An AI receptionist for a multi-speciality hospital in India costs, with us, from ₹2 per billed minute of talk time — from the very first minute, with no subscription, no minimum volume and no lock-in. Add ₹2,000 per 30 days if we supply the phone number and its first line, and ₹1,500 per 30 days for each additional concurrent line, because a hospital's 9am OPD rush is a concurrency problem before it is a minutes problem. A 60-bed, six-speciality hospital handling roughly 1,800 inbound calls and 800 outbound reminder calls a month — about 5,700 billed minutes on three lines — spends ₹16,400 a month once the founding Care Plan period ends, and ₹21,400 a month during it. The full arithmetic, including what leaves the credit balance in month one, is worked out below. We are publishing that number at the top because the pages ranking for this query on 26 Jul 2026 do not. We read all three while writing this. HuskyVoice's ranking hospital page publishes a single number, a ₹1,999/month entry plan, and its healthcare page carried no rate at all when we read it; their own pricing FAQ says pricing depends on call volume, workflows, integrations and setup, and huskyvoice.ai/pricing returns Page Not Found (the ₹1,999, the FAQ wording and the dead pricing page are all recorded in our price ledger, checked 21 Jul 2026). Caller Digital's hospital guide gives a market-wide per-minute range rather than a rate of its own, and settles for a relative conclusion — a 1,200-appointment clinic paying less than one receptionist's loaded salary — with no rupee total attached. Tatkal Doctor's clinic-and-hospital page carried no price of any kind when we read it. None of the three shows a hospital what a month actually costs. A hospital is not a large clinic, and this page is not a rewritten clinic page. The call mix is different: report-status calls from the lab and radiology, "is Dr Menon sitting today", TPA and cashless empanelment questions, room tariffs, admission and discharge queries, health-check package bookings, and the calls that must never touch an AI at all — casualty, ambulance and anything that sounds like an emergency. Those get their own sections, including the one that governs everything else here. Said plainly before any pricing: the agent books, routes, confirms and informs. It never triages, never assesses urgency, never reads out a lab or radiology result, and never discusses medication. That is our own hard rule and it does not depend on anyone's reading of the law. It also tracks India's Telemedicine Practice Guidelines of 2020, appended to the Indian Medical Council (Professional Conduct, Etiquette and Ethics) Regulations, 2002, which as we understand them do not permit an artificial-intelligence or machine-learning platform to counsel or prescribe — the final advice has to come from a registered medical practitioner. We could not retrieve the ministry's own text while writing this, so treat that as our reading rather than as legal advice, and see the boundary section for how we source it.
What does an AI receptionist cost a multi-speciality hospital, line by line?
Every rupee that can leave your account. These are our own published figures, and there is nothing else underneath them.
| Line item | Amount | When it applies |
|---|---|---|
| Voice talk time | from ₹2 per billed minute | Bharat Standard — 22 scheduled Indian languages plus English — is ₹2/min, our permanent base rate and not a founding discount. Studio HD ₹3/min. Global voices for non-Indian languages ₹4–5/min |
| Billing granularity | Whole minutes | A 40-second "is the doctor sitting today" call bills 1 minute. A 2 min 20 s booking bills 3 |
| Chat conversations | ₹2 per conversation (24h window) | Website chat, WhatsApp and Instagram text handling |
| Phone number + 1 concurrent line | ₹2,000 per 30 days | Only if we supply the number. Auto-debited from your credit balance |
| Each extra concurrent line | ₹1,500 per 30 days | The line that matters most for a hospital — see the concurrency section |
| Each extra phone number | ₹750 per 30 days | A separate health-check or international-patient line, for example |
| Call recording | +₹0.10 per minute | Optional, only if you switch it on |
| One-time custom build | ₹40,000 list — free for our 10 founding clients | The trade is stated plainly: the Care Plan at the founding rate for the first 3 months, and an honest testimonial plus a short case study after go-live |
| Care Plan | ₹12,000/mo list · ₹5,000/mo founding rate, locked 6 months | Required for the first 3 months of a founding build, optional from month 4. Debited from the same credit balance as your calls |
| Minimum first credit load | ₹10,000 | The floor, not a hospital's month-1 budget — see below |
| Later top-ups | From ₹500 | Credits never expire |
| WhatsApp message fees | Billed by Meta directly to your own account | We add zero markup |
| GST | 18%, applied when you load credits | Prices above are exclusive of GST. A ₹25,000 load is billed ₹29,500 — ₹25,000 usable plus ₹4,500 GST |
What is not on that list
- No monthly platform or SaaS subscription.
- No per-seat, per-department or per-doctor fee.
- No annual contract. Credits never expire, and at zero balance the agent pauses gracefully rather than failing mid-call.
- No charge to add another agent type later. The seven agents — Receptionist, Leads, Sales, WhatsApp, Instagram, Booking, Payments — draw on one shared credit balance.
The month-one reality for a hospital, stated honestly
The ₹10,000 minimum first load is a floor designed for a salon or a single-chair clinic. A hospital will burn through it in about a week, and we would rather say so now than have you discover it. Here is what the minimum load actually does at hospital volume:
| Month-1 debit from a ₹10,000 load | Amount |
|---|---|
| Phone number + 1 line, 30 days | ₹2,000 |
| Care Plan, founding rate | ₹5,000 |
| Left for actual talk time | ≈₹3,000 = about 1,500 billed minutes |
At 5,700 billed minutes a month, 1,500 minutes is about eight days at that run rate. So for a 60-bed hospital the sensible first load is ₹25,000, billed ₹29,500 with GST, which covers the number, two extra lines, the founding Care Plan and roughly 7,500 minutes of talk time — about six weeks at Hospital A's volume below. For a 150-bed hospital, ₹50,000 (billed ₹59,000). Top up from ₹500 whenever you choose.
Note the Care Plan point specifically, because it is the one buyers miss: during the founding first three months the ₹5,000/month Care Plan is debited from the same credit balance as your calls, not invoiced separately. It is not an extra bill; it is ₹5,000 less talk time each month unless you load for it.
Which India AI receptionist vendors publish a rate, and what does it take to get it?
This is the table the ranking pages do not have. Every figure below was read on the vendor's own page and recorded with the sentence it came from, checked 21 Jul 2026. Nothing here comes from a search snippet, an AI summary or a rival's comparison page — each of those produced at least one false number when we tried them.
| Vendor | Published per-minute rate | What it takes to get that rate | Checked |
|---|---|---|---|
| Dvaarik | ₹2/min (Bharat Standard) | Nothing. First minute, no subscription, no minimum volume, no lock-in. ₹2,000/30 days if we supply the number and its first line; ₹1,500/30 days per extra concurrent line | 21 Jul 2026 |
| Agni by Ravan.ai | ₹2/min headline, "all-in from ₹2/min" | ₹2/min appears on the quote-only Enterprise tier, qualified on their own homepage as the "Enterprise volume rate at 10,000+ minutes". The buyable published tiers are ₹2,999/mo (300 min, ₹8/min overage), ₹5,999/mo (1,000 min, ₹6/min), ₹12,999/mo (2,500 min, ₹4/min). Twilio telephony is passed through separately at ~₹0.71/min inbound, ~₹1.08/min outbound. Starter is Hindi and English only. Prices exclude 18% GST | 21 Jul 2026 |
| HuskyVoice.AI | ₹4/min | Published strictly as "Enterprise volume pricing as low as ₹4/min" — an at-volume floor, not an entry rate. The buyable entry is ₹1,999/mo including 1 Indian number and 100 voice credits at 2 credits = 1 minute. There is no canonical pricing page; huskyvoice.ai/pricing returns Page Not Found, and their own pricing FAQ says pricing depends on call volume, workflows, integrations and setup | 21 Jul 2026 |
| Scalify Labs | ₹0.40/min | A self-labelled estimate — their own page states "Rates are estimates based on publicly available information and direct conversations. Always get a formal quote." It sits on top of a mandatory monthly plan from ₹15,000/mo, with ₹15,000–₹40,000 one-time setup, a stated minimum engagement of 10,000 minutes/month and a 3-month initial commitment. Voice minutes and telecom are billed separately from the plan. Hindi and English are standard | 21 Jul 2026 |
| Trikon | ₹5/min flat | No subscription and no setup fee, and they state the rate covers platform, LLM, STT, TTS, recording, dashboard and transfer. Telephony and DID numbers are billed separately by your own carrier — you bring your own | 21 Jul 2026 |
| Bolna AI | 6.00¢/min, which their page renders as ₹5.52/min | No subscription; prepaid credits from $10. The ₹ figure is Bolna's own conversion at a hard-coded 92 INR/USD, not an independently priced India rate. Their docs state total call cost = voice processing + telephony + "a flat per-minute fee charged by Bolna on top of your provider costs" — and that platform fee amount is not published | 21 Jul 2026 |
| ConnectAI | ₹4/min (8 credits/min) | Requires a subscription: ₹800/mo standalone, or ₹499/mo as an add-on to their ₹2,499/mo clinic suite. One-time ₹1,000 onboarding for new clinics. Pricing is genuinely public with no demo gate. Their own comparison table claims Hindi and Indian English only | 21 Jul 2026 |
| Edesy | ₹4–6/min | Their pricing page publishes "₹4-6 Per Minute". Pay As You Go is ₹6/min plus telephony at $0.07/min, with a ₹500 minimum recharge and no monthly commitment; ₹4/min is the Ultra rate and needs a ₹14,999/mo plan including 3,500 minutes. A ₹20,000 onboarding package is also published, converting to ₹20,000 of credits. Their own comparison tool quotes a different ₹1.50/min for themselves — a figure that appears nowhere on their pricing page, so we do not use it | 21 Jul 2026 |
| Aixclerate | ₹7/min | An overage rate, not a standalone rate — it applies beyond the bundle on plans of ₹9,999/mo (500 min) or ₹24,999/mo (2,000 min). 1 Indian business number included; extra number ₹699, extra concurrency ₹499, recording ₹0.20/min | 21 Jul 2026 |
| Botsense | ₹9/min · ₹7/min · ₹5/min | Three open per-minute tiers with "No monthly commitment" and "no setup fees" on their AI-calling page. Starter caps at 1,000 min/mo and is Hindi plus English only; Growth caps at 10,000 min/mo; the ₹5/min Enterprise tier routes to a custom quote. Every CTA on the page goes to WhatsApp or sales, not a checkout | 21 Jul 2026 |
| MyOperator | No per-minute rate published anywhere | ₹10,000/mo for the standalone AI voice agent including 2,000 minutes; extra usage is billed "₹8 per conversation", not per minute. ₹20,000 one-time dedicated onboarding, extra users ₹2,000 each, all plans billed annually, GST applicable | 21 Jul 2026 |
| SquadStack | No rate published | Their pricing page markets "per-minute pricing with no seat fees, no platform charges" but publishes no figure: "The per-minute rate depends on language mix, call complexity, integration depth, and the engagement scale... pricing is shared after the discovery call." A one-time setup fee applies per use-case, amount not stated. Entry tier carries a 90-day commitment | 21 Jul 2026 |
Read the middle column, not the left one. Six of the sixteen vendors we checked require a subscription to place a single call. Two of the three lowest headline rates in this table — Agni's ₹2 and Scalify's ₹0.40 — are not buyable at the number shown: one is a quote-only Enterprise tier at 10,000+ minutes, the other is a self-declared estimate layered on a ₹15,000/month plan with a three-month commitment. That is the honest shape of this market, and the reason our claim is about terms rather than about being lowest.
Numbers we found and refused to publish
While researching this page, search results and AI answers offered us a Botsense plan price of ₹9,999/month, a Bolna rate of ₹8–₹11/min, and various hospital ROI multiples. The ₹9,999 string appears zero times on Botsense's own site — we counted it, twice, across both of their pricing pages. The ₹8–₹11 Bolna figure comes from a rival's comparison page and contradicts Bolna's own published ₹5.52/min — yet it is the number Google's own AI answer repeats, because it is the version that got written down in a citeable shape. Scalify's blog also publishes estimated rates for several other platforms, Yellow.ai, Ozonetel, Knowlarity and Vapi among them; those are one vendor's guesses about others and are not in this table. The hospital ROI multiples could not be traced to a named hospital or a primary source, so they appear nowhere on this page. If a vendor quotes you a competitor's price, ask to see it on the competitor's own page.
What does 5,700 minutes a month actually cost at each vendor?
The table above is what vendors publish. This one is what a 60-bed hospital would pay, at 5,700 billed minutes a month with three concurrent lines. The rupee totals in this table are our arithmetic applied to the vendors' own published numbers, not quotes from them — the inputs are shown in every row so you can check the sum yourself. All Dvaarik figures exclude GST; Agni and MyOperator publish that their prices exclude GST too, while several vendors publish nothing about GST either way, so check before you compare.
| Vendor | The sum, from their published figures | Monthly total at 5,700 min | Per billed minute |
|---|---|---|---|
| Dvaarik (month 4 onward) | 5,700 × ₹2 = ₹11,400, plus ₹2,000 number and 2 extra lines at ₹1,500 | ₹16,400 | ₹2.88 |
| Dvaarik (founding months 1–3) | The above plus ₹5,000 Care Plan, debited from the same credits | ₹21,400 | ₹3.75 |
| Scalify Labs | ₹15,000 plan + (5,700 × ₹0.40 estimate) ₹2,280 + amortised setup ₹2,500–₹3,500 — and note 5,700 sits below their stated 10,000-minute minimum | ₹19,780–₹20,780 | ₹3.47–₹3.65 |
| ConnectAI | ₹800/mo + (5,700 × ₹4) ₹22,800 | ₹23,600 | ₹4.14 |
| Edesy (Ultra plan) | ₹14,999 incl. 3,500 min + (2,200 × ₹4) ₹8,800, before telephony | ₹23,799 + telephony | ₹4.18 + telephony |
| Agni Scale | ₹12,999 incl. 2,500 min + (3,200 × ₹4) ₹12,800, before Twilio pass-through | ₹25,799 + telephony | ₹4.53 + telephony |
| Trikon | 5,700 × ₹5, telephony billed by your own carrier on top | ₹28,500 + telephony | ₹5.00 + telephony |
| Bolna AI | 5,700 × ₹5.52 (their own conversion), plus provider, telephony and an unpublished platform fee | ₹31,464 + unknowns | ₹5.52 + unknowns |
| Edesy (Pay As You Go) | 5,700 × ₹6, plus telephony at $0.07/min | ₹34,200 + telephony | ₹6.00 + telephony |
| Botsense Growth | 5,700 × ₹7 | ₹39,900 | ₹7.00 |
| Aixclerate Business | ₹24,999 incl. 2,000 min + (3,700 × ₹7) ₹25,900 | ₹50,899 | ₹8.93 |
| MyOperator | ₹10,000 incl. 2,000 min; the 3,700 overage minutes cannot be priced because overage is published per conversation (₹8), not per minute | Not computable | Not computable |
| HuskyVoice.AI | ₹1,999 includes 100 voice credits at 2 credits = 1 minute, which is 50 billed minutes by our arithmetic on their inputs; the overage rate for the remaining minutes is not published | Not computable | Not computable |
| SquadStack | No rate published | Not computable | Not computable |
Four things this table does not say
It does not say we are cheapest. Look at the Scalify row. If their ₹0.40/min estimate holds at their own 10,000-minute minimum, they land below us per minute — and above 10,000 minutes a month, a volume-priced vendor will beat our flat ₹2. What their row also carries is a ₹15,000/month plan you pay whether or not you use it, ₹15,000–₹40,000 of setup, a three-month commitment, minutes billed separately from the plan, and their own instruction to get a formal quote because the rate is an estimate. The honest claim we will make is about terms, not about the number: ₹2/min from the first minute, no subscription, no minimum volume, no lock-in.
It does not include your telephony everywhere. Our ₹2,000 per 30 days is the number and one line, stated as a separate line item rather than buried in a per-minute rate. Trikon, Bolna and Edesy bill telephony or DID separately at amounts that depend on your own carrier and that we cannot compute for you; Agni publishes a Twilio pass-through of roughly ₹0.71/min inbound and ₹1.08/min outbound on top of its plan. HuskyVoice and Aixclerate include a number in the plan. Compare like with like.
Our own rate gets worse at low volume, not better. At 5,700 minutes the fixed ₹5,000 of number and lines works out to ₹0.88/min on top of ₹2. At 800 minutes a month it would be ₹6.25/min on top — a blended ₹8.25 — and you should be sceptical of us at that volume exactly as you should be of everyone else.
Whole-minute rounding is real. A hospital's call mix is full of 30-to-45-second calls: visiting hours, is the doctor in, where is the OPD block. Every one of those bills a full minute. If your average call is 45 seconds, whole-minute billing costs you roughly a third more than per-second billing would. We do not offer per-second billing and do not claim to.
Three worked hospital scenarios, with the founding months shown separately
All three hospitals are invented and sized to be realistic rather than flattering. Minutes are billed minutes, meaning every call is already rounded up to the next whole minute. All talk time is at the ₹2/min Bharat Standard rate.
Hospital A — 60 beds, six specialities, one location
OPD-led, with a small in-house lab. Sixty inbound calls a day (1,800 a month) averaging 2.5 billed minutes, plus 800 outbound calls a month — appointment reminders, report-ready notifications, health-check recalls — averaging 1.5 billed minutes. Three concurrent lines so the 9am booking rush does not queue.
| Line | Founding months 1–3 | Month 4 onward |
|---|---|---|
| Talk time (5,700 billed min × ₹2) | ₹11,400 | ₹11,400 |
| Number + 1 concurrent line | ₹2,000 | ₹2,000 |
| 2 extra concurrent lines (₹1,500 each) | ₹3,000 | ₹3,000 |
| Care Plan, founding rate, from the same credits | ₹5,000 | ₹0 if you drop it |
| Monthly total | ₹21,400 | ₹16,400 |
| Blended cost per billed minute | ₹3.75 | ₹2.88 |
Suggested first credit load: ₹25,000, billed ₹29,500 with GST. That covers the ₹21,400 first month and leaves ₹3,600 rolling into month two.
Hospital B — 150 beds, twelve specialities, diagnostics and a health-check unit
150 inbound calls a day (4,500 a month) at 2.5 billed minutes, plus 2,500 outbound calls at 1.5 billed minutes. Fifteen thousand billed minutes. Five concurrent lines, and a second published number for the health-check and corporate-package desk.
| Line | Founding months 1–3 | Month 4 onward |
|---|---|---|
| Talk time (15,000 billed min × ₹2) | ₹30,000 | ₹30,000 |
| Number + 1 concurrent line | ₹2,000 | ₹2,000 |
| 4 extra concurrent lines | ₹6,000 | ₹6,000 |
| Second phone number | ₹750 | ₹750 |
| Care Plan, founding rate | ₹5,000 | ₹0 if dropped |
| Monthly total | ₹43,750 | ₹38,750 |
| Blended cost per billed minute | ₹2.92 | ₹2.58 |
The fixed component has almost disappeared here — ₹8,750 of numbers and lines across 15,000 minutes is ₹0.58/min. This is the volume band where a pay-as-you-go model is at its strongest against a bundled plan. For reference, the same 15,000 minutes on Aixclerate Business, from their published figures, would be ₹24,999 plus 13,000 overage minutes at ₹7 = ₹1,15,999 (our arithmetic, their numbers, checked 21 Jul 2026). On Botsense Growth at ₹7/min it would be ₹1,05,000, and 15,000 minutes exceeds Growth's published 10,000-minute cap.
Suggested first credit load: ₹50,000, billed ₹59,000 with GST — that covers the ₹43,750 first month and leaves ₹6,250 rolling forward.
Hospital C — 500-bed chain, four cities, running a full HIS
Central call centre, twenty-plus simultaneous calls at peak, an existing hospital information system that appointments must be written into, a procurement process with a security questionnaire, and an uptime SLA with penalties.
We would decline this, or take a single department at a single unit as a pilot. The reasons are in the section on where we are the wrong choice, and they are not marketing modesty. A 500-bed chain should be buying from a company with a support rota, not from one engineer.
Run your own numbers against your current missed-call volume with the missed call calculator, or see the full pricing page.
What a hospital front desk actually gets asked — and which of it the agent handles
Described mechanically, with no performance claims attached. All hospital and doctor names below are invented.
1. OPD appointment booking, routed by speciality
This is the call type most hospitals want handled first, and it is structurally harder than a clinic booking, because "I want an appointment" is not an appointment — it is a routing decision across departments, consultants and sitting hours that change by day. The agent asks what the appointment is for in the caller's own words, maps it to the department you have configured, offers real open slots for the right consultant, locks the slot while it confirms details, and sends a WhatsApp confirmation.
> Caller: My mother has knee pain, I want to show a doctor. > Agent: This is Anandi Multispeciality. For knee pain that would be Orthopaedics. Dr Rekha Iyer has 11:20am tomorrow or 5:40pm on Thursday. Which suits you better.
What it does not do is decide which department is clinically correct in an ambiguous case. Where your rules do not cleanly map a complaint to a department, it routes to a human rather than guessing.
2. "Is Dr Menon sitting today"
This question arrives all day, and answering it repeatedly is rarely the best use of a front-desk person standing in front of a queue. The agent answers it from a consultant schedule you keep updated, including leave and OT days, and offers the next available slot in the same breath rather than leaving the caller to call again. Pull your own call log before you buy anything and see what share of your inbound calls this one question is.
3. Lab and radiology report-status calls
Another large bucket, and the one with the sharpest boundary. The agent tells a caller whether a report is ready, when it is expected, and where or how to collect it — and can send the collection link on WhatsApp. It does not read out values, does not say whether a result is normal, and does not interpret anything. If a caller asks what the result means, it says it cannot discuss results and routes to the department. That refusal wording is something you sign off on before go-live.
4. TPA, cashless and insurance queries
"Do you accept my insurance" is a call your billing desk answers over and over. The agent reads from the empanelment list you supply — which insurers and TPAs, which government schemes, what documents to bring, pre-authorisation desk timings — and routes anything about a specific claim, approval status or amount to a human. It does not estimate what will be covered.
5. Room tariffs, admission and package prices
The agent quotes only what you authorise it to quote: published room categories, fixed health-check package prices, standard consultation fees. Anything variable — a surgical package, an ICU estimate, a treatment plan — goes to the billing or admissions desk with a callback logged. Getting this wrong is expensive, so the default is to route rather than quote.
6. Discharge, billing and attendant queries
Visiting hours, ICU visiting policy, attendant passes, what documents a discharge needs, where the pharmacy is, parking. Low value individually, large in aggregate, and available at 11pm when nobody is at the desk.
7. Outbound, to your own patients only
Appointment reminders the day before. Report-ready calls. Health-check package recalls at six and twelve months. Rebooking after a no-show. Scheduling the post-discharge follow-up OPD visit — the scheduling of it, never anything clinical about it. Every outbound call identifies your hospital by name, runs only inside the calling hours you set, and stops permanently for anyone who says stop.
We call only your own patients and enquiries. No purchased lists, no scraped numbers, no databases, ever. See outbound calls for how campaigns are structured.
8. Language, the way district patients actually speak
A city hospital's callers are not all city callers. The agent handles the 22 scheduled Indian languages plus code-mixed Hinglish, Tenglish and Tanglish, switching mid-sentence rather than asking anyone to press a key for a language — which is precisely where a keypad IVR loses an elderly caller from a district. Indian languages and English are on the ₹2/min rate; a Global voice for a non-Indian language is ₹4–5/min. More at multilingual and AI voice agent vs IVR.
9. Deposits and payment links
For health-check packages and elective procedures where a no-show costs a slot, the agent can send a payment link mid-conversation and confirm on receipt. We take no commission on what you collect. See payments.
The emergency call and the lab result: where the agent stops
This is the section a hospital should read before the pricing section, and it is the reason we would rather write it plainly than sell around it.
The rule first, because it is ours and it stands on its own: the agent never triages, never assesses urgency, never reads out a lab or radiology result, and never discusses medication. We would hold that line for a hospital even if nothing in Indian law required it.
It also matches India's Telemedicine Practice Guidelines of 2020, appended to the Indian Medical Council (Professional Conduct, Etiquette and Ethics) Regulations, 2002, which as we read them do not permit artificial-intelligence or machine-learning platforms to counsel or prescribe, and require the final advice to come from a registered medical practitioner.
A note on sourcing, because this page is otherwise strict about it: the ministry's own text was not retrievable while we were writing, so we are describing the guidelines rather than quoting them, and we are not citing a section number or a secondary paper we did not read end to end. Confirm the current wording with your own legal adviser before relying on it. We are not lawyers, and regulations change. Nothing about our own boundary depends on how that check comes out.
The emergency line should not touch the agent at all
Our first recommendation to any hospital is architectural rather than commercial: publish your casualty and ambulance number as a separate DID that never routes through the agent. No software should sit between a person in crisis and your emergency desk, ours included. That is free advice and it costs us minutes.
On the main line, the agent applies a routing rule, not a clinical assessment. It listens for urgency language — chest pain, breathlessness, bleeding, unconsciousness, accident, seizure, a fall in an elderly patient — and on hearing it stops trying to book anything and escalates immediately.
| What the agent does | What the agent never does |
|---|---|
| Recognises urgency language and transfers at once to casualty | Grades severity or decides how urgent something is |
| States plainly that it cannot give medical advice | Advises whether to come in, wait, or go elsewhere |
| Gives your casualty number and ambulance number verbatim | Suggests medication, dosage or first aid |
| Takes a callback number and a verbatim note if no human answers | Explains what might be wrong or what caused it |
| Alerts your configured staff immediately with the transcript | Reassures a caller that something is not serious |
| Confirms whether a report is ready and where to collect it | Reads out a lab or radiology value, or says whether it is normal |
| Routes result questions to the department | Interprets a result, compares it to a range, or discusses next steps |
> Caller: My father is having chest pain since one hour, what should we do. > Agent: I am not able to give any medical advice. Our casualty number is [x] and our ambulance number is [y] — I am connecting you to casualty right now. If the call does not connect, I have your number and someone will call you back immediately.
You decide the escalation numbers, the after-hours rule and the exact wording of every refusal. What is not configurable is whether the agent can triage, advise or read results. It cannot — for any hospital, under any configuration, at any price.
Patient data, and what we will not claim
The list is yours; you load it, and we do not sell, share or reuse it. Call recording is optional and priced at +₹0.10/min only if you enable it — if you do not, calls are not recorded. Transcripts and outcomes are available to you.
On the DPDP Act 2023: some competitor pages cite section numbers at you. We could not retrieve the official text while writing this, so we are not going to quote sections we did not read. The operational position is that consent for calling and messaging sits with the hospital as the party holding the patient relationship. Take your own legal advice on consent, retention and data residency before running an outbound campaign at scale, and ask any vendor reciting the Act whether their architecture actually implements what they are quoting. We make no certification, residency or retention-window claim on this page. If those are procurement requirements, ask us at contact and we will answer honestly, including where the answer is no.
Why concurrency, not minutes, is a hospital's real cost driver
A clinic's calls arrive in a trickle. A hospital's arrive in waves, and the wave is the whole problem.
Ask your own front desk and you will usually hear the same shape: OPD booking calls clustering through the morning, report-status calls clustering in the late afternoon once the lab publishes, discharge and billing calls clustering around late morning, and quiet lines outside those windows. Check it against your own call log rather than taking our word for it — the pattern is the thing that decides how many lines you buy. A hospital that averages 60 calls a day is not receiving 2.5 calls an hour; it may be taking a dozen or more in one hour and none in the next.
This matters to your bill because minutes are elastic and lines are not. Talk time costs what it costs whenever it happens. But a call arriving when every line is busy does not cost you ₹2 — it costs you the patient, exactly as it does today when your front desk has two handsets and eleven people calling.
| Concurrent lines | Cost per 30 days | Roughly what it absorbs |
|---|---|---|
| 1 (included with the number) | ₹2,000 | A small nursing home outside peak, or an after-hours-only deployment |
| 3 | ₹5,000 | A 60-bed hospital with a defined 8–11am booking rush |
| 5 | ₹8,000 | A 150-bed hospital with diagnostics and a health-check desk |
| 20+ | Not our model | This is call-centre territory — buy from a vendor built for it |
Two things follow from that table. First, size your lines to your peak, not your average, and expect the number-and-line component to be ₹5,000–₹8,000 a month for a real hospital rather than the ₹2,000 a single-clinic page would quote you. Second, the fixed component stops mattering as volume rises: it is ₹0.88 per billed minute at 5,700 minutes and ₹0.58 at 15,000.
There is also a deployment shape worth considering before you buy anything. We would often suggest putting the agent on overflow and after-hours only — your desk answers first, and the agent picks up what rolls over and everything from 8pm to 8am. Your minutes drop, your line count drops, and your staff keep the calls where a human genuinely helps. It is a smaller sale for us and often the right first step.
Where we are the wrong choice for a hospital
Five situations where another vendor genuinely serves you better. This section exists because the rest of the page is only credible if this one is honest.
1. You run a hospital information system and appointments must land inside it
This is our largest real gap and it is a bigger gap for hospitals than for clinics. We have no HIS or HMIS integration. Our booking is calendar-based with slot-locking, plus WhatsApp confirmations. That fits hospitals whose OPD scheduling still lives in a register, a spreadsheet or a shared calendar. It fits badly if your registration desk works inside an HIS all day, because a clerk retyping our bookings into your system is work we created, not work we removed. If that is you, buy from a vendor with a native integration to your specific system, and ask them to name it.
2. You need twenty-plus simultaneous calls, an SLA and a support rota
Our model is a number plus lines at ₹1,500 each per 30 days. That scales to a handful of lines, not to a central call centre with penalty-backed uptime commitments. MyOperator sells enterprise AI tiers with managed onboarding and account management; SquadStack runs dedicated pods for high-volume programmes. If your procurement needs an SLA, a security questionnaire answered and a named support tier at 3am, buy from a company that has those things. We are one engineer.
3. Your decision requires proof it worked at a comparable hospital
We have no hospital case study and no published outcome numbers, because our founding seats are still open and we have a hard rule against publishing proof we do not have. Competitor pages in this category routinely show ROI multiples, no-show reduction percentages and answer-rate figures. We traced several of them while writing this page and could not find a primary source for any. Our only counter is that you can call the live demo yourself and judge the thing directly instead of judging a screenshot. If your board needs a reference site, we cannot give you one today, and we would rather you heard it here.
4. Finance wants one flat annual line item
A ₹10,000-a-month plan billed annually is easier to put through a hospital's approval process than prepaid credits with a variable monthly draw and a number recharge. Our model often comes out lower on total cost at the hospital volumes worked above — the arithmetic shows where — but it loses on predictability, and predictability has genuine value in an institution. If your finance team wants a fixed number, a flat plan is a legitimate choice even at a higher effective rate.
5. You want the agent to do anything clinical
If what you actually want is symptom triage, a nurse-line replacement, or an agent that reads results to patients, we are not just the wrong vendor — you should be sceptical of any vendor who says yes. See the boundary section above.
Twelve questions to ask any vendor selling an AI receptionist to a hospital
Take this list to us and to everyone else. The answers separate vendors faster than a demo does.
- What is your per-minute rate in rupees, on a page I can read without booking a demo. If the answer is "it depends on your volume", note that they wrote a pricing page without a price on it.
- Is that rate buyable today, or is it a volume tier? Ask specifically whether the headline number sits on an Enterprise tier or behind a minimum-minutes commitment.
- Is billing per second or per whole minute? If your average call runs around 45 seconds, that difference is worth roughly a third of your voice bill. Ours is whole-minute, and we say so.
- How many calls can it answer simultaneously, and what does each additional line cost? Then ask what happens to call eleven at 9:15am.
- What do I pay in month one before a single call is answered — setup, plan, number, minimum load, onboarding, and whether GST sits on top.
- Is telephony included in the per-minute rate, or billed separately by my own carrier?
- Read me the exact words the agent says when a caller describes chest pain at 3am. Ask for the literal script, not a description of it.
- Will it ever read out a lab or radiology result? The only acceptable answer is no.
- Does it write appointments into our HIS, or into a calendar? Make them name your system.
- Where does the WhatsApp message cost land — on your invoice with a markup, or on our own Meta account?
- Every statistic on your website — which hospital, over what period, measured how. If it cannot be sourced, discount the page. Apply this to us.
- Will you ever call numbers I did not give you? The only acceptable answer is no.
Related reading: AI receptionist for clinics in India, dental clinic AI receptionist cost, AI receptionist vs call centre cost, and the clinics overview.
Tell us your problem on WhatsApp at +91 93923 98750 and describe how your hospital's calls are handled today — which lines, which hours, and what your desk drops at 9am. If an agent fits, we will say so and quote the numbers on this page. If an HIS-integrated vendor suits you better, we will say that too.
Frequently asked questions
What will an AI receptionist actually cost my hospital every month?
With Dvaarik, voice talk time is from ₹2 per billed minute with no subscription and no minimum volume, plus ₹2,000 per 30 days if we supply the phone number and its first line and ₹1,500 per 30 days for each extra concurrent line. A 60-bed hospital running about 5,700 billed minutes a month across three lines pays ₹16,400 a month from month four, or ₹21,400 during the founding first three months when the ₹5,000 Care Plan is debited from the same credit balance. A 150-bed hospital at 15,000 billed minutes on five lines and two numbers pays ₹38,750 a month. All prices exclude GST; 18% is applied when you load credits, so a ₹25,000 load is billed ₹29,500. The ₹40,000 one-time build is free for our 10 founding clients, in return for the Care Plan at the founding rate for the first three months and an honest testimonial plus a short case study after go-live.
Can it handle the 9am OPD rush when ten people call at once?
Only if you buy the lines for it. Each concurrent line costs ₹1,500 per 30 days on top of the ₹2,000 that covers the number and the first line, so three lines is ₹5,000 a month and five is ₹8,000. Size your lines to your peak hour, not your daily average, because hospital calls typically arrive in waves — booking through the morning, report calls in the late afternoon. Check the shape against your own call log before you decide. If you genuinely need twenty-plus simultaneous calls with an uptime SLA, we are the wrong vendor and you should buy from a company built for call-centre scale. A common middle path is to put the agent on overflow and after-hours only, which cuts both your minutes and your line count.
What happens if someone calls with chest pain at 3am?
Two things. First, our standing recommendation is that your casualty and ambulance number should be a separate line that never routes through the agent at all — no software should sit between a person in crisis and your emergency desk, ours included. Second, on the main line the agent applies a routing rule, not a clinical assessment: it recognises urgency language, states plainly that it cannot give medical advice, reads out your casualty and ambulance numbers, transfers immediately, and if no human answers it takes a callback number and alerts your staff with the transcript. It never grades severity, never advises whether to come in or wait, and never mentions medication. That is our own hard rule; it also matches India's Telemedicine Practice Guidelines of 2020 as we read them, which do not permit AI or machine-learning platforms to counsel or prescribe and require the final advice to come from a registered medical practitioner. We could not retrieve the ministry's own text, so confirm the wording with your legal adviser — our boundary does not change either way, and it is not configurable at any price.
Will it tell patients their blood test or scan results?
No, and this is a hard limit rather than a setting. The agent will tell a caller whether a report is ready, when it is expected, and where or how to collect it, and it can send the collection link on WhatsApp. It will not read out a value, will not say whether a result is normal, and will not interpret anything. If a caller asks what a result means, it says it cannot discuss results and routes the call to the department. You sign off on the exact refusal wording before go-live.
Does it connect to our hospital information system?
No, and this is our clearest weakness for hospitals. We book into a calendar with slot-locking and confirm on WhatsApp. We have no HIS or HMIS integration. That works well if your OPD scheduling lives in a register, a spreadsheet or a shared calendar. It works badly if your registration desk lives inside an HIS all day, because someone would have to retype our bookings into it — that is work we created, not work we removed. If that is your situation, choose a vendor with a native integration to your specific system and make them name it.
Can it tell callers whether a particular consultant is sitting today?
Yes, from a consultant schedule you keep updated, including leave and OT days. This question arrives all day at most front desks, and answering it repeatedly is rarely the best use of someone standing in front of a queue — pull your own call log and see what share of inbound calls it is. The agent answers it and offers the next available slot with that consultant in the same call, rather than leaving the caller to ring back. It routes by speciality too, mapping a caller's own description of the problem to the department you have configured — and where your rules do not map cleanly, it hands the call to a human instead of guessing.
Is there a subscription or an annual contract, and how does GST work?
There is no usage subscription and no lock-in on call spend. You load prepaid credits, they never expire, top-ups start at ₹500, and at zero balance the agent pauses gracefully rather than failing mid-call. The one commitment applies only to founding clients: taking the free ₹40,000 build means keeping the Care Plan at ₹5,000 a month for the first three months, optional from month four, with the founding rate locked for six months. On GST, every price we publish is exclusive of it — 18% is applied at the point of loading credits, so a ₹25,000 load is billed ₹29,500 and a ₹50,000 load is billed ₹59,000. We do not add a platform fee of any kind.
Is anyone in India actually cheaper than ₹2 a minute?
On a headline number, possibly. Scalify Labs publishes ₹0.40 a minute, but their own page labels that an estimate and tells you to get a formal quote, and it sits on top of a monthly plan from ₹15,000, one-time setup of ₹15,000 to ₹40,000, a stated minimum of 10,000 minutes a month and a three-month commitment, with voice minutes billed separately from the plan. Agni advertises ₹2 a minute, but their own homepage qualifies it as the Enterprise volume rate at 10,000-plus minutes; the buyable tiers are ₹2,999 a month with ₹8 a minute overage, up to ₹12,999 with ₹4. HuskyVoice's ₹4 a minute is published as an enterprise volume floor, not an entry rate. All checked 21 Jul 2026 on the vendors' own pages. We will not claim to be cheapest — above roughly 10,000 minutes a month a volume-priced vendor will beat us. What we will claim is the terms: ₹2 a minute from the very first minute, no subscription, no minimum volume, no lock-in.
Tell us your call volume on WhatsApp and we will work out your real monthly number before you commit to anything.
Tell us your problem
Tell us your problemTagged

Written by
Rohith Sriramula
Founder & CEO, Dvaarik AI
A laid-off engineer who went all in on Dvaarik AI — he builds every custom AI voice agent personally. This is written from hands-on work with Indian businesses, not theory.