Companies shopping for public AI voice agents on Oct. 20, 2026 faced a pricing landscape with no single standard. Vendors are charging everything from flat all-in rates to layered per-minute and token-based billing, and carrier fees are frequently tacked on separately.
For buyers, the sticker price is only the starting point. The structure of each plan determines what a call actually costs once carrier charges are added.
Three billing models in the same market
The public pricing pages show three broad approaches. Some vendors quote a flat all-in rate, which bundles the voice agent and associated usage into one figure. Others bill per minute, with the rate applied to each unit of call time. A third group uses token-based billing, charging according to the text the agent processes rather than the duration of the call.
Each model shifts risk differently. Flat rates make budgeting straightforward but can overcharge low-volume users. Per-minute billing tracks call length but leaves token consumption unpriced. Token billing follows the computational work but can be harder to forecast when conversations run long or complex.
Carrier fees sit outside the headline rate
Carrier fees are often extra, regardless of which model a vendor uses. That means the advertised rate doesn't reflect the full cost of connecting a call. Buyers comparing quotes need to ask whether telephony charges are included or added at the end of the month.
The result is a market where two vendors can advertise similar headline numbers and still deliver very different invoices. A flat all-in rate with carrier fees excluded isn't the same product as a per-minute rate that includes them.
What buyers should check before signing
The first question is whether the quoted rate covers carrier fees. The second is what unit the vendor bills on — call time, tokens, or a flat fee per interaction. The third is how overages are handled when usage exceeds the plan.
None of this is hidden, but it's scattered across pricing pages and contract terms. Buyers who compare only the headline number are comparing incomplete figures.
As of Oct. 20, 2026, no single model has emerged as the default. Vendors are still splitting across flat, per-minute, and token-based approaches, and carrier fees remain a separate line item for many of them.
