Contact centres and voice agents

Call centre speech analytics in the languages your customers speak.

Most contact centres review a tiny sample of calls, and most analytics tools only understand a few languages. ADB Voices transcribes every call, writes a short summary, scores sentiment so unhappy customers surface, and masks card and phone numbers.

Transcribing· Customer call · Kiswahili

00:31
  • 00:12 Nimesubiri wiki mbili na bado sijapata huduma.
  • 00:20 Namba ya kadi yangu ni 4417 1234 5678 9113.
  • 00:31 I am sorry, I will escalate this to billing today.
Example outputSummary · sentiment negative

What you get

  • Summary after every call
  • Unhappy calls surfaced
  • Card and phone numbers masked

Features used

  • Transcription
  • Summary
  • Sentiment
  • PII redaction
  • Speaker labels

Deployment: Managed or licensed

In use in Longa.

How it works

  1. 1.

    Send recordings

    Calls go in through the API as they finish. A signed webhook tells you when each is done.

  2. 2.

    Analyse

    Transcript, summary and sentiment per call, with card and phone numbers masked.

  3. 3.

    Act

    Supervisors see the calls that went badly instead of a random sample.

Questions

Is real-time transcription available for voice agents?

A live preview is in the console and on the API: lines arrive a few seconds after each pause, which suits captions and supervisors listening in. Voice agents that must answer in under a second need a streaming model, which is in development.

What does it cost?

$0.30 per hour of audio, less at volume. Summary, sentiment and redaction are priced per audio minute. New accounts get 1 free hour.

Can it follow a caller who switches languages?

Yes. Switching mid-sentence, for example between Kiswahili and English, comes out in one transcript.