Personalizing Emails at Scale
Sending the same email to fifty people is easy. Sending fifty emails that each feel like they were written specifically for that person is hard, and it used to require either a large team or a lot of very late nights. AI tools change that equation, letting one person personalise outreach at a volume that would previously have needed real automation infrastructure and a copywriting budget. This lesson covers how to build a strong template once, then use AI to genuinely personalise it for each recipient, rather than just swapping in a first name and calling it personal.
Why first-name personalisation is not enough
Mail merge tools have let people insert a recipient’s first name into a template for decades, and most experienced professionals have learned to recognise and discount this kind of personalisation instantly, because the rest of the email is obviously identical to everyone else’s. Real personalisation changes something substantive: a reference to the recipient’s specific situation, a detail that proves you looked at their company or their role, an example chosen because it is relevant to their industry rather than generic. AI makes this level of personalisation achievable at volume because it can take a short, factual brief about each recipient and weave it naturally into a consistent template, far faster than a human writing each version by hand.
The two-layer method
The most efficient approach separates the email into two layers. The first layer is a fixed core: your offer, your main value proposition, your call to action, written once and refined until it is strong. The second layer is a set of personalised variables specific to each recipient: their name, company, a relevant detail, and sometimes one custom sentence. Write the fixed core yourself with AI help using the techniques from earlier lessons, then use AI to generate the personalised layer for each recipient from a short data point you supply.
A practical prompt for the second layer: “Here is my email template with a placeholder marked [PERSONAL DETAIL]. For each of the following five companies and one fact about each, write a single sentence that fits naturally into that placeholder, referencing their specific situation. Company 1: Northwind Logistics, recently expanded into three new UK regions. Company 2: Bellcrest Retail, posted about struggling with return rates last month. [continue for each].” This produces five distinct, relevant sentences in one response, each ready to drop into the template, rather than five entirely separate emails to draft from scratch.
A simple personalisation workflow
| Step | What you do | What the AI does |
|---|---|---|
| 1. Build the core | Write and refine one strong template using techniques from earlier lessons | Helps draft and tighten the fixed template |
| 2. Gather recipient facts | Collect one or two real, specific facts per recipient, five minutes of research each | Nothing, this step is yours, AI cannot know these facts |
| 3. Generate personal lines | Supply the facts in a batch prompt | Writes a tailored sentence or opening line per recipient |
| 4. Merge and review | Combine template and personal line for each recipient | Can help proofread the merged version if pasted back in |
| 5. Spot-check before sending | Read three or four at random in full before sending the batch | Nothing, this check must be human |
Where the research still has to be human
AI cannot browse a hundred LinkedIn profiles or company websites for you inside a single chat unless it has live tool access, and even then, accuracy on individual facts about specific people needs verification, since models can occasionally produce plausible-sounding but incorrect details about real companies or people, particularly for anything outside their training data or requiring current information. Treat any AI-suggested fact about a specific recipient as a claim to verify, not a claim to trust, especially before referencing it directly to that person, since a wrong detail in a personalised email is worse than no personalisation at all, it signals carelessness rather than attention.
Segment-level personalisation as a middle ground
When individual research for every recipient is not realistic, personalising by segment is a strong middle ground. Group recipients by shared traits, industry, company size, job function, and write one tailored version per segment rather than one universal email or fifty individual ones. A prompt such as “Write three versions of this email, one for HR directors focused on retention costs, one for finance directors focused on budget efficiency, one for operations directors focused on process time savings, same core offer, different framing for each” gives you meaningfully differentiated messaging for a fraction of the effort of fully individual writing.
Quality control at volume
The risk of scale is that errors also scale. Before sending any batch of AI-personalised emails, spot-check a genuine random sample, not just the first two, for merge errors, wrong names attached to wrong facts, awkward phrasing where a personal detail was forced into a sentence that does not quite fit. A quick prompt to catch this systematically: “Here are ten merged emails. Flag any where the personal detail feels forced, awkward, or factually inconsistent with the rest of the message.” This kind of self-review pass, discussed further in the lesson on editing and proofreading, is what separates AI-scaled outreach that still feels human from outreach that obviously was not.
Personalisation handles emails that go well. The next lesson turns to the opposite challenge: emails nobody enjoys writing, complaints, apologies and saying no, where AI’s tone control becomes especially valuable precisely because getting these wrong carries real relationship cost.
