AI calling agents for follow-ups in India: consent, DND, and calling hours

An AI calling agent in India can follow up with new enquiries, remind people about appointments, and confirm bookings, provided three things are settled first: the person has agreed to be called, DND preferences are honoured, and calls go out only within permitted hours.

This guide covers each in plain terms and shows how to build them into the agent. It describes the rules only in general terms and is not legal advice, so confirm the details with your own adviser.

What counts as a follow-up call

“Follow-up” covers several different calls, and they do not all carry the same risk. The closer a call is to something the person has just asked for, the easier it is to justify and the more welcome it tends to be.

Follow-up call	How recent is the contact	What to check first
Callback to a new enquiry	Minutes or hours	Consent captured on that lead source, calling hours
Appointment or site visit reminder	An existing booking	Calling hours, how the person prefers to be reminded
Confirmation call	An existing booking	Calling hours, an easy way to cancel or reschedule
Enquiry that went quiet	Days or weeks	Consent still valid, any opt-out since
Old enquiry list	Months or more	Consent, DND status, and whether the call is promotional

The top of the table is where most businesses should start. A callback to someone who filled in your form ten minutes ago is still a warm call. A call to a list from two years ago is, for the person receiving it, much closer to a cold one. Our guide to inbound and outbound voice agents explains why fresh enquiries are the lower-risk place to begin.

Why the rules matter more when software dials

A person working through a callback list notices things. They see that it is late, that a number has already been tried three times, or that a note says “asked not to be called”. An AI agent notices none of that unless it has been built to check.

It can also place as many calls at once as your phone lines allow, so a mistake in a list or a calling window is repeated across every call in the batch, not made once.

So the rules have to live in the system, as checks that run before every call.

It also helps to be clear about who owns what. The business placing the calls is responsible for its own compliance, which in general terms means registration where required, honouring the DND registry, calling only within permitted hours, and keeping evidence of consent. In our builds, consent capture, calling windows, and opt-out handling are configured as part of the work, not added later.

Consent: know where every number came from

In India, commercial calls and messages are regulated by TRAI under the Telecom Commercial Communications Customer Preference Regulations, and consent sits at the centre of them. In broad terms, the rules treat promotional calls differently from service and transactional ones, and the consent you hold affects which calls you may make. Which category your follow-up calls fall into is a question for your adviser.

On the practical side, list every lead source and answer four questions for each:

1. What did the person agree to? Agreeing to be contacted about an enquiry is not the same as agreeing to hear about future offers.
2. Where is that recorded? The form wording, the checkbox, the date, and the source should travel with the lead into your CRM.
3. Could you show it later? Evidence of consent is something to keep from the start, not reconstruct after a complaint.
4. Who controls the source? Leads from property portals or ad platforms need particular care: find out what the enquirer agreed to when they submitted, and note it.

Then make the agent depend on the answers: it only dials numbers whose CRM record carries a consent note from a source you have reviewed, and anything else goes to a queue for a person to check.

DND and opt-outs: honour every preference

The DND registry lets people state which commercial calls they do not want to receive. How it applies to a particular call depends on the kind of call and the consent you hold, so settle that with your adviser before any campaign.

Two practical habits make DND easier to respect:

• Check before each campaign, not once. Preferences change. A number that was fine to call last quarter may not be now.
• Treat old lists with the most care. They are where consent is weakest and DND checks matter most, and the people on them are the least likely to expect your call.

Opt-outs on the call itself

The registry is not the only preference that counts. If someone tells your agent “please don’t call me again”, that should end the matter everywhere, not just on that call.

The agent should recognise the request in English, Hindi, Gujarati, or Hinglish, confirm it politely, and tag the record. Your workflow should then remove the number from every future call and message sequence, including WhatsApp follow-ups. If the person is annoyed, the agent should offer to pass them to someone from your team, with the transcript attached.

Calling hours: enforce the window in the system

Outbound calls should only go out within permitted hours, and the agent should check the clock itself before every dial. Confirm the permitted hours for your type of call with your adviser, and write them into the agent’s configuration rather than a policy document nobody opens.

Then decide your own window, which can be narrower. Being allowed to call at a given hour does not make it a good hour to call. A clinic, for instance, might keep reminder calls to its working day.

Three settings sit alongside the window:

• What happens to leads that arrive outside it. A WhatsApp acknowledgement can go out straight away, with the call placed when the window opens.
• How many attempts, and how far apart. For example, two call attempts on different days, then a WhatsApp message instead. That is an illustration, not a rule; choose what suits your customers.
• Public holidays and festivals. Decide in advance whether follow-ups pause, and put the dates in the calendar the agent checks.

Our article on how AI voice agents qualify real estate leads shows how calling hours fit into an instant callback flow.

The first ten seconds of the call

What the agent says first matters as much as when it calls. We recommend that it says it is an AI in its opening line, on outbound calls as well as inbound ones. In our deployments, disclosure raises trust and does not reduce completion rates.

A good opening does four things: it says who is calling, says why, checks it is a good time, and offers a person. For example:

“Hello, I’m the AI assistant calling from [your business]. You asked about a quote on our website this morning. Is now a good time for two quick questions? If you’d rather speak to someone from the team, I can arrange that.”

Our guide to writing a voice agent script covers the rest of the wording.

How the checks fit into a follow-up agent

The checks work best as steps in one workflow, so nobody has to remember them. A typical flow, built on n8n, looks like this:

1. An enquiry lands from a form, portal, or ad and is logged in your CRM with its source and consent note.
2. The workflow checks consent, DND status, any earlier opt-out, and the calling window.
3. If everything passes, the agent calls. If not, the lead gets a WhatsApp acknowledgement or goes to a queue for a person.
4. The agent qualifies, books, or reminds, and transfers to a person when needed.
5. A summary, tags, recording link, and outcome are written back to the CRM, and any opt-out is applied everywhere.

Recordings and transcripts contain personal data, so decide how long you keep them and who can see them, and check your set-up against India’s Digital Personal Data Protection Act, 2023 with your adviser.

Our workflow automation work connects these steps across your CRM, telephony, and WhatsApp. On the voice side, a typical single-intent agent goes from call audit to live traffic in under three weeks, with human escalation in place from day one. Start narrow: fresh enquiries from a single lead source, inside agreed calling hours. After launch, review transcripts each week for opt-out requests and complaints, and confirm that no call went out of hours. Our AI voice agent development page shows how callbacks, reminders, and confirmations fit into the full build.

Where to start

Before you switch on any follow-up calls, write a one-page sheet: every lead source and how it captures consent, how you check DND, your calling window and retry limit, how opt-outs are handled, and the agent’s opening line. Take it to your adviser, then start with one lead source. If you want help turning that sheet into a working agent, book a free call audit.