use case
verified July 2026
Replacing IVR menus with voice agents.
An IVR never had a latency problem — a menu can take its time. The moment you replace it with a conversation, the caller’s 200-millisecond turn-taking reflex applies to a machine, and the machine has to keep up on every line at once.
01
What changes when the menu talks back
- Latency becomes a product feature: menus tolerated seconds; conversations start breaking past one. First audio here: 107 ms, measured.
- Barge-in becomes mandatory: callers trained on “press 1” interrupt constantly, and the agent must stop mid-word and answer the interruption.
- Concurrency becomes visible: an IVR port maps one-to-one onto a line — count your busy-hour ports and you have sized the fleet.
02
The migration math
IVR replacement is a high-minute workload from day one, because it inherits the call center’s entire volume. Twenty ports at 8,000 minutes each is 160,000 minutes a month: $4,800 to $16,000 on published meters, $3,000 on twenty flat lines. The forecast for finance is one multiplication.
Table 1 — A 20-port IVR, migrated
| Volume | On a meter | On 20 lines |
|---|---|---|
| 160,000 min / mo | $4,800 – $16,000 | $3,000 |
| volume doubles | $9,600 – $32,000 | $3,000 |
03
The rollout
Nobody cuts over a call center in one night. Standby failover lines ($20 a month each) run beside the legacy stack; burst lines absorb the days the forecast was wrong; and the 20-line pilot is sized to shadow one real queue for one real month.
See also
Related sheets.
use case
8,000 min
TTS for AI customer-support agents
The latency budget and cost arithmetic for a voice support desk: why first audio decides whether callers interrupt, and what 8,000 minutes a month costs.
use case
12,000 min
TTS for order-status calls
An order-status call is one fact delivered fast. The latency and seasonal cost shape of where-is-my-order traffic, with burst lines for the peak weeks.
capability
$1,000
The pilot: a month on real traffic
The pilot is 20 lines for 30 days at $1,000, credited toward the first annual invoice — sized to shadow one real queue with your production prompts.
glossary
×2
Barge-in
Barge-in is the caller interrupting the agent mid-sentence. What the TTS layer must do to support it: stop instantly, resynthesize fast, stay cheap.
Thirty minutes, your production script, the live latency readout — measured in front of you.
Bring this use case to the call