What Concurrent Calls Actually Means, and How Many You Need

Every voice AI pricing table has a concurrency row. Almost none of them explain it, and buyers reasonably assume bigger is better and pay for headroom they will never touch — or, more expensively, size it from a monthly call volume and discover the limit during their first campaign.

Here is what the number is, and how to work out your own.

Not calls per day. Not calls per month. The number of phone conversations your agent can hold simultaneously.

Five concurrent calls means five people can be talking to your agent at 11:04 on Tuesday. It says nothing about how many can be talking at 11:05. If your calls average three minutes, five concurrent lines can carry roughly a hundred calls an hour — which is a lot more capacity than the number 5 sounds like.

The formula is short, and it is the same one call centres have used for decades:

concurrency needed = calls per hour × average call length in minutes ÷ 60

A clinic taking 40 calls a day, clustered into maybe four busy hours, at two minutes a call: that is 10 calls an hour × 2 ÷ 60 = 0.33. One line covers it, and five is comfortable headroom for the moments when three people ring at once.

A D2C brand confirming 2,000 COD orders in a morning is a different shape entirely. Outbound is worse than inbound here, because you control the timing and will naturally fire the whole list at once. 2,000 calls over three hours at 90 seconds each is 667 an hour × 1.5 ÷ 60 = about 17 concurrent lines.

Inbound is bursty and you do not control it. Your average hour is irrelevant; what matters is your worst ten minutes. A campaign, a festival, an outage at a competitor — everyone rings at once. Size inbound for the peak, not the mean, because the cost of being short is a caller who hears a busy tone and does not ring back.

Outbound is smooth and you do control it. The dialler decides the pace, so concurrency sets your throughput rather than your reliability. Being short here means the campaign takes longer, which is usually survivable — and sometimes actively good, because a list of 2,000 calls placed in twenty minutes is a spike your own team cannot handle when people start calling back.

On inbound, a caller arriving above the limit gets a busy signal or falls through to whatever your number does next. This is why sizing inbound on averages is a false economy: the failure is invisible to you and completely visible to the customer.

On outbound, the dialler simply holds the row until a line frees. Nothing is lost; the campaign runs longer.

Concurrency is separate from cost. A minute costs the same whether it is your first simultaneous call or your twenty-fifth — you are not billed for idle lines, and raising the ceiling does not raise the rate. Which means the honest way to size it is to the traffic you actually have, not the traffic you would like to describe.

Sizing is not delicate. Take your busiest realistic hour, apply the formula, double it.

• Starter — 5 concurrent at ₹2,999/month.
• Growth — 25 concurrent at ₹7,999/month.
• Scale — 100 concurrent at ₹19,999/month.

Most inbound businesses — a clinic, a dental practice, a diagnostics lab — never approach five. Outbound campaigns are what move you up the ladder, and outbound is a Growth feature, which is usually the real reason to be on that tier rather than the concurrency number next to it.

Is concurrency a hard limit or a soft one? A hard limit means call twenty-six is refused. A soft one means it is queued or throttled. Both are defensible; they fail very differently, and it should be in writing which one you have bought.

And a second: does the limit apply per agent or per account? If you run three agents on one account, a per-account ceiling is shared between them — which is fine, as long as you knew.

Ours are listed on the pricing page, next to what the plan actually includes.