AI Receptionist India: ₹2/min, No Subscription

There is no single winner, so judge on two things: whether the vendor publishes a rate, and what you have to commit to before that rate applies. On the vendors we checked on 21 Jul 2026, Dvaarik is the strongest fit if you want the system built for you and want ₹2/min to apply from the first minute with no subscription, no minimum volume and no lock-in. Trikon is a sensible pick if you would rather run it yourself on a flat ₹5/min with no setup fee and bring your own telephony. ConnectAI fits clinics already living inside its clinic suite, since the receptionist is ₹800/mo (₹499/mo bundled) plus ₹4/min. HuskyVoice suits voice-heavy teams comfortable starting on a ₹1,999/mo plan that includes 100 credits at 2 credits per minute. If a vendor will not put a number on a page — SquadStack shares its rate only after a discovery call — treat that as a cost in itself. Message us on WhatsApp and we will tell you honestly which of these fits your call volume, even when it is not ours.

Across the six receptionist vendors compared here, the per-minute rates published on their own pages on 21 Jul 2026 run from ₹0.40/min to ₹5/min; other India voice-AI vendors we checked publish higher rates still. The headline number is rarely what you pay. Trikon publishes a flat ₹5/min excluding telephony. ConnectAI publishes ₹4/min on top of a required ₹800/mo subscription. HuskyVoice's ₹4/min is enterprise volume pricing, not its entry rate. Agni's ₹2/min is a quote-only Enterprise rate at 10,000+ minutes, with ₹8/min overage on its ₹2,999/mo entry plan. Scalify Labs' ₹0.40/min carries its own disclaimer that the rates are estimates, and sits above a plan from ₹15,000/mo with a 10,000 minutes/month minimum. Dvaarik is ₹2/min from the very first minute with no subscription and no minimum volume — whole-minute billing, so a 40-second call bills as one minute. The question worth asking any vendor is not what the rate is, but what has to be true before you get it.

It depends entirely on the vendor, and it is the single biggest difference between them. Of the sixteen India voice-AI vendors we checked on 21 Jul 2026, six require a monthly plan before a single call can be placed — on Agni, Aixclerate, ConnectAI, MyOperator, Scalify Labs and VaniAgent, usage is billed on top of a recurring fee rather than instead of one, and MyOperator publishes no per-minute rate at all. Dvaarik has no subscription, no plans and no tiers: usage runs on prepaid credits at ₹2/min for voice and ₹2 per chat conversation, the minimum first credit load is ₹10,000, later top-ups start at ₹500, and credits never expire. What is not free is the build — setup is ₹40,000 one-time, currently free for our 10 founding clients in return for the Care Plan at the founding ₹5,000/month for the first three months and an honest testimonial after go-live. See the full pricing breakdown.

Answering your own inbound calls with an AI raises no special issue — the caller rang your business, and the agent identifies your business when it picks up. Outbound is where the care is needed, and it is worth asking any vendor exactly how they handle it. The agents we build call only your own leads and enquiries, never purchased or scraped lists, they respect calling hours, and the agent states which business is calling at the start. Every conversation produces a text transcript with the caller's number and outcome, so you can audit what was actually said rather than take a summary on trust. We go through the rules and what they mean for a small business in our note on whether AI calling is legal in India.

It depends almost entirely on your call volume, so here is the actual arithmetic rather than a market average. With Dvaarik, voice is ₹2/min on simple whole-minute billing, so 300 calls in a month averaging 2 minutes each is 600 minutes = ₹1,200 in voice credits. Add ₹2,000 per 30 days for the phone number (only if we supply it), and for founding clients ₹5,000/month for the Care Plan in the first 3 months, and that month lands near ₹8,200. A quiet month costs far less, because you only pay for minutes actually used and unused credits sit there and never expire. Two things to plan for at the start: the minimum first credit load is ₹10,000, and setup is ₹40,000 one-time — currently free for our 10 founding clients. See the full pricing breakdown or our AI receptionist cost guide. If the real decision is hiring a front-desk person instead of automating, we cost out both routes side by side in AI vs a human receptionist in India.

You can keep your existing number — the usual setup is call forwarding, where your current line rings as always and forwards unanswered or after-hours calls to the AI. The alternative is that we supply a dedicated number, which costs ₹2,000 per 30 days, auto-debited from your credits. Forwarding is the lower-commitment route: you can start by forwarding only busy hours and holidays and widen it once you have heard the agent handle real calls, and nothing about your printed number, boards or listings has to change. Which route suits you is something we work out when we scope your agent. Details of both are on the phone setup page.

It should hand the conversation to a human rather than guess, and that escape route is something you should test in any tool you shortlist. In the agents we build, the AI is given explicit boundaries — anything outside them (a pricing exception, a complaint, a medical or legal question) triggers a transfer to your number or a WhatsApp summary sent to you with the caller's number and what they wanted, so you can call back. The failure mode to avoid is an AI that invents an answer to sound helpful; that costs more trust than a missed call ever did.

Yes — taking several calls at once is the clearest structural advantage over a single human front desk, which can only ever hold one caller while the rest hear a ring or a busy tone. Concurrency is worth asking any vendor about directly, because some meter it and some do not. With Dvaarik there is no published per-line add-on: the telephony cost is ₹2,000 per 30 days for a number we supply, auto-debited from credits, and the capacity your agent needs is part of what we scope before we build. Work out your peak, not your average, and tell us that number: a clinic taking calls at a steady pace all day is a different shape of problem from a restaurant that gets five calls in the same two minutes at 8pm on Saturday.

If your call volume is genuinely low and you already answer nearly every call yourself, an AI receptionist is not worth it yet — the recurring costs (₹2,000 per 30 days for the number, if we supply one, and ₹5,000/month for the Care Plan in the founding first 3 months) have to be earned back by calls you are currently losing, and a handful of calls will not do it. It also fits badly when almost every enquiry needs a genuine expert judgement call from you personally, or when your prices and offers change constantly and nobody will keep the agent updated. The honest test is your own numbers, not a rule of thumb: count the calls you missed last month and multiply by your average order value with the missed-call loss calculator. If the loss does not comfortably exceed what you would spend, wait — and tell us so, we would rather say so now than build you something you do not need.

Days, not months — provided you can describe your business clearly and get your phone or WhatsApp access sorted. The bottleneck is almost never the AI: it is you gathering your service list, prices, timings, address, and the ten questions customers ask most. Dvaarik works to a scope-and-quote within a day and a working agent in days after that, because the founder builds and tunes each one personally rather than handing you a dashboard and a blank prompt box to fill in yourself. If you are comparing us with a self-serve platform, the question to ask is not how fast you can switch it on, but who does the configuration and the tuning afterwards, and how long that takes you — see how the build works.

Code-mixed speech is the real test — a customer saying "bhaiya do bedroom flat ka rate kya hai" is completely normal in India, and it is worth checking how any tool built for a different market copes with it. The agents we build handle all 22 scheduled Indian languages plus Hinglish, Tenglish and Tanglish, along with most world languages, so a caller who mixes English into Hindi or Telugu is handled as ordinary speech rather than as an error. Background noise is a genuine limitation for everyone: a call from a busy street or a workshop floor degrades accuracy for any voice AI, and for human receptionists too, which is why a clean fallback to a WhatsApp message matters. Before you commit to any vendor, call their demo line yourself from a noisy place in your worst-case language — see multilingual handling.

Insist on transcripts before you buy — an AI receptionist you cannot audit is a liability, not an employee. Every conversation the agents we build handle produces a text transcript with the caller's number and the outcome, so you can read exactly what was said rather than take a summary on trust. Care Plan clients also get a monthly report, and we change the agent's behaviour when the transcripts show it getting something wrong. If you want audio recording as well as transcripts, raise it when we scope your build and we will tell you plainly what is possible. Also confirm how any vendor handles caller consent and calling hours for outbound calling — we call only your own leads and enquiries, never purchased lists; the rules are covered in our note on whether AI calling is legal in India.