How to Set Up Cold Email Autopilot (Settings That Keep You Out of Spam)

Quick answer: Cold email autopilot is safe when it is narrow and slow: one or two niches and a few cities per automation, a minimum score so only businesses with a visible problem are emailed, 5 to 10 emails a day from your own mailbox, one follow-up that stops the moment anyone replies, and a hard rule that no business is ever emailed twice. Set it up once, then review it every week for the first month.

The promise of autopilot is that outreach keeps happening on the weeks you are busy delivering work, which are exactly the weeks it usually stops. The risk is the same thing seen from the other side: a system that keeps sending after it should have stopped, to people who should never have been on the list.

Everything below is about getting the first without the second. If you want the case for automating the finding as well as the sending, start with how to automate cold outreach without buying a list. This post is the setup guide.

What autopilot should actually do

A real outreach autopilot runs the whole loop, not just the send button:

• Finds businesses that match your niche and cities.
• Audits each one, so there is a specific reason to write.
• Writes the email from that audit.
• Sends it from your own mailbox, at a human pace.
• Follows up once, and stops on any reply.
• Never contacts the same business twice, and honours unsubscribes everywhere.

If a tool only does the last three, you will still spend a day a week building lists. If it does all six, the settings below decide whether it works.

1. Pick niches and cities as combinations

Each niche in each city is its own search: dentists in Leeds, dentists in York, physiotherapists in Leeds. Start with one or two niches you can genuinely speak to, and two or three cities. That gives you four to six searches, which is plenty.

The temptation is to tick ten trades across a whole region. Resist it. The narrower the target, the more specific the emails stay, and specific emails are the ones that get replies. You can widen once you know which combinations answer. If you have not settled on a niche yet, our guide to choosing an agency niche helps.

2. Decide what you are selling

If you sell one thing, say website redesigns, pick that service. Businesses are then ranked by how badly they need it, and every email leads with it.

If you sell several things, let the system pitch each business its biggest problem instead. A dentist with a fast site but a 3.8 rating hears about reviews. A gym with a nine-second mobile load hears about speed. One automation, several offers, and every email is about something real.

3. Set a minimum score

This is the most important setting, and the one most people set too low. Every business found gets a score for how much visible need it has. Only those above your threshold are emailed.

Start around 60 out of 100. Below that you are writing to businesses whose site is mostly fine, the email has little to say, and they ignore it. Ignored email is what teaches mailbox providers to filter you. If nothing in a niche clears the bar this week, the correct behaviour is to send nothing, not to lower it. Our lead scoring guide explains what goes into a score like this.

4. Keep the volume human

Five to ten emails a day from one mailbox looks like a person doing outreach. Fifty a day looks like software, and is treated that way. Run it on weekdays, so nothing lands on a Sunday morning.

A rough guide:

• New mailbox or new domain: start at 3 to 5 a day for the first two weeks.
• Established mailbox you also use for normal work: 5 to 10 a day.
• More than that: add a second mailbox rather than pushing one harder.

Send from your own inbox, not a shared sending pool. Replies come straight back to you, and the reputation you build belongs to you. The rest of the deliverability picture is in our spam checklist for agencies.

5. One follow-up, cancelled on reply

Many replies come after the second message, not the first, so a single follow-up is worth turning on. Three days later is a good gap: long enough not to feel pushy, short enough that they still remember the first email.

The rule that matters is that it stops the moment they answer, whatever the answer is. A follow-up that arrives after someone has replied "not interested" is how you get marked as spam. Our follow-up guide covers why one or two is enough.

6. Give them something to look at

An email that says "your website could be better" asks the owner to take your word for it. An email with a link to something about their own business asks them to look. Two things work well:

• An audit report with their own findings: speed, booking, reviews against nearby competitors.
• A concept homepage built from their public details, showing what a new site could look like. More on that in how to use a website mockup in cold outreach.

Pick one per automation and keep the email short around it. The link does the persuading.

7. Write your own template, or let the audit write it

You have two honest options. Let each email be drafted from that business's audit, so the first line is always a fact about them. Or write your own template with merge fields for the specifics: their name, their rating, their biggest problem, their link.

If you write your own, keep the specific part specific. A template where the only personal detail is the business name reads as a template. And include a short opt-out line, so anyone who does not want to hear from you can say so without reporting you. Why merge tags alone are not personalisation is covered in this guide.

What it should handle without you

Once it is running, a good autopilot takes care of the boring edges:

• Running out of leads. When the businesses it found have all been emailed, it searches again for more in the same niches and cities.
• Duplicates. The same practice will turn up in more than one search. It gets one email, ever, across every campaign you have run.
• Unsubscribes. Anyone who opts out, or is on your suppression list, is never emailed again by anything.
• Problems. If the mailbox stops accepting logins or you run out of leads on your plan, it pauses itself and tells you, rather than failing silently.

The weekly review for the first month

Autopilot is set-and-review, not set-and-forget. Every week for the first month, spend fifteen minutes on this:

• Replies by niche and city. If one combination has sent forty emails without a reply, pause it and put the volume somewhere that answers.
• What the replies say. Three price objections is a signal about your offer, not your targeting. Three "we already have someone" means the niche is well served.
• Bounces and complaints. More than the odd bounce means the contact data in that market is poor. Any complaint means the email or the targeting needs another look.
• The score threshold. If replies are mostly from the highest-scoring businesses, raise the bar and send fewer, better emails.

After a month you will know which combinations work, and the review drops to a glance at the reply count.

Autopilot in Tellsign

Tellsign runs this whole loop. You pick niches, cities and the service you sell, set a minimum score, a daily volume and a schedule, and it finds businesses on Google Maps, audits and scores each one, writes the email from its findings and sends it from your own mailbox. It follows up once after three days unless they have replied, and it never emails the same business twice. It can link a concept homepage or an audit report in every email, or send your own template instead, and replies land in one inbox with the conversation threaded. See everything it does.