Call health & QA
Every AI call gets a health check. Bad calls stand out.
Call log · today
- Rahul Sharma · 2:41Healthy callDemo_Booked
- Meena Iyer · 1:12Healthy callCallback
- Unknown · 0:19Probably an answering machineIncomplete
- Vikram Rao · 3:05Long silence (4.1 s) mid-callInterested
- Sana Khan · 0:47The agent could not hear the callerIncomplete
- Arjun Mehta · 2:10Healthy call · transferred to repHot_Lead
What it catches
Plain-language verdicts, not a wall of metrics.
The agent could not hear the caller
Speech was detected but nothing usable came through. Often a bad line or a very noisy room.
The agent never spoke
The caller spoke and the agent produced no audio. Rare, and always worth a look.
Voice synthesis stalled
The voice engine stopped mid-reply. The agent reconnects and repeats, and the call is flagged.
Long silence
A gap of 3.5 seconds or more during a real conversation, with what the agent was doing at the time.
Caller answers were discarded
Something the caller said never reached the conversation. The exact words are shown.
The agent kept restarting a reply
Two near-identical replies in a row (three turns it red): usually a script or interruption problem.
Probably an answering machine
An estimate only. It never changes the outcome on its own.
Transfer failed · slow responses
A handover that didn't connect, or replies that took too long, measured on the call.
Honest numbers
If we couldn't measure it, we say so.
A quality dashboard that shows zero when it means “we don't know” is worse than no dashboard. On Telleo, a signal that wasn't measured on a call is shown as “not measured”.
When a caller's answer is lost, you see the exact words, not just a count. When there is a long silence, you see what the agent was doing at that moment.
- Verdict
- Long silence
- Worst gap
- 4.1 s · after the caller's turn
- Interruptions handled
- 3
- Answers discarded
- 0
- Repeated replies
- 0
- Speech-to-text delay
- not measured
The call log
Search, listen, export.
- A health dot on every AI call, with the headline on hover.
- Transcript, recording (when switched on) and timings on every call.
- Live status while a call is in progress.
- Search, filter and export by outcome, date, team, counsellor and call type.
- Full phone numbers only for roles allowed to see them; diagnostics for admins.
Search & filter
By outcome, counsellor, date
Live status
Watch calls as they happen
Role-based access
Numbers and diagnostics by role
Export
Take calls into your QA
A 20-minute weekly QA routine.
What we recommend to every team running AI calls. It catches script problems early and keeps the agent sounding like you.
- 1
Scan for red and amber
Scan the call log's health dots and open the week's red calls first, then a sample of amber ones. Read the headline before you listen.
- 2
Read three healthy calls too
Pick three green calls at random. A healthy call can still have a weak script.
- 3
Check outcomes against transcripts
Did “Demo_Booked” really have a day and time? Did “Not_Interested” really say no?
- 4
Fix the script, not the symptom
Change the opening line, a question or an outcome, then use the feedback tool to revise and re-test on your own phone.
Frequently asked questions
Is call health the same as call intelligence?+
No. Call intelligence is about the conversation: outcome, objections, sentiment. Call health is about the machinery: could the agent hear, did it speak on time, did anything go silent or get lost. You get both on every AI call.
What does ‘not measured’ mean?+
Some signals can't be measured on every call. When that happens, the panel says so instead of showing zero, so a clean-looking number is never a guess.
Can we hear the call?+
With recording switched on for your line, yes: play the recording next to the transcript and the timings. Every AI call has a full transcript either way.
Who can see transcripts?+
Health verdicts and technical diagnostics are shown to your account admins. Anyone with access to the call log can open transcripts; they are not restricted per role. Health verdicts and technical diagnostics are shown only to your account admins, and on the call log you choose which roles see full phone numbers., and on the call log you choose which roles see full phone numbers.
Can we export calls for our own QA?+
Yes. Search and filter the call log by outcome, date, team, counsellor and more, and export the results.
Hear it on a real phone call.
Book a 20-minute demo and we'll build a first agent around your script, or ring our test line and talk to one right now.