How to Run a Customer Reactivation Campaign With AI Calling
Every established business is sitting on a list of people who bought once and then stopped. They are the cheapest audience you will ever address: they already know you, they have already trusted you with money, and you do not have to pay to reach them again. What stops most companies calling that list is capacity, not doubt. This is a practical playbook for running a customer reactivation campaign with AI calling — how to segment the base before you dial, what the call should actually say, the compliance question to settle before the first number, and how to tell a real result from a busy one.
Why the dormant base is the list worth calling
A dormant customer is not a cold lead. They found you once, chose you, and completed a transaction. Whatever objection a new prospect has about whether you are real, they do not have. The reason they stopped is usually mundane — they moved, the need passed, a competitor was in front of them at the right moment, or nobody ever contacted them again.
The list is also already paid for. You spent the acquisition cost years ago. Yet in most companies it goes uncalled, not because anyone decided against it but because the sales team's day is full of newer names. That is a capacity problem, and it is exactly the kind automation is suited to.
Segment before you dial
Calling the whole base with one script is the most common way this fails. Dormancy has causes, and the causes need different conversations. Four segments cover most businesses.
- Recently lapsed — 60 to 180 daysStill remember you, still in the habit. The call is a reminder rather than a pitch, and this segment usually produces the best response rate. Start here, because it also gives you a clean read on whether the script works.
- Long lapsed — one to two yearsThe relationship is cold but recoverable. This call needs a reason to return: something new, something changed, or an acknowledgement that they have been away.
- One-time buyers who never repeatedOften the largest group and the most interesting. Something prevented a second purchase, and the call is worth using to find out what — the answers are frequently more valuable than the sales.
- Unfinished businessPeople who enquired, quoted, booked or abandoned and never completed. Technically not customers, but the warmest names on the list and usually the fastest conversions.
Two exclusions before any of these are called: anyone who asked not to be contacted, and anyone with an open complaint. The first is a legal and reputational necessity; the second turns a reactivation call into an escalation you did not plan for.
What the call should actually say
The reflex is to lead with a discount. It is usually the wrong move: it trains the base to wait for offers, and it fails to find out why they left. A better structure has four beats.
- Say who you are and why you are callingImmediately and plainly. A caller who cannot tell within five seconds who this is and what it is about will hang up, and they will be right to.
- Acknowledge the gap'We noticed it has been a while' is honest and it opens the useful conversation. Pretending no time has passed makes the call feel automated in the worst sense.
- Ask before offeringOne question — what changed, or whether the need still exists. This is where the value is. A campaign that returns a hundred reasons people stopped buying has paid for itself even if it sells nothing.
- Then give a reason to actSomething new, a relevant offer, or simply an easy next step. Match it to what they just told you rather than reading it regardless.
Write the ending as carefully as the opening. A call that finishes without a defined next step — a callback, a booking, a message with details, or a clean removal from the list — produces an outcome nobody can act on.
Settle the compliance question first
This section is short and it is not optional. Calling existing customers is not automatically permitted, and automated calling is frequently held to a stricter standard than a person dialling manually.
- Consent: on what basis are you contacting this person, and is that basis recorded? An old purchase is not consent in every jurisdiction.
- Do-not-call registries and internal suppression: both must be checked before the list is finalised, not after complaints arrive.
- Disclosure: some markets require you to state that the caller is an automated system. Establish whether yours is one of them.
- Recording: notice and consent obligations for recorded calls vary by country and apply to AI calls too.
- Data age: a database that has not been touched in three years may contain people who have exercised a deletion right you did not action.
Get local advice for each market you dial into, before the first campaign rather than after the first complaint. This is a one-off cost that makes every subsequent campaign routine.
Running the campaign without burning the base
A dormant base can be worked once carelessly and never again. Six operational rules protect it.
- Clean first: remove records with no number, and remove duplicates. Calling the same person twice in one campaign is the fastest route to a complaint.
- Call in your customers' hours, not yours. An evening call to a lapsed customer reads as intrusion rather than service.
- Cap the daily volume. Working through 5,000 numbers in two days leaves you nothing to learn from and nobody to fix a bad script for.
- Limit retries. Two or three attempts, spaced out. Beyond that you are not persistent, you are a nuisance.
- Honour a refusal permanently, and make sure that refusal is written back to the customer record rather than just to the campaign.
- Start with a few hundred, listen to the recordings, then scale. The first hundred calls of any campaign are a test, whether or not you treat them as one.
Sequence matters too. Run the recently lapsed segment first: it is the most forgiving audience, and the script you refine on them is the script you take to the colder segments.
What to measure
Reactivation is unusually easy to declare successful and unusually easy to measure honestly. Choose the second.
- Reach rate: how many numbers were actually answered. A low figure here is a data problem, not a script problem, and no amount of message tuning will fix it.
- Outcome mix: interested, call back later, declined, no answer. The shape of this tells you more than any single conversion number.
- Stated reasons for lapsing, aggregated. Frequently the most valuable output of the entire campaign.
- Revenue from reactivated customers, tracked for at least ninety days. A reactivated customer who buys once and disappears again has not been reactivated.
- Opt-outs and complaints, watched closely. If this rises, stop and fix the script before continuing.
- Cost per reactivated customer, compared with your cost of acquiring a new one. This is the number that decides whether reactivation becomes a standing programme.
Deliberately excluded: total calls made. It is the easiest number to produce and it says nothing about whether the campaign worked.
How Vexvon runs a reactivation campaign
The audience is built as a CRM filter rather than a static file — 'no contact in 90 days', 'status = interested', 'source = Instagram' — and the filter is recalculated when the campaign starts, so the list is current at launch rather than at export. An uploaded Excel or CSV file works too, with columns mapped to the system's fields.
Before anything is dialled, the panel shows how many people the campaign will actually reach, with records that have no number and duplicate numbers listed separately and excluded. Calls go out only inside the days and hours you set, with a daily cap and a ceiling on concurrent lines, and a running campaign can be paused, resumed or cancelled — so a wrong script is caught at call fifty rather than call three thousand.
The agent speaks from the same knowledge base as the chatbot, so what it says about prices and terms matches every other channel. Each call is recorded with an outcome — interested, no answer, call back later, declined — plus a transcript, an audio recording, a one-sentence summary and the fields you defined in the script, including why the customer stopped buying if the agent was able to establish it. A no-answer is scheduled for a retry, a call-back-later creates a reminder, and anyone who shows interest reaches your sales team immediately through a Telegram notification the manager can act on directly.
Frequently asked questions
- What is a customer reactivation campaign?A structured effort to bring back customers who have stopped buying, usually by contacting a segmented list of lapsed customers with a relevant reason to return.
- How long does a customer have to be inactive to count as dormant?It depends on your purchase cycle rather than on a universal figure. A useful working definition is two to three times your median gap between purchases — for a monthly product that might be 90 days, for an annual one it might be two years.
- Should we lead with a discount?Usually not. It trains the base to wait for offers and it skips the more valuable step of finding out why they stopped. Ask first, then offer something matched to the answer.
- Is it legal to call our own former customers?It depends on your market and on what consent you hold. A previous purchase is not consent everywhere, and automated calling is often regulated more strictly than manual calling. Get local advice before the first campaign.
- How many attempts should we make per number?Two or three, spaced out, then stop. Beyond that the return falls and the complaint risk rises.
- How do we know it worked?Track revenue from reactivated customers for at least ninety days and compare cost per reactivated customer with your cost of acquiring a new one. Anything shorter measures a conversation, not a customer.
Start with one segment
The correct size for a first reactivation campaign is a few hundred numbers from your most recently lapsed segment — not the whole base. You will refine the script, hear the real reasons people stopped, and have a measured cost per reactivated customer within two weeks. Everything after that is a decision about volume rather than a decision about whether this works.