standard.

Why it gets abandoned

Because doing it properly takes eight to twelve minutes per company — not for the writing, but for the reading that makes the writing possible. At twenty companies a day that is roughly three hours, every day, before anything else happens.

Almost nobody sustains that. So outreach gets done in panicked bursts with generic copy, produces nothing, and gets abandoned entirely. The gap between what works and what people actually do is a time problem, not a knowledge problem.

What changes with AI, and what does not

A language model can read a company's public pages and write a specific opening line in seconds rather than minutes. That genuinely collapses the cost of the research step, which is the step that made personalisation unaffordable.

What it does not change:

A practical safeguard: before any batch goes out, compare every message against every other one. If two share so much as a six-word sequence, something has collapsed into a template and the batch needs rewriting. This is the difference between a hundred individual emails and one email sent a hundred times.

The line you should not cross

Personal life, family, health, anything from a personal social account, anything they did not publish in a professional context. It is unsettling, it damages you the moment it is noticed, and in jurisdictions with privacy legislation it can also be unlawful.

The test is simple: would you be comfortable if they asked how you knew that? Public business information passes. Anything else does not.

Common questions

Can AI write personalised cold emails that do not sound artificial?

Yes, when constrained properly: give it the company's own published material, require it to reference something specific from that material, forbid flattery and superlatives, and check output for repeated phrasing across the batch. Left unconstrained, it produces confident, generic, slightly enthusiastic text that reads exactly like what it is.

How much personalisation is enough?

One genuinely specific sentence, usually the opener. The rest can follow a consistent structure without harm. The purpose is to demonstrate that a person considered this company — one true sentence does that, and five paragraphs of research does not do it five times better.

Is it obvious when a cold email is automated?

It is obvious when it is generic, which is not the same thing. A message that references something real about the business reads as human whether a person or a machine assembled it. A message that could have been sent to anyone reads as automated even if someone typed it by hand.

What information is acceptable to use?

Anything the business published about itself in a professional context: their website, their services, their job listings, their public work, their business social accounts. Not personal accounts, not personal circumstances, not data from breaches or scraped personal profiles. If you would be uncomfortable explaining how you found it, do not use it.

Keep reading
How to write a cold email Finding business emails ethically Building an ideal customer profile

Hire a salesperson once.
Keep them forever.

The Cold Outreach Machine finds the companies you sell to, writes a different email for each one, and sends them on its own — paced, inside working hours. US$299.99, paid once, live setup call included.

See the machine Buy once vs subscribe