You have a list of 200 local businesses and an afternoon to write to them. An AI tool can produce 200 emails before your coffee cools. The question is whether those emails will get replies or get you ignored, and what you should change before you hit send.
This guide gives you an honest look at both sides, then a workflow you can use today.
What AI actually does well in cold email
AI is a strong drafting assistant. Used that way, it helps with:
- Getting past the blank page. Ask for five subject lines or three opening lines and you have something to react to instead of staring at an empty draft.
- Tightening long emails. Paste your draft and ask it to cut to under 100 words. It is usually good at trimming filler.
- Rewriting for a different audience. One core offer can become a version for electricians and another for daycares without starting over.
- Spotting unclear sentences. Ask "What would a busy owner misunderstand here?" and you often get a useful answer.
- Writing follow-up variations. A second and third angle on the same offer is tedious to write by hand.
Where AI-written cold emails go wrong
The problems are less about grammar and more about trust.
- It sounds like everyone else. Phrases like "I hope this email finds you well" or "I came across your impressive business" are easy for an owner to recognise as generic, and they signal you did not look closely.
- It invents details. Ask AI to personalise from a business name and it may guess a specialty, an award or a recent project that does not exist. One wrong "fact" about someone's own business kills credibility.
- It overpromises. AI tends to add claims like "double your bookings". If you cannot back that up, it reads as spam and can create real problems for you.
- It writes too much. Default output is often long, with several asks. Keep one call to action per email.
- It hides weak offers. Polished wording cannot fix an offer nobody needs. If the core pitch is vague, AI just makes vagueness sound smooth.
Pros and cons at a glance
| Pros | Cons | |
|---|---|---|
| Speed | Drafts in seconds | Easy to send without checking |
| Personalisation | Can adapt tone by industry | May invent facts about the prospect |
| Quality | Clean grammar, concise edits | Generic phrasing, sameness |
| Scale | Many variations quickly | Mistakes repeat across the whole list |
The last row matters most. A human makes one mistake in one email. A bad AI template makes the same mistake 200 times.
A workflow that keeps the speed and fixes the risks
Here is a process you can run this afternoon.
1. Write the core message yourself
Before opening any AI tool, write three things in plain words:
- Who you are writing to (for example, independent personal trainers in one city).
- The one problem you solve for them.
- The one small next step you want (a reply, a 10-minute call, a free audit).
This is the part AI cannot do for you, because it depends on what you actually offer.
2. Give AI tight instructions
A prompt that works better than "write a cold email":
Write a cold email under 90 words to the owner of a local [category] business. Offer: [your offer]. One question as the call to action. No compliments, no claims with numbers, no phrases like "hope this finds you well". Use {{params.firstname}} for the first name and {{params.name}} for the business name. Do not invent any details about the business.
Telling it not to invent details is important. Telling it the exact merge tags keeps the template usable in your sending tool.
3. Personalise with real data, not guesses
Personalisation should come from facts you have, such as the business category, city and website. If you build lists from Google Maps, you already have these fields. Our guide on how to scrape Google Maps without code covers collecting them.
In GetClient, the list and the campaign live in the same place, so tags like {{params.city}}, {{params.category}} and {{params.website}} are filled in from each lead when you send. That way the personal parts are real, and AI only writes the parts that stay the same.
4. Edit like a sceptical owner
Read every draft out loud and ask:
- Would I say this to the owner's face?
- Is there any claim I cannot prove?
- Is there more than one ask?
- Does any line sound like it could be sent to anyone?
Cut or rewrite anything that fails.
5. Preview real examples before sending
Check how the email looks with actual lead data filled in, including leads with missing first names or odd business names. A preview of a handful of real recipients catches broken tags and awkward sentences. GetClient shows a per-step preview so you can do this before launch. Our cold email launch checklist covers the rest of the pre-send checks.
6. Use AI for follow-ups, with a new angle
Follow-ups should add something, not just "bumping this". Ask AI for a follow-up that offers a different reason to reply, such as a quick tip or a simpler next step. Space them sensibly; see how long to wait between follow-ups.
Rules that still apply, whoever writes the email
AI does not change your responsibilities. Wherever you send from, include who you are, avoid misleading subject lines, and give people an easy way to opt out. Rules differ by country, so check what applies to you and your recipients. This is not legal advice.
Verify your list before sending, keep daily volume reasonable for your accounts, and stop emailing anyone who unsubscribes.
So, should you use AI for cold emails?
Yes, as a drafting and editing partner. No, as an unsupervised writer. Let AI handle structure, trimming and variations. Keep the offer, the facts and the final read in your own hands. That split gives you most of the time savings with few of the risks.