What can an AI voice agent actually do on a clinic's phone line?

"AI receptionist" gets used for two different things: software that answers your clinic's phone, and software that answers on WhatsApp. This post is about the phone version only, what it can genuinely do today, where it still needs a person, what India's consent-to-recording rules say, and what it actually costs where a vendor has published a number.

What is an AI voice agent for a clinic phone line?

An AI voice agent picks up an incoming call the same way a person would, listens to what the caller says using speech recognition, decides how to respond, and speaks back in a synthetic voice. It is built on a clinic-specific script: timings, fees, department names, doctor availability and a booking flow, so it can answer inside that script without a person on the line. It is a different product from a WhatsApp chatbot, which never touches a phone call at all and only handles typed messages.

Vendors selling this into Indian clinics include ConnectAI, HuskyVoice, Tatkal Doctor and Patient Square, alongside general-purpose voice-AI platforms such as Bland, Vapi and Retell that are not built specifically for healthcare but get adapted for it. Most describe roughly the same core loop: answer, understand, respond, book, confirm by WhatsApp.

What can it handle today?

The routine, repeatable part of a front-desk call is what these tools are built for, and vendor pages agree closely on the list.

• OPD timings and doctor availability. Quoting when the clinic is open and which doctor is in on which day, per Tatkal Doctor's own product page (opens in a new tab).
• Consultation fees. Stating a fee already fixed by the clinic, the same information NMC 2002 already requires a clinic to display on its chamber board.
• Booking, rescheduling and cancelling a routine appointment, then sending a WhatsApp confirmation once it is done.
• Department or doctor routing, directing the caller to the right person or line when a clinic has more than one department or branch.

This overlaps almost entirely with what a WhatsApp AI receptionist does on text. The difference is the channel, not the underlying task.

What can it not handle yet?

Every vendor page that discusses limits at all agrees on the same edge: anything clinical, urgent, or outside the script goes to a person, not the AI.

Is it legal to record a patient's call in India?

Most AI voice agents record and transcribe every call, both to work and to give the clinic a log afterwards. Whether a clinic can do this without telling the caller is a real question, and the honest answer has two separate parts.

On interception law, the position has been settled for decades. In R.M. Malkani v State of Maharashtra (1972) (opens in a new tab), the Supreme Court held that a person recording a conversation they are themselves a party to is not unlawful interception, even without the other side's knowledge; this is the basis for India's widely cited "one-party consent" position. What the courts have treated as a privacy violation, in People's Union for Civil Liberties v Union of India (1996) (opens in a new tab), is a third party secretly tapping a call neither side agreed to, which is a different situation from a clinic recording its own line.

No vendor page we checked spelled out its own recording-consent line to the caller. Ask any vendor to show you, word for word, what the AI says at the start of a call and whether that line is something you can edit.

What does an AI voice agent cost in India?

Most India-focused vendors sell this on a custom quote, so a published, comparable number is rare. The table below uses only prices each vendor states on its own page.

As with WhatsApp chatbot pricing, the number on the page is rarely the full monthly bill. Telephony (the phone line itself), speech-to-text and text-to-speech processing, and any compliance layer a healthcare deployment needs are commonly billed as separate line items on top of the headline per-minute rate, whichever vendor you compare.

What is it like for the patient calling in?

Global research on healthcare AI, not specific to India, gives a useful shape to patient comfort even without an India-specific survey. A Salesforce study of 3,200 adults across eight countries found patients are three times more likely to trust an AI agent built into their own provider's system than a public chatbot, and that a clear option to reach a person is close to universal: 89% said it was essential for administrative AI and 90% said the same for anything medical.

For an Indian clinic, the practical read is the same as the survey's headline: patients seem willing to talk to a provider's own AI for routine things, as long as reaching a person is genuinely one step away, not buried behind a script that keeps looping.

A note on tone

A caller who is unwell, in pain, or anxious about a result is a common reason people call a clinic instead of messaging. A voice agent that sounds cheerful and scripted in that moment can land worse than the same words typed on WhatsApp, where the caller reads at their own pace. This is a design choice each vendor makes differently, and it is worth listening for specifically when you evaluate one.

How does it compare with WhatsApp-first replies?

Voice and WhatsApp are not competing for the same moment. A voice agent answers someone who has already picked up the phone and wants a live answer now. A WhatsApp-first reply catches a text enquiry directly, and can also follow up a call that was missed, without needing the caller to ring back.

Neither replaces the other outright. A clinic that gets most enquiries as calls has a real case for voice; one where patients already message first, or where a written record matters for follow-up, gets more immediate value from WhatsApp. Read what AI can and cannot do for a clinic for the boundary that applies to both channels, or see how an AI receptionist actually books an appointment on WhatsApp if text is the channel you are considering instead. The full picture of the category, including cost and what to ask any vendor, is in the AI receptionist for clinics guide.