The DILR Voice Index · Q2 2026 (April–June)
Production benchmarks for
enterprise voice AI.
The DILR Voice Index publishes anonymized production metrics from a live multi-agent voice platform running across 30+ languages and five regulated industries. Q2 2026 (April–June) covers inbound and outbound enterprise traffic, reporting resolution rate, escalation rate, time-to-human, cost-per-resolved-call, and voice latency by language. The numbers below are aggregated across all opted-in customers; no single customer is identifiable.
Method
All metrics are computed from production call logs over the period 2026-04-01 to 2026-06-30. Data is aggregated and anonymized before publication; no customer, caller, or transcript content is exposed. Customers can opt out of inclusion at any time; opted-out traffic is excluded from all aggregates.
- Resolution rate — share of inbound/outbound calls that completed the intended workflow without human handover.
- Escalation rate — share of calls that triggered a human handover at any point. Excluded from resolution numerator.
- Time-to-human — median seconds from escalation trigger to human agent connected, across all escalated calls.
- Cost per resolved call — total platform + telephony cost divided by resolved-call count, by industry.
- Voice latency by language — P50 round-trip response time, measured user-speech-end to AI-speech-start.
Headline · Q2 2026 (April–June)
TODO_v1 — populated once Q2 2026 (April–June) extracts complete.
Resolution rate by industry — Q2 2026 (April–June)
| Metric | Resolution rate | Calls (n) |
|---|---|---|
| Insurance — claims intake | TODO_v1 | TODO_v1 |
| Healthcare — appointment booking | TODO_v1 | TODO_v1 |
| Financial services — KYC | TODO_v1 | TODO_v1 |
| Real estate — lead qualification | TODO_v1 | TODO_v1 |
| Education — admissions follow-up | TODO_v1 | TODO_v1 |
Escalation and time-to-human — Q2 2026 (April–June)
| Metric | Value | Notes |
|---|---|---|
| Escalation rate | TODO_v1 | Calls handed to a human agent at any point |
| Median time-to-human | TODO_v1 | Across all escalated calls |
| Context-completeness on handover | TODO_v1 | Pre-resolved intent + transcript + attempted actions |
Cost per resolved call — Q2 2026 (April–June)
| Metric | USD |
|---|---|
| Median cost per resolved call | TODO_v1 |
| Human-only baseline (same workflow) | TODO_v1 |
| Net cost delta | TODO_v1 |
Voice latency P50 by language — Q2 2026 (April–June)
| Metric | Latency (ms) |
|---|---|
| English (UK) | TODO_v1 |
| English (US) | TODO_v1 |
| Hindi | TODO_v1 |
| Spanish | TODO_v1 |
| Mandarin | TODO_v1 |
| Arabic | TODO_v1 |
Citation
When citing the DILR Voice Index, please use:
Dilr.ai (2026). DILR Voice Index — Q2 2026 (April–June). Retrieved from https://www.dilr.ai/voice-index
Published under CC BY 4.0. Attribution required.
Want the next Index direct to your inbox?
Quarterly cadence. No marketing email.