Almost every voice AI product a hotel can buy today is billed by the minute. That reads like an accounting detail. It is not. The unit you are billed in is the behavior your vendor is paid to produce, and a per-minute unit pays your vendor, structurally, to get your guest off the phone.
The billing unit is an incentive with an invoice attached
Per-minute is the industry default because the horizontal platforms underneath most voice products price that way. Vapi publishes per-minute rates. So does Retell. That transparency is to their credit, and per-minute is an honest way to sell infrastructure.
But follow the money one layer up, to the vendor selling your hotel a finished voice agent. Their cost is per-minute and compounding. Every additional minute carries more speech and telephony, billed per minute as Aircall’s breakdown of voice agent costs lays out. And the tokens pile up: as one 2026 breakdown of voice-agent pricing shows, the model is handed the full conversation transcript on every turn, so the context grows as the call runs. Minute eight of a call costs the vendor more than minute two.
Which means the vendor’s margin improves every time your guest hangs up early. CX Foundation’s analysis of conversational AI pricing models says it plainly: per-minute billing incentivizes short calls, and the industry is now moving toward outcome and resolution-based pricing precisely to repair that misalignment.
Nobody tunes an agent to hang up on guests. The mechanism is quieter. Every prompt revision, every latency fix, every “efficiency” release gets evaluated against a cost line that only improves when calls get shorter. The pressure points one way, forever.
A hotel call is not a ticket to be closed fast
Here is the tension, and we will state it as a tension rather than a proven fact, because nobody has published clean causal data: a reservation call is often a patience game. The anniversary caller wants to talk through the view from the corner king. The family of five has four questions before the credit card comes out. We cannot prove the longer call always converts better. We can observe that the hotel wants its agent to have zero reasons to prefer the short one, and per-minute economics hands it a standing reason.
In a support queue, shorter is usually better, because the ticket is a cost. On a reservation line, the call is not a cost. It is the highest-intent revenue channel the property owns. Importing a billing model built for cost centers into a revenue center is the category’s quiet mistake.
You cannot audit an incentive you cannot see
It gets harder, because hospitality voice pricing is strikingly opaque. The major hotel-specific vendors sell through contact-sales motions with no public number; Hotel Tech Report’s coverage of Canary’s voice product, for instance, can only estimate pricing by room count. That is the category norm, not a knock on any one vendor. But it means most hotels signing a voice AI contract cannot say what a minute costs them, let alone what a booking does.
Cost per Captured Call is the number the invoice should show
In April we argued for Call Capture Rate: the percentage of inbound calls that reach a human or AI capable of closing the booking. CCR tells you whether the phone got answered. It does not tell you what you paid for the answer.
So here is the companion metric: Cost per Captured Call (CPCC). Total voice spend for the period, divided by the number of calls where the system captured the outcome the guest called for: the booking closed, the stay saved, the request resolved. Not dollars per minute. Dollars per result.
CPCC does what per-minute pricing structurally cannot: it makes patience free and failure expensive. Two vendors at identical per-minute rates can produce wildly different CPCCs, and the one whose agent lingers ninety extra seconds to close the anniversary booking wins the metric while losing the stopwatch. It also gives operators one number to demand in the sales process. Ask the vendor what your CPCC will be. If the answer is a rate card denominated in minutes, you have just learned what they optimize.
The industry is already drifting toward outcome-based pricing, because minutes were never the product. Operators can get ahead of the drift by measuring in outcomes now, whatever unit the invoice arrives in.
A minute is an input. A booking is the point. Vendors bill in their unit. You should measure in yours.
Cost per Captured Call is how you keep score.