What cold calling is, and why AI changes only part of it
Cold calling is calling someone with no prior enquiry — a purchased database, a scraped list, a directory. The hard parts are not the dialling. They are that most people do not pick up unknown numbers, many of those who do are annoyed, and Indian regulation treats these calls as promotional.
AI changes the economics of each attempt: it can work through a list with as many calls at once as you have channels, give every connected person the same well-tuned pitch, and log objections word for word. It does not change who wants to hear from you, or the law.
The legal picture first
Under TRAI's Telecom Commercial Communications Customer Preference Regulations, 2018 (TCCCPR), unsolicited promotional calls are regulated commercial communication. In practice that means:
- —Numbers registered on the National Customer Preference Register (DND) must not receive promotional calls in the categories they have blocked.
- —Promotional voice calls are meant to come from registered telemarketers using the designated 140-series numbering, and senders must register on the DLT platform.
- —TRAI has directed telecom operators to act against unregistered senders making promotional calls from ordinary 10-digit numbers, up to disconnection and blacklisting.
- —Calls should stay within daytime hours; late-night promotional calling invites complaints.
- —The DPDP Act 2023 governs how you collect and use the personal data on the list.
An AI voice agent is exempt from none of this. Outpero calls from a DLT-registered line and screens bulk lists against DND, but a bought database with no consent is still a bought database. Before any cold campaign, confirm with your vendor — Outpero included — how promotional calls will be classified and routed, and take compliance advice if the list's origin is unclear.
Cold, semi-warm and warm: where AI calling fits
| List type | Example | Regulatory footing | Fit for AI calling |
|---|---|---|---|
| Truly cold | A purchased database of "HNI contacts" | Promotional; DND applies; consent doubtful | Poor — low pickup, high complaint risk |
| Semi-warm B2B | Businesses that sent you an IndiaMART enquiry, trade-fair visitors who shared a card | Enquiry-based; still check DND for promotional pitches | Good — there is a clear reason for the call |
| Lapsed customers | Patients not seen in a year, gym members whose plan ended | Existing relationship; keep the call relevant | Good — reactivation calls |
| Warm inbound leads | Meta lead ad, website form, portal enquiry | The person requested contact | Best — call within 30 seconds |
The pattern is clear: the more the person expects the call, the better AI calling performs and the safer it is. That is why Outpero leads with instant lead callback and treats cold calling as a secondary use case. A useful exercise before any cold campaign is to ask what share of your list you could defend, in writing, as having a reason to hear from you; call that share first.
Realistic expectations on pickup
Hear it on a real call before you read further?
Pickup on cold lists in India is low for anyone, human or AI, because people screen unknown numbers and caller-ID apps flag high-volume callers as spam. An AI agent does not fix that. What it does change is the cost of trying: Outpero bills only connected talk time, so numbers that ring out or are switched off cost nothing, and a 30-second "not interested" costs a couple of rupees rather than a telecaller's time.
Measure cold campaigns on cost per qualified conversation, not cost per dial. If the first few hundred numbers produce almost no interest, the list is the problem, not the caller. For one-way offers where no conversation is needed, Outpero's voice broadcast plays a fixed message at a lower rate than AI minutes — but it is subject to exactly the same DND rules.
How an AI cold call should be structured
- 01Say who is calling and why in the first sentence, and — good practice — that it is an AI assistant.
- 02Ask for 30 seconds, and end the call politely if the answer is no.
- 03Ask one qualifying question before pitching.
- 04Offer a low-commitment next step: a WhatsApp brochure, a callback from a specialist, a demo slot.
- 05Record "do not call again" requests as a captured field and honour them across every campaign.
ನಮಸ್ಕಾರ, ನಾನು Bright Solar ಕಡೆಯಿಂದ AI assistant ಮಾತಾಡ್ತಿದ್ದೀನಿ. ನಿಮ್ಮ ಮನೆಗೆ rooftop solar ಬಗ್ಗೆ ಒಂದು ನಿಮಿಷ ಮಾತಾಡಬಹುದಾ?
Cost of AI cold calling
Outpero has no setup fee, no monthly base and no contract. Credits cost ₹1 each, never expire, and are spent only on connected minutes: ₹5/min on Standard (the usual choice for bulk campaigns), ₹6.8 on Premium. A dedicated number is ₹649/month. Compare a human telecaller at roughly ₹18,000 a month who spends much of a cold-calling shift listening to phones ring.
Choosing an AI cold calling tool
- —Insist on DND screening and ask exactly how promotional calls are classified and routed.
- —Look for per-connected-minute billing, so dead numbers do not cost money.
- —Check retry logic and calling windows, so nobody is redialled five times in a day.
- —Make sure opt-out requests are captured as data, not buried in recordings.
- —Check whether it can answer when prospects call back the number — on Outpero, the same agent picks up 24/7.
When not to use AI for cold calling
Do not use it to blast purchased lists without consent, to make product recommendations on loans, insurance or investments that RBI, IRDAI and SEBI expect a qualified person to make, or to open high-ticket B2B deals where the first call should come from a senior salesperson. Use it instead to call the people who already raised a hand — faster than anyone else does.
- TRAI — TCCCP Regulations 2018, consolidated with amendments (21 May 2026)
- TRAI — Direction on disconnecting and blacklisting unregistered senders (13 Aug 2024)
- TRAI — Telecom Commercial Communications Customer Preference Regulations, 2018
- MeitY — Digital Personal Data Protection Act, 2023
- SEBI — (Investment Advisers) Regulations, 2013 (Gazette)
