Know exactly where your outbound bot loses the call.
Outbound AI agents fail in small, repeatable ways: a pause that is a second too long, a question asked twice, a corrected phone number that never sticks. VoiceLens grades every attempt so you can fix the prompt instead of guessing from a sample.
Free to start · Works with any call platform · Same rubric for bots and people
How VoiceLens helps
From transcript to a scorecard your team acts on.
Every attempt graded, not a sample
POST each transcript or recording as the call ends and get a health score with faults, latency and captured slots back for all of them.
Latency outliers against your own baseline
Per-turn response times are summarised as median, P95 and maximum and graded against the latency threshold you set in the rubric, so a slow release shows up as a spike, not as an anecdote.
Compare versions on one scorecard
Run the same rubric across agent versions or channels and see which one captures more slots with fewer faults before you roll it out.
Fault detection
What VoiceLens catches on these calls
Every fault is timestamped and pinned to the transcript turn it came from, for AI and human agents alike.
Dead air
Silence long enough for the caller to wonder if anyone is there.
Question repetition
The agent asks for something the caller already gave.
Agent latency
Slow replies after the caller finishes speaking, measured per turn.
Context loss
The agent forgets what was established earlier in the call.
Customer had to prompt
The caller has to nudge the agent to continue or respond.
- Fault codesThe fourteen faults VoiceLens detects, what triggers each one, its severity, the evidence attached, and which ones you can switch off per rubric.
- Latency and dead airHow per-turn agent latency is measured, which statistics are reported, and how the rubric's thresholds turn slow turns into faults.
- Ingest APISend call recordings into a channel: multipart upload, download from a URL, or a pre-signed upload URL. Metadata fields, responses, idempotency and errors.
Related solutions
Inbound support automation
Catch missed pivots and empathy gaps, track sentiment turn by turn and escalate flagged calls to your QA queue.
Learn moreHuman agent teams
Grade every call, coach with timestamped evidence and benchmark human agents against your bots on one rubric.
Learn moreQuality and compliance
Severity-graded fault feed, required and forbidden phrases checked on every call, evidence pinned to the transcript.
Learn moreSee what your QA process has been missing.
Create a workspace, drop in a recording or a transcript, and read the first scorecard in minutes.