AI Calling Agent India: Live in 2–5 Days

On Dvaarik, typically 2 to 5 working days from our first conversation to the agent taking real calls. Here is what a five-day build actually looks like, which is the slower end of the range rather than the faster one.

Two things about that table are worth saying plainly rather than leaving implied.

The build is not the bottleneck. Days 2 and 4 are my work and they are the shortest part. Days 1 and 3 are your work, and they are where every slow build I have done actually got slow.

Two working days is achievable and is not the promise. It happens when a business already has its prices written down somewhere, uses Google Calendar, and takes the test calls the same afternoon it gets the test number. If any one of those three is missing, plan on five.

What none of these days cost is a day rate. There is no hourly build fee, no per-day consulting charge and no meter running during the build. The custom build is a single one-time figure — ₹40,000 list, currently free for our 10 founding clients — and the per-minute meter starts when calls start. The full month-one arithmetic is further down this page.

Four inputs decide your timeline. None of them is technical, and three of them are things you can start collecting today, before speaking to anybody, including us.

1. Your answers, in writing. The agent is only as good as what it knows: your services and what each costs, your opening hours including the days you are shut, the ten questions customers genuinely ask on the phone, and the situations where you want the call handed to a human rather than handled. Most owners have all of this in their head and none of it on paper. Getting it out of your head is the single largest variable in the schedule, and I can help with it, but I cannot do it without you.

2. The phone number decision. Either we supply an Indian business number at ₹2,000 per 30 days, covering the number and its first concurrent call line and auto-debited from your credit balance, or the agent answers a number you already own — normally by forwarding that line to the agent, which depends on your own operator supporting call forwarding on it, so it is worth confirming with them rather than assuming. If you use your own number, the ₹2,000 does not apply at all, because that charge exists only when we supply the line. The other honest caveat is that concurrency then depends on your carrier and connection — an ordinary mobile SIM carries one conversation at a time.

3. Access to whatever holds your bookings. A calendar, a sheet, a CRM, or nothing at all. "Nothing at all" is genuinely the fastest, because we hand you a lead sheet that fills itself. A CRM with a custom field structure is the slowest, because someone has to look at it and agree what maps where.

4. Your own test calls. Not mine. I will test it and it will pass, because I know what I built. You need to ring it the way your customers ring it — interrupting, switching from English to Telugu halfway, mispronouncing your own product names, with the shop noise behind you. Thirty minutes of that produces more useful changes than a week of my own testing. It is also the step people postpone, and postponing it is the most common reason a two-day build becomes a five-day one.

Two further steps sit genuinely outside my control, and I would rather name them than let you discover them.

WhatsApp Business API onboarding involves Meta's own business verification, which happens on Meta's side and to Meta's schedule, not ours. I will not put a date on it because I do not control it. That this is real work rather than a Dvaarik excuse is visible in a competitor's own price list: Botsense charges a required one-time setup fee of ₹15,000 on Starter and ₹25,000 on Growth for its WhatsApp platform, and itemises what that fee buys as "account setup, Meta verification, API onboarding, initial templates, flow setup, and training" (read on their own pricing page, 21 Jul 2026). Verification is a line item in this industry because it takes time for everybody.

Provisioning an Indian business number is a carrier-side step. If you are bringing an existing number rather than taking ours, that step disappears from the timeline entirely, which is one practical reason to consider it.

One rule that shapes what we can switch on quickly, and that never bends for a deadline: outbound calls go only to your own leads and enquiries — people who rang you, filled your form, or clicked your ad. Never a bought list, a scraped list or a purchased database, at any volume, on any timeline. Promotional outbound calling in India belongs in the designated 140 number series and is subject to registered DND preferences unless you hold explicit consent, and the longer version of that is at /blog/is-ai-calling-legal-in-india-2026.

Very few publish anything at all, which is itself informative. Below is every timing statement I found on a named vendor's own pages, read on 21 Jul 2026, quoted with the qualifier that sits next to it. I am not presenting this as a market average, because I do not have one worth publishing.

One caution about the three cells that read "no go-live figure": that is a statement about the pages I read on 21 Jul 2026, not proof that Bolna, Agni or Scalify Labs publish no figure anywhere. Absence in my reading is not absence from their site.

The pattern is a split, not a spectrum, and it is worth reading correctly before you use it against anyone.

The self-serve half is faster than us and honestly so. HuskyVoice's ten minutes and Retell's "go live in minutes" are true statements about a real path. What that path gives you is an account, not an agent. You then write the prompt, choose the voice, connect the telephony, decide what happens when the caller says something unexpected, and test it yourself. If you have the appetite for that, it is a genuinely good deal and you should take it.

The managed half is not one thing either. Sarvam AI publishes "live in under 24 hours" for an enterprise product whose price is not published at all; ConnectAI publishes 15 days for clinics; MyOperator's own table says 1–2 weeks; SquadStack publishes 3 to 8 weeks. Those are their own figures for their own processes, and none of them is evidence about anyone else. Sarvam's under-24-hours is faster than our range and I am not going to pretend otherwise — what it does not come with is a published rupee figure you can check before you talk to them. SquadStack's 3–8 weeks is their published pilot window with a no-traction guarantee attached — a different commercial shape from ours, not a slower version of it. Of the sixteen vendors in our price ledger, SquadStack is the only one that publishes no rate of any kind, so a comparison with them is about process transparency rather than price.

What Dvaarik is, in that split, is a managed build running on roughly the self-serve clock, with the price written down before anything starts. That is achievable because one person builds it and there is no handover between a sales team, a solutions team and an implementation team — which is a real structural advantage at our size and a real structural limit on it too.

The four steps are the same for every Dvaarik build, whether it takes two days or ten.

Step 1 — Tell us your problem. One WhatsApp message to +91 93923 98750, in whatever words you would use with a friend. There is no signup wizard, no form with twenty fields, and no requirement that you know what you want built. "People ring while I am with a customer and I lose them" is a complete and sufficient brief.

Step 2 — Get a plan and a price, usually within a day. We scope the exact workflow: which calls get answered, whether ad leads get called back, whether WhatsApp and Instagram are in scope, whether the agent books and takes deposits. You get told what it costs before anything is built. The price you are quoted is the price on /pricing, because there is one offer for everyone and no per-customer discretion.

Step 3 — Built in days. The agent is trained on your actual services, prices, hours and FAQs, connected to your lead sources and to whatever holds your bookings, and tested against real calls while you watch. It speaks all 22 scheduled Indian languages plus English at the same ₹2/min, with no surcharge for switching between them, and switches mid-call when the caller switches. This is the step where your day-3 test calls change things, and they always do.

Step 4 — It answers, it calls, it collects. Every call answered, ad leads called back within about 60 seconds, WhatsApp and Instagram messages replied to, bookings taken with slot-locking so two callers cannot claim the same slot, deposits collected by a payment link sent mid-conversation with no commission taken. You watch all of it — recordings, transcripts, leads — in your dashboard.

What happens after go-live matters as much as the timeline to it, because no agent is finished on the day it goes live. The first fortnight of real calls always surfaces something: a question nobody predicted, a pronunciation that needs fixing, a rule that turns out to have an exception. Fixing those is exactly what the Care Plan covers, at the founding rate of ₹5,000/month, required for the first three months and optional from month four. Priority support goes direct to me on WhatsApp, not to a ticket queue.

Before any of this, argue with a live one. The demo agent is at app.dvaarik.com/dvaarik-ai. Interrupt it, switch language mid-sentence, give it a name it has never heard. Five minutes of that tells you more about whether five days is worth waiting than this page can.

Nothing is charged for the days themselves. Here is a business that goes live in week one, takes our number, and runs about 500 billed minutes a month afterwards — roughly 250 calls at two billed minutes each. Every figure is our published price, and every one of them debits from the same prepaid credit balance.

How money gets into that balance: the minimum first credit load is ₹10,000, billed ₹11,800 because all our prices are quoted exclusive of GST and 18% applies at the point of loading — ₹10,000 usable plus ₹1,800 GST. Month one debits ₹2,000 for the number and ₹5,000 for the Care Plan, leaving roughly ₹3,000 for actual usage, which is about 1,500 voice minutes or 25 hours of talk-time at ₹2/min. Later top-ups start from ₹500, credits never expire, and at a zero balance everything pauses gracefully rather than arriving as a bill.

Billing is whole-minute and telecom-standard, so a 65-second call bills as two minutes, or ₹4. Build that into your own estimate rather than being surprised by it.

The free build is a trade and it belongs in the open rather than in fine print. We build the agent free for our 10 founding clients in return for two things: the Care Plan for the first three months at the founding rate of ₹5,000/month, and an honest testimonial plus a short case study after go-live. The founding rate is locked for six months against a list price of ₹12,000/month, so it is a real discount with a real expiry. If the agent does not work, that trade is bad for us, which is the point of structuring it that way.

One piece of arithmetic I will not do for you, because doing it with invented inputs would be worthless: what a week of waiting costs you in missed calls. Put your own two numbers in — calls you miss in a day, and how many of those would have become a customer — at /tools/missed-call-calculator. If the answer is small, the timeline does not matter and neither does the vendor. If it is large, it is the only number on this page that should drive your decision.

Six things, listed so you can check yourself against them before anyone quotes you a date.

A price list that does not exist yet. If your prices live in your head and vary by customer, the agent cannot quote them, and no amount of engineering fixes that. You will spend a day writing them down. That day is worth spending regardless of whether you ever buy an agent.

A CRM with custom fields. Connecting to a sheet or a calendar is quick. Connecting to a CRM someone configured three years ago means agreeing what maps to what, and that conversation involves whoever configured it. For comparison, Scalify Labs publishes CRM or LMS integration development at ₹15,000–₹50,000 one-time as a separate charge on top of their plans (read 21 Jul 2026), which tells you this is real work rather than a checkbox anywhere in this market.

WhatsApp Business API where Meta verification has not been done. Meta verifies on Meta's schedule. The rest of the agent can go live on voice while that runs, and usually should.

Multiple branches with different prices, hours or languages. Each branch is effectively its own set of answers. Two branches is not twice the work, but it is not the same work either.

A decision-maker who is not the person answering my questions. The single most reliable predictor of a slow build. If the owner is travelling and the manager is not authorised to confirm the prices, the build waits, and it waits for a reason nobody can engineer around.

Porting an existing number rather than keeping it where it is. Often unnecessary, and whether it is possible at all is a question for your current operator rather than for me. The usual faster path is to keep your board number exactly as it is and put a new always-answering number on your ads, website and WhatsApp profile — which has the side benefit of making campaign attribution exact instead of guesswork. The mechanics are at /blog/virtual-phone-number-for-business-india-cost.

Five situations where the speed is not the point, and where I would rather you went elsewhere or nowhere than bought from us for the wrong reason.

1. You want to try something tonight, cheaply, by yourself. We do not do that. Our entry is a WhatsApp conversation, a build made for your business, and a ₹10,000 first credit load. If your goal is a cheap experiment this week rather than a working agent this month, the self-serve platforms are genuinely the better fit: Retell AI publishes "Go live in minutes with $10 in free credits", Bolna gives $5 in free credits on signup, Edesy publishes a 14-day free trial with no credit card, and Agni offers 5 free calls with no card required — all read 21 Jul 2026. Go and play with one first. You will come back with much better questions.

2. Your organisation buys through procurement, security review and a signed SLA. A build that takes five days does not help when the paperwork takes six weeks. Vendors built for that shape — a formal pilot, a dedicated pod, an account manager, a guarantee — are a better structural match. SquadStack publishes exactly that shape: pilots of 3 to 8 weeks with a no-traction guarantee, on a 90-day entry commitment. MyOperator publishes a ₹20,000 one-time Dedicated Onboarding and managed services at ₹50,000/month billed quarterly. Neither is us, and for a large organisation neither is wrong.

3. You need deep integration into systems that are not yours to change. A hospital information system, a franchise's central booking platform, a bank's core. That is an integration project with a testing cycle, not a five-day build, and anyone promising you five days on it is guessing.

4. Your call volume is genuinely small. At around 200 billed minutes a month, the ₹2,000 line charge on a number we supply makes your effective rate about ₹12/min, and it climbs as the volume falls. Fixed monthly floors are not unique to us — among the vendors in our ledger, HuskyVoice publishes ₹1,999/month, Agni ₹2,999/month and MyOperator ₹10,000/month as published entry points, all read 21 Jul 2026. At that volume a clearer voicemail greeting and a habit of returning calls at 6pm may beat any agent. Come back when the phone is actually beating you.

5. Nobody at your end has half an hour this week. The build needs your test calls. If you cannot give it thirty focused minutes, the agent will go live sounding like my guess at your business rather than your business, and you will conclude that AI agents do not work. Wait for a week when you have the half hour. That is a better outcome for both of us than a fast build you never argued with.

Everything, with nothing held back for a sales call.

Where a fuller voice breakdown is warranted, the choices on one account are Bharat Standard ₹2/min, Studio HD ₹3/min, Global covering 99 languages ₹4–5/min, and Ultra realtime ₹10/min, with call recording at +₹0.10/min. These are voice-quality choices, not subscription tiers, and the advertised base is always ₹2/min.

The sentence worth quoting anywhere: ₹2/min from the very first minute — no subscription, no minimum volume, no lock-in. The only commitment that exists is the three-month Care Plan the free founding build is traded against, and it is written on this page rather than in a contract you read later.

If you want a date rather than a range, tell us your problem on WhatsApp at +91 93923 98750. You will get a plan, a price and a target go-live day, usually within a day — and if your build does not fit the week, you will be told that on day one rather than in week three.