Transcript first.
Call ends → transcript written immediately. Analysis runs async so hangup stays fast.
Call Analytics
AI call analytics for better voice-agent workflows.
AI call analytics turns completed calls into reviewable transcripts, summaries, outcomes, and patterns—so your team can see why callers contacted you, where a workflow broke down, and what to improve next.
What is AI call analytics?
AI call analytics helps teams review conversations at scale: inspect individual transcripts, group calls by outcome, and find the patterns that deserve a workflow, knowledge-base, or handoff change.
Wirevox organizes supported post-call details into a record your team can inspect alongside the transcript, making patterns easier to trace back to an individual conversation.
Pick an outcome.
Booked, transferred, message taken, inquiry, cancelled, rescheduled, intake, or other.
Leads & hooks.
Enrich or create Contacts, optionally email a summary, fire call.ended if configured.
Logs
History with the full transcript.
Logs (`/w/:id/logs`) holds call history, transcripts, and chat logs — the place to open a specific conversation after Analytics shows the trend.
- Outcome, summary notes, and caller fields on the call record
- Phone backfilled from caller ID when the caller never said it
- Lead status mapped from outcomes into Contacts
Metrics that lead to action
What should teams measure in AI calls?
Start with measures that reveal whether callers reached the right next step. A useful review connects each metric to a transcript and a specific workflow change.
Call outcomes
Group calls by booked, transferred, message taken, inquiry, changed appointment, intake, or other. Review unexpected outcomes instead of treating every completed call as a success.
Handoff quality
Check whether the right calls reached a person and whether the summary gave that person enough context to continue.
Knowledge gaps
Find repeated questions the agent could not answer reliably, then update the approved knowledge source and retest the scenario.
Workflow completion
Compare the caller’s stated goal with the recorded outcome. Investigate transfers, early endings, and fallback paths that occur more often than expected.
Outcomes the pipeline understands.
Post-call analysis classifies each transcript into a strict set of outcomes — or skips re-classification when a mid-call tool already set a higher-priority result like booked.
BOOKED
Appointment scheduled
A successful booking — often already set by calendar or Open Dental tools during the call.
TRANSFER
Handed to a human
transfer_call bridged the caller; outcome reflects the escalation.
MESSAGE
Message taken
Caller left details for the team via take_message or equivalent flow.
INQUIRY
Questions only
Asked for information without booking or leaving a message.
CHANGE
Cancelled / rescheduled
Existing appointment moved or cancelled — mapped cleanly into lead status.
INTAKE
CRM intake
Caller info submitted to Clio, Jobber, Filevine, or similar intake tools.
OTHER
Everything else
Wrong number, early disconnect, or outcomes that do not fit the list above.
Post-call pipeline
Seconds after hangup.
After a call ends, the workflow can save a transcript, prepare a summary, classify an outcome, and deliver the next approved action for your team to review.
01 / TRANSCRIPT
Save immediately.
Database write on WebSocket close — call marked ended even if analysis later fails.
02 / ANALYZE
Summary + details.
Structured extraction: outcome (when allowed), 1–2 sentence summary, name, phone, email, reason.
03 / LEADS
Enrich or create.
Update a lead created mid-call, or insert a new Contact so every caller appears.
04 / DELIVER
Email & webhooks.
Optional email_send_summary to the owner; call.ended POST with transcript payload.
Priority & backfill
Never downgrade a booking.
If a tool already set booked during the call, post-call analysis skips re-classification and only enriches the summary and caller fields. Phone numbers fall back to Telnyx caller ID when the transcript never stated one.
01
Outcome priority.
booked beats rescheduled, cancelled, intake/transferred, message_taken, inquiry, and other — lower priorities never overwrite higher ones.
02
Skip re-classify when tools won.
Existing outcomes other than inquiry/other are trusted; the LLM still writes the summary and extracts details.
03
Phone backfill.
If extraction misses a number, Wirevox uses call.callerNumber (caller ID) or a phone captured by tools mid-call.
04
Email summary is post-call only.
email_send_summary never runs live mid-call — it fires after hangup so the digest includes the complete conversation.
05
Review downstream delivery.
Confirm configured post-call actions and their outcomes before making them part of an operational dependency.
— In the product
Two places to look after every call.
Dashboard trends
Call volume, usage trends, and active call indicators for the weekly ops review.
Call-level detail
Open the transcript, outcome, and notes for a single conversation when something looks off.
Follow-up context
Approved call details can support follow-up, with status mapped from outcomes such as booked, intake, or transferred.
FROM THE DOCS
Section 7 — Post-Call Processing & Webhooks
Transcript save, outcome classification, lead sync, email summaries, and webhook delivery — plus Analytics & Logs in Getting Started.
Turn call volume into clarity.
Book a demo and we’ll walk Analytics, Logs, and a live post-call summary for your workflow.