Outbound that earns a reply

Volume stopped being the strategy

Sending ten thousand near-identical messages used to produce a predictable number of meetings. It now produces spam complaints, a damaged sending domain, and a list burned for anyone who tries later.

The economics inverted. A hundred contacts with a specific reason for contacting each one outperforms ten thousand without, and it does not destroy the asset on the way through.

• Fewer accounts, chosen because something about them makes now the right time
• A first line that could not have been sent to a different company
• A reason to reply that is about them rather than about your capabilities
• A sequence that stops immediately on any reply, including a negative one

Research is the part AI actually changes

The reason outbound was generic was never laziness. Researching a hundred companies properly took a week, so people sent generic messages because the alternative did not fit in the day.

That constraint moved. Reading a company's site, recent announcements, job postings and public activity to find a genuine trigger is now fast enough to do at scale. What has not changed is that the trigger has to be real. A generated compliment about someone's website is more obviously automated than no personalisation at all.

If the personalisation could apply to fifty companies, it is not personalisation. It is a template with a variable in it, and everyone can tell.

Deliverability is the whole game underneath

None of the message quality matters if the message lands in spam. Deliverability is a technical discipline with real rules, and most failed outbound programmes failed here rather than on copy.

That means separate sending domains so the main one is never at risk, authentication configured properly, sending volume warmed gradually, and lists validated before use. It also means watching reply and complaint rates as health metrics rather than as performance metrics.

• A separate sending domain, so the primary business domain is never exposed
• SPF, DKIM and DMARC configured and verified rather than assumed
• Volume increased gradually rather than switched on at full rate
• Lists validated before sending, since bounces damage the sender faster than anything else

What happens after a reply

Most outbound programmes invest everything in getting the reply and nothing in handling it. A response that sits unanswered for a day undoes the work that produced it.

Replies should reach a person quickly, sequences should stop the moment one arrives, and a negative reply should be recorded so the same person is not contacted again next quarter by a different campaign.

What this costs to run

Outbound is an ongoing operation. There is a setup phase with an end: domains bought and warmed, records prepared, the first sequences written and tested. After that it is monthly, because lists decay, deliverability moves, and the only version of this that works is one somebody is watching every week.

The cost drivers are specific. How many separate audiences you are approaching, since each needs its own research angle and its own copy. Data quality, because an unverified list produces bounces and bounces damage the sending domains you have just paid to warm. And whether you want meetings booked and handed over or only replies passed on, which is a different amount of human involvement entirely.

Before a number exists, agree what happens after a reply. Outbound that generates interest nobody follows up on is the most expensive marketing there is, and a provider who quotes the sending without asking who answers has quoted half a system.

• How many distinct audiences and messages are being run
• The state of the data, and whether verification is included
• Sending infrastructure: how many domains and mailboxes are needed
• Whether the engagement ends at a reply or at a booked meeting

When outbound has nowhere to aim

If the buyer is not identifiable, outbound has nowhere to aim. Categories where the purchase is impulsive or the buyer is anonymous are better served by demand capture than by prospecting.

If nobody is available to run the conversations outbound creates, it is a machine for generating unanswered replies. The capacity to respond is a prerequisite, not a detail to arrange later.

What happens when somebody actually replies

Outbound is the part of this work where the machine should do the least. Research and sequencing automate well. The reply does not. When a prospect answers, the conversation hands over to a person immediately with the thread and the research attached, because a second scheduled message landing after a real human being has typed something is how a warm reply becomes a spam complaint.

Anything that commits the company waits for a person too. A price, a discount, a promise about a delivery date, a response to somebody who is annoyed. Those are labelled augmented before the workflow goes live, meaning the machine drafts and somebody signs off. You see the label before you approve the workflow, so nobody is left guessing whether a human was meant to be involved.

Ali runs marketing and sales here, and Wobble stays answerable for what sequences do in your name. That matters more in outbound than anywhere else, because the damage lands on a domain you have to keep using for years. The domain, the sending accounts and the data are yours. You own the lists and the infrastructure, and if you leave you take both.

Where outbound will not work: a market too small to prospect into, an offer nobody has bought yet, or a sending reputation that has to be repaired before a single message goes out. Doing it in-house is entirely reasonable if somebody there understands deliverability and enjoys research. What gets underestimated is warm-up. The audit takes the first week, sending starts small and slow, and a sequence worth judging is usually a month old rather than a fortnight.

Common questions

Does cold email still work?

It works at low volume with genuine research and fails at high volume without it. Filters and recipients have both adapted, so the version that produces meetings now is a small number of contacts each with a specific reason for being contacted. The high-volume version mostly produces complaints and a damaged sending domain.

How does AI improve outbound prospecting?

By removing the research constraint. Reading a company's website, recent announcements, job postings and public activity to find a genuine trigger used to take long enough that people sent generic messages instead. Doing that at scale is now practical. The trigger still has to be real, because generated flattery is more obviously automated than no personalisation at all.

What ruins email deliverability?

Sending from the main business domain, skipping authentication, ramping volume too quickly, and mailing unvalidated lists. Bounces damage a sender faster than anything else. Most outbound programmes that fail, fail here rather than on the writing.

How many prospects should be contacted per week?

Fewer than most tools encourage. The right volume is set by how many you can research properly and how many replies the team can actually handle, since an unanswered reply wastes the work that produced it. Sending capacity is rarely the binding constraint.

Is automated outbound legal?

Rules differ by market and cover consent, identification of the sender, and a working way to opt out. Business-to-business email is treated differently from consumer email in most jurisdictions, and messaging channels such as WhatsApp have their own separate consent requirements. It needs checking per market rather than assuming one answer travels.

What should happen when a prospect replies?

The sequence stops immediately, including on a negative reply, and a person picks the conversation up quickly. Negative replies are recorded so the same person is not contacted again next quarter by a different campaign, which is one of the fastest ways to convert a soft no into a complaint.