The unchanged part
An AI voice call includes more than model processing. Telephony and other session costs can remain unchanged after a model switch. Comparing model prices alone applies the discount to parts of the workflow that did not change.
A simple illustration
Suppose model processing represents half of an all-in cost. Halving that component reduces the total by 25%, assuming everything else stays constant. It does not halve the total. This arithmetic is illustrative, not a measured AIReceptionist result.
How the portfolio uses it
The AIReceptionist planning range of 30–50% whole-call savings includes both model and non-model costs under a fixed speech workload. Its price snapshot is September 1, 2026. The configuration work makes the option available; it does not establish those savings or equal caller quality in production.
Make the estimate useful
List each input, identify its source and date, and show which assumptions would change the decision. Then measure usage and accepted outcomes on the same workload. A cost model earns trust when someone else can inspect and challenge it.
About the evidence
Historical pricing assumptions. Configuration tested locally; savings modeled.
Project detail and source links ↗