Automated calling in India sits under TRAI's commercial-communication rules, and the practical consequences of getting it wrong are not fines so much as your numbers being blocked and your calls stopping. This is an operational risk, not just a legal one.
What follows is how to run a calling programme sensibly. It is not legal advice, and if you are calling at real scale you should get some.
The distinction everything rests on
Indian rules turn on whether a call is promotional or transactional, and the difference is not what the call is about — it is the relationship you already have with the person.
- Transactional / service — the person has an existing relationship with you and the call concerns it. Somebody applied for your job. A customer placed your order. A student enrolled in your course.
- Promotional — you are marketing to someone who has not asked for it. Cold outreach to a purchased list.
Recruitment screening a candidate who applied is squarely in the first category — they initiated contact by applying. Calling a database of people who never applied is squarely in the second, and is where the risk lives.
Five habits worth building in
1. Keep the provenance of every number
For each phone number, be able to answer: where did this come from, and when? "They applied through our careers page on 12 March" is a defensible answer. "It was in a spreadsheet someone sent" is not — and you will not reconstruct it later.
2. Honour opt-outs immediately and permanently
When someone says stop, they must stop — across every campaign, not just the one they were on. This is the most common operational failure: a person opts out of one batch and gets called by the next because the suppression lived in that batch rather than at the account level.
A keypad menu makes this clean and auditable: "press 9 to opt out", and the number is suppressed for all future calls automatically, with a logged keypress showing exactly when.
3. Respect calling hours
Promotional calls are restricted to 9 AM–9 PM. Even where a call is transactional and technically permitted, calling at 8 AM produces complaints, and complaints are what get numbers flagged. Set a calling window and let the system enforce it rather than trusting whoever starts the batch.
4. Identify yourself in the first sentence
Not a rule so much as the thing that prevents everything downstream. "Hi, this is Maya calling from Acme about the Backend Engineer role you applied for" answers who, why and on whose behalf before the person has to ask. Calls that start vaguely get treated as spam by the recipient regardless of legality.
5. Do not disguise the automation
If someone asks whether they are talking to a machine, the agent should say yes. Beyond the ethics, denying it destroys trust for the entire programme the moment one person posts a recording — and someone eventually will.
Practical settings
Most of this is configuration rather than policy:
- Set the calling window on the agent, so a batch started at midnight dials in the morning.
- Put the identification line in the welcome message, where it is spoken before anything else.
- Use a keypad opt-out on any campaign large enough to generate complaints, and make sure suppression is account-wide.
- Keep the recordings and transcripts — the record of what was actually said is your best evidence if a complaint arrives.
The blunt version
Calling people who asked to hear from you is low risk and works well. Calling people who did not is higher risk, converts worse, and is the thing that gets caller IDs blocked. The compliant approach and the effective one are the same approach.