Voice layer · 15
Conversations, annotated
Every conversation with Aiera is transcribed and scored, but the value is easiest to see one call at a time. This is a real enquiry, walked turn by turn, with the reasoning behind each response made explicit alongside it.
Sessions, trailing 30d
15,630
▲ 9%vs prior 30d
Success rate
80%
▲ 5%vs prior 30d
Voice share of sessions
68%
▲ 3%vs prior 30d
Avg. call length
4m 12s
▼ 4%tighter qualification
Inbound enquiry, voice
Annotated transcript — Aiera qualifying a 30-bed requirement
Session volume and success rate
Daily, trailing 14 days
Language of the conversation
Native scoring, not translation
5,530
Sessions
Aiera sessions across the book
Volume and success rate, by role and channel
| Agent | Role | Sessions | Success rate | Channel |
|---|---|---|---|---|
| Aiera | Inbound enquiry, voice | 2,410 | 68% | Voice |
| Aiera | Inbound enquiry, WhatsApp | 4,180 | 71% | |
| Aiera | Renewal and expansion, voice | 1,240 | 83% | Voice |
| Aiera | Patient support, WhatsApp | 6,840 | 89% | |
| Aiera | Referrer bed, voice | 960 | 74% | Voice |
15,630 sessions in the window shown here, held to the same annotation standard as the call above.
Where a call ends
Outcome split, all sessions
Funnel from answer to visit
What the transcripts show at scale
Aiera qualifies before it quotes
PatternIt establishes bed count and move-in window before any number is given, which is why quoted availability holds up on follow-up.
Kannada-first calls close visits at a comparable rate to English
LanguageNative scoring rather than translation appears to be the reason — tone and hesitation are read correctly in the source language.
Budget confirmation is the single largest drop-off
Funnel1,324 of 5,432 requirement-captured calls stall here; a rate anchor earlier in the script is being tested against this cohort.
What the transcript shows
Aiera does not simply send the price list. It establishes the two variables that determine what the number actually is, qualifies the move-in window before quoting anything, answers with named availability drawn from live inventory, pre-empts the growth question with same-building expansion capacity, and closes on a held suite and a named slot rather than a vague follow-up.
Why this is shown one call at a time
A success rate is a useful summary and a bad argument. The way to judge whether a voice agent is any good is to read what it actually said, in order, and check whether each turn earns the next one. This is the same standard a sales manager would apply to a strong human call, applied here at a volume no manager could review by hand.
