Solutions · Quality assurance

Call Center Quality Assurance

Every call your Hamsa voice agents handle reaches QA already recorded, transcribed with speaker labels, summarised and reduced to the outcomes you define. Recordings from your other systems can be transcribed in Arabic and summarised as well.

  • Transcripts with speaker labels
  • Outcomes you define, as JSON
  • CSAT, NPS, FCR and escalation
  • Silent live listen-in
  • Excel and API exports

QA today

How manual call QA usually works

Most QA teams follow a version of the same routine. It is the starting point for what changes when the agent on the call is a Hamsa agent.

01

Pick a sample

Reviewers choose a few calls per agent, because listening to a call takes as long as the call itself.

02

Listen and mark

Each call is played back and checked against an evaluation form, with notes on what went well and what did not.

03

Coach and repeat

Supervisors turn the notes into feedback, then review a new sample to see whether anything changed.

Automated workflow

Automated call QA for every call a Hamsa agent handles

What happens to each agent call after it ends, and what your team can do while it is live.

  1. 01

    Recorded and transcribed

    Every call is recorded and transcribed, with each line labelled by speaker and timestamped.

  2. 02

    Summarised

    An AI summary and a resolution status, resolved or escalated, are added when the call ends. Sentiment is added too for agents where you turn it on.

  3. 03

    Outcomes extracted

    Fields you define, such as reason for call, whether it was resolved and whether follow-up is needed, are extracted as structured data and sent by webhook.

  4. 04

    Found in call history

    Filter calls by date, status, channel and agent, or search by call ID or phone number, then open the transcript, logs and outcome.

  5. 05

    Watched live

    Supervisors can listen to an active call silently, follow its transcript, send the agent instructions or end the call.

  6. 06

    Fed back into the agent

    Transcripts show where the flow guessed wrong, which questions the knowledge base could not answer, and where callers asked for a person. That is what you fix and test next.

Other recordings

Call center analytics in Arabic for recordings from other systems

Calls handled by your human team or another platform do not appear in the agent dashboards. You can still transcribe and summarise them as separate jobs.

01

Conversational transcription

Send any recording by URL to the Hamsa-Conversational-V1.0 model, built for calls, meetings and dialogues. It separates speakers and returns word timestamps and punctuation.

02

Dialect detected for you

Speech recognition detects the Arabic dialect automatically and handles speakers who mix Arabic and English.

03

Summaries and key topics

Generate a summary, key topics and keywords from each transcript, in Arabic, English or both, or request your own output with a custom prompt.

Specifications

What QA can see on each agent call

Everything in this table applies to calls handled by Hamsa agents. Sentiment is off by default, enabled per agent, and best read as a directional indicator.

ItemDetail
Call recordTime, agent, channel, duration, status and cost
TranscriptLabelled by speaker, with timestamps
RecordingKept for every call, shared through a signed URL
After the callAI summary and resolution status (resolved or escalated)
SentimentPositive, neutral or negative, per agent when enabled
OutcomesYour JSON schema: string, number, boolean, enum, object and array fields
DashboardsVolume and duration, forwarded to human %, ASR, LLM and TTS times, error rate, CSAT, NPS, first-call resolution, escalation rate
ExportsCall history as CSV, dashboards as Excel, conversations export and analytics API
Live monitoringSilent listen-in, live transcript, instructions to the agent, end the call

Privacy and access

Arabic call monitoring with access you control

01

Access by role

Call history access is controlled by role. Owners and admins can monitor live calls; the user role cannot, and permissions can be adjusted per member.

02

Separate projects

Each project keeps its own agents, numbers, call history and members, so teams and environments do not see each other’s calls.

03

Signed recording links

Recording links returned by the conversation API are signed URLs, not open links.

04

Tell callers

Live listening is silent, so the caller is not notified. Make sure your monitoring follows the laws that apply to you, such as telling callers they may be monitored.

FAQ

Questions about AI call center quality assurance

Which calls does Hamsa QA cover?

+

The call history, dashboards, outcomes and live monitoring cover conversations handled by Hamsa agents, and call history includes phone, web and chat sessions. Recordings from other systems can be transcribed and summarised as separate jobs, but they do not appear in the agent dashboards.

Can Hamsa transcribe recordings from our existing call center?

+

Yes, as transcription jobs. Send each recording by URL to the Hamsa-Conversational-V1.0 model, which separates speakers, adds word timestamps and punctuation, and detects the Arabic dialect automatically. You can then generate a summary, key topics and keywords from each transcript.

Is sentiment analysis included?

+

For calls handled by Hamsa agents, sentiment can be enabled per agent; it is off by default. The dashboard then shows the split between positive, neutral and negative calls. Treat it as a directional indicator, not an exact measurement.

Can supervisors listen to live calls?

+

Yes. Authorised team members can join an active agent call as a silent listener, follow the live transcript, send the agent instructions or end the call. The user role does not have live monitoring. Make sure your monitoring follows the laws that apply to you.

How do we get QA data into our own reporting?

+

Export call history as CSV and dashboards as an Excel file with overview, performance and satisfaction sheets. The analytics API returns the same figures, a conversations export returns an Excel file, and outcomes reach your systems by webhook after each call.

Who can see call recordings and transcripts?

+

Access to call history is controlled by role, and each project keeps its own calls and members. Recording links from the API are signed URLs. Permissions can also be set for each member individually.

Keep reading

Partner ecosystem

Trusted Across the
Technology Ecosystem

Hamsa works with leading technology companies, integrators, and solution providers to bring enterprise Voice AI to organizations across the region.