Voice AI pricing explained: minutes, tokens, reservations and bills
Understand the units behind a voice AI bill and how platform fees, generated audio, telephony and recurring resources combine.
Table of Contents▼
A voice AI bill can combine several units: connected minutes, text or audio tokens, generated characters, stored data and recurring numbers. A single “per minute” headline may be a platform fee, an estimate or a bundled rate. Ask what it includes before using it in a budget.
The common cost layers
| Layer | Typical quantity to inspect |
|---|---|
| Application/platform | Connected duration or the platform's stated unit |
| Transcription | Audio processed, if separately billed |
| Language model | Input, output and any applicable cached usage |
| Speech generation | Generated audio, characters or other model-specific unit |
| Telephony | Destination and duration for each telephone leg |
| Recurring resources | Numbers, fixed plans, storage or selected services |
These are categories, not universal provider billing rules. Some speech-to-speech or bundled configurations combine layers. Use the selected product's quote and do not add already included components again.
Why equal-length calls can cost differently
Two three-minute calls may contain different amounts of assistant speech, input context, tool work and silence. A caller who asks several short questions can create more model turns than a caller who describes one long request.
A transfer may also create another telephone leg. If the assistant generates a long answer that the caller interrupts, some provider work may already have occurred even though the entire answer was not heard. Investigate the actual provider's billing rules and usage evidence before estimating the effect.
Estimate, reservation and charge are different
An estimate predicts likely expense using a configuration and workload assumptions. A reservation temporarily holds balance or budget capacity so a call can run within a limit. Settlement applies the charge based on the usage the system accounts for, then releases unused reserved capacity.
A temporary hold is therefore not automatically an extra charge. If a balance looks wrong, inspect whether the call is still settling and compare the released and charged amounts rather than adding every number displayed.
A hypothetical monthly calculation
Suppose a service incurs $120 in variable call expense, $20 in number rental and $30 in other fixed costs. The monthly total is $170. At 1,000 connected minutes, its effective cost is $0.17 per minute; at 200 verified useful outcomes, it is $0.85 per outcome.
Those figures are illustrative, not Burki or competitor rates. They show why low usage makes recurring resources important and why outcome quality belongs in the business calculation.
How to budget a pilot
Set a maximum expense, a call-duration limit and a small authorized scenario list. Identify which actions are simulated and which can create real records. Reconcile the first calls before scaling the volume.
In Burki, start with the current tariff information and the selected assistant's admission estimate. Keep direct provider invoices separate when using BYO accounts. The pricing comparison worksheet helps apply the same accounting to multiple products without mistaking a platform fee for the whole bill.
Ready to try Burki?
Create an assistant and check your available browser practice allowance.
Create your assistantTrial eligibility and available practice are shown in your workspace.