Old customer list reactivation calls that people answer

The list sitting in your Excel file

Every business that has been running for a few years has one. An Excel file or an old billing system with two, five or ten thousand customers who bought once and never heard from you again. A diagnostic lab with patients from 2023. A car service centre with owners whose cars were last serviced 14 months ago. A jewellery shop with festival buyers.

These people already know you. Calling them is usually far cheaper than buying new leads. The reason nobody does it is that calling 4,000 numbers by hand is a month of a telecaller's life.

An AI caller changes that maths. It can work through the whole list in days, inside your calling hours, retry the ones who did not pick up, and hand your team only the people who said yes. The hard part is no longer the dialling. It is cleaning the list and knowing what to say.

First, consent and DND

Before a single call, clean the list. This is not optional.

• Remove anyone who asked not to be called. Put them on your do-not-call list so no campaign reaches them again.
• Check DND. India's TRAI rules restrict promotional calls to numbers registered on the national do-not-disturb registry unless you have the person's consent. A service reminder about something they bought is different from a sales offer; know which one you are making.
• Use data for the purpose it was given. The DPDP Act expects personal data to be used for the purpose people agreed to. A customer who gave a number for a bill did not necessarily agree to every offer.
• Remove duplicates and dead numbers. They waste attempts and skew your results.

ElevateBox checks the do-not-call list, consent records and calling hours before each call. How to load those is in do-not-call and consent. If you are unsure where the line is, read is AI calling legal in India before you start.

Segment before you call

One script for the whole list is the most common mistake. Split it into groups that need different conversations:

Each group becomes its own bulk campaign, with its own short brief. Smaller groups also make results easy to read.

Keep the fields that make a call sound personal: first name, what they bought, roughly when, and which branch. "Your Swift was last serviced at our Kukatpally branch in August last year" gets a different response from "We are calling our old customers". Drop the fields you are not sure are correct; a wrong detail is worse than none.

Start with the segment most likely to say yes, usually the people who are clearly due for something. Early wins tell you the brief works before you spend minutes on harder groups.

What the AI caller should say

Lead with the relationship, give one reason, ask one question. For a car service centre calling an owner in Telugu:

• "Namaskaram andi, [business name] service centre nunchi matladutunna. Mee car last service ayyi one year daatindi andi." (Hello, calling from the service centre. It has been over a year since your car's last service.)
• "Ee week lo free check-up slot undi. Thursday leda Saturday, edi convenient?" (There is a free check-up slot this week. Thursday or Saturday, which suits you?)

And in Hindi for a lab: "Aapka last health check-up pichhle saal hua tha. Is mahine home collection ka slot book kar doon?" (Your last check-up was last year. Shall I book a home collection slot this month?)

If someone says "don't call me again", the AI employee should say sorry, end politely, and that number goes on the do-not-call list. That is part of the brief, not an afterthought.

Running it as a bulk campaign

1. Upload the cleaned segment as a list with name, phone and one or two useful fields, like last purchase date.
2. Write the brief for that segment, using those fields so each call sounds personal.
3. Set calling hours and how many times to retry a no-answer.
4. Start with 100 to 200 numbers. Listen to ten recordings. Fix what sounds off.
5. Release the rest of the segment.

The general mechanics of lists, retries and pacing are in the bulk calling campaign guide and bulk campaigns help.

What to measure

• How many picked up, by segment
• How many booked a slot, visit or order
• How many asked not to be called again (if this is high, your reason to call is weak)
• Wrong numbers, so you can clean the master list
• Cost per booking: talk minutes at ₹7.98 a minute, divided by bookings

What it costs, and what to expect

An illustrative example, not a promise: you call 1,000 old customers. Say 450 pick up at some point and the average conversation is about a minute and a half. That is roughly 675 minutes, or about ₹5,387 at ₹7.98 a minute, plus GST. The 550 who never answered cost nothing.

Expect old lists to pick up less than fresh leads, and expect more wrong numbers. Do not judge the campaign on pick-ups. Judge it on bookings and orders that would not have happened otherwise. If one segment books well and another does not, call more of the first and rethink the reason for the second.

Repeat the good segments every few months. A service centre that calls owners as their service falls due, every month, turns a one-time clean-up into a steady stream of bookings.

When a person should take over

• A customer brings up an old complaint. The AI employee should apologise, note it, and mark it for your manager to call back the same day.
• A high-value buyer asks for the owner or a known salesperson by name.
• Someone wants a price or discount the brief does not allow.
• A family member answers and says the customer has passed away. The AI should apologise, end the call, and the number should come off every list.

Read the summaries every evening during the campaign. The handovers are where the goodwill is won or lost.