Of the three channels AIVA runs on, voice was the hardest to price. Web chat and SMS are naturally discrete — a conversation is a conversation, a message is a message. A phone call doesn't come in fixed units. It could be answered and resolved in forty seconds or run six minutes long. We looked at three models before settling on ₹4 a minute, and the two we rejected are worth explaining, because the reasons we said no to them are the reasons per-minute protects you.
Option one: a flat rate per call
The simplest model would have been a flat fee per call, regardless of length. It's easy to understand and easy to budget against.
It's also wrong in a way that matters, and the math makes it obvious. Say we'd priced a flat ₹15 per call, set to roughly break even against an average two-minute call. A business whose calls are mostly quick "what are your hours" questions — resolved in twenty or thirty seconds — would still pay ₹15 for a call that, at our actual per-minute rate, would have cost closer to ₹2. Meanwhile, a business whose calls are long, detailed bookings running six or seven minutes would pay the same ₹15, well under what those minutes actually cost to handle. The flat rate isn't neutral — it's a transfer from the business with simple calls to the business with complicated ones, hidden inside a number that looks fair because it's the same for everyone.
That's not a small distortion, either. Call length varies enormously by industry — a quick-service restaurant confirming a takeaway order and a law firm's intake call for a new matter are not the same shape of conversation, and a flat per-call rate can't tell the difference between them.
This isn't an unfamiliar idea to anyone who's used a phone in India. Mobile billing had almost this exact debate years ago, when TRAI made per-second billing mandatory, replacing per-minute rounding as the fairer default, precisely because rounding every call up to the next full minute quietly overcharged short calls to subsidize the billing system's own convenience. We didn't want to rebuild that same quiet overcharge into a brand-new product on day one.
Option two: a monthly bucket of minutes
The second option was the standard telecom move: sell a bucket of included minutes for a flat monthly price, then charge overage. Say a plan like this offered 200 minutes for ₹800 a month, with overage priced at ₹6 a minute past that — a fairly typical shape for this kind of plan. A business that only used 120 minutes that month effectively paid ₹6.67 a minute for the minutes it did use, because the other 80 minutes were paid for and wasted. A business that ran over into 260 minutes paid ₹800 plus 60 × ₹6 = ₹1,160 total — a blended rate of ₹4.46 a minute, and a bill that jumped in a way that's hard to predict from the outside.
We rejected this for the same reason we rejected a subscription generally — it reintroduces the problem of paying for capacity you didn't use, and it adds a cliff-edge overage rate that's usually priced to sting exactly when your volume is already higher than you planned for.
What per-minute actually does
Per-minute pricing means the bill tracks the thing that's actually costing anything — time on a live call, with work actually being done on the line. Nothing is amortised across other customers, nothing is estimated in advance and reconciled later. A forty-second call to confirm a booking costs about ₹2.70. A six-minute call working through an insurance question costs about ₹24. Both are correct, in the sense that both reflect what actually happened on the line — not an average, not a bucket, not an estimate.
Per-minute pricing means the same thing has to be true of every call AIVA answers: the faster it resolves you correctly, the less you pay. That's not a coincidence — it's the whole point.
The honest edge case: does this penalize businesses with naturally long calls?
It's a fair question, and worth answering directly rather than glossing over. Some businesses just have longer calls by nature — a financial advisor walking a new client through options, a clinic explaining a treatment plan, an immigration consultant answering a complicated case-specific question. Those businesses will spend more per call than a business whose calls are mostly quick confirmations, and per-minute pricing doesn't hide that.
But that's the correct outcome, not a flaw. A six-minute call takes six minutes of work to handle well regardless of who's paying for it — a human receptionist doing that same call isn't free for six minutes either, they're just paid on a salary that doesn't itemise the call. Per-minute pricing doesn't create the cost of a long call; it just makes a cost that already existed visible, instead of burying it inside a flat rate that overcharges every other business to compensate. For a business that runs a lot of longer calls, the more useful comparison isn't "would a flat rate be cheaper" — it's what a booked appointment or resolved call is actually worth once you have a real number to divide by.
The part that protects you
Here's the mechanism that matters most, and it's easy to miss: per-minute pricing puts AIVA's own incentives on your side of the table. If we'd priced per call, there'd be a perverse case for AIVA to move fast and hang up quickly even when a caller needed more time — cheaper for us to run, worse for you. If we'd priced on a padded monthly bucket, there'd be no real pressure on us to keep calls efficient at all, because you'd already paid for the minutes whether we used them well or not.
Per-minute pricing removes both distortions. AIVA earns nothing extra by keeping you on the line, and we lose nothing by resolving you fast — which is exactly why we've spent real engineering effort getting the voice pipeline to respond in around 198 milliseconds and resolve about 96% of calls without a human handoff. A slow, meandering AI on a per-minute meter is a bad experience and an expensive one. A fast, accurate one is good for the caller and cheap for you. Per-minute pricing is what makes sure those two things stay pointed the same direction.
What this looks like against the rest of the pricing
Voice doesn't sit in isolation — it's one of three channels drawing from the same balance, alongside web chat at ₹2 a conversation and SMS at ₹1 a message, with no monthly fee sitting underneath any of it. The same logic that shaped the ₹4-a-minute rate shaped the other two: price the unit that actually reflects the work done, not a proxy for it. Every new account also starts with ₹500 in free credit, which at this rate is enough to see your own real call-length distribution — not a hypothetical one — before deciding anything.
For businesses running high call volume every month, the standard rate is still the fair default, but it's not the only option — custom pricing and invoicing exists for accounts where a negotiated rate makes more sense than the per-unit one. For everyone else, the per-minute number is designed to do one specific job: make sure the fastest correct answer is also the cheapest one, for both of us.
That's what ₹4 a minute is actually doing on your invoice. Not just metering time — making sure the pricing itself rewards getting your callers off the phone quickly and correctly, instead of quietly rewarding the opposite.