Audio Analysis · Call analytics

Call Analytics and Quality Assurance

Every call your Hamsa agents handle is recorded, transcribed and summarized, with the outcomes you define pulled out as data. Dashboards, exports and live listen-in turn those calls into QA, and batch transcription covers recordings from your other systems.

  • Transcripts with speaker labels
  • Outcomes as structured JSON
  • CSAT, NPS, FCR and escalation rate
  • Excel export and analytics API
  • Silent listen-in

Capabilities

Arabic call analytics for every agent call

01

A history of every call

Each call is logged with its time, agent, channel, duration, status and cost. Filter by date, status, channel or agent, search by call ID or number, and export as CSV.

02

Transcripts and recordings

The conversation view shows who said what, with timestamps, next to the call recording and the call’s logs.

03

Summary and resolution

When a call ends, it gets an AI-generated summary and a resolution status: resolved or escalated.

04

Outcomes you define

Describe the data you want from each call as a JSON schema, in a visual builder or a JSON editor. Results appear on the call and go to your systems by webhook.

05

Sentiment, when you want it

Sentiment is off by default and switched on per agent. Read it as a directional signal, not an exact measurement.

06

Listen in live

Join an active call as a silent listener, follow the transcript, send the agent instructions or end the call.

QA routine

A call monitoring routine your QA team can run every week

How a QA lead uses the tools above, from deciding what to measure to changing the agent.

  1. 01

    Decide what to measure

    Define the outcome fields your team reports on, such as an appointment booked or a payment promised, and turn on sentiment for the agents where it helps.

  2. 02

    Watch live calls

    Supervisors whose role allows it listen silently and send the agent instructions when a call drifts.

  3. 03

    Review finished calls

    Filter by status, channel or agent to pick the calls to review, then read each one’s summary, transcript, logs and outcome.

  4. 04

    Track the trends

    Dashboards show volume, duration, hand-offs to people, latency, CSAT, NPS, first-call resolution and escalation rate.

  5. 05

    Report and adjust

    Export the dashboards to Excel or pull the numbers through the analytics API. Then change the agent’s prompt, flow or knowledge, and compare.

Metrics

Agent performance analytics, view by view

Dashboards, call history and outcomes cover calls handled by Hamsa agents, on the phone and on the web. Call history also lists chat sessions.

ViewWhat it shows
OverviewLive and total sessions, average and total duration, calls under 30 s, 30 to 120 s and over 120 s, words per AI response, share forwarded to a human agent
PerformanceSpeech recognition, language model and speech synthesis times, latency, error rate
Satisfaction & OutcomeCSAT, NPS, first-call resolution, escalation rate, sentiment distribution (positive, neutral, negative)
Call detailsOverview, conversation with speaker labels and timestamps, logs, outcome
Call History exportCSV
Dashboard exportExcel, with Overview, Performance and Satisfaction sheets
Analytics APIOverview, performance, satisfaction and logs; conversations export as Excel
WebhooksCall events and extracted outcomes, sent over HTTPS with bearer token support

Other recordings

Arabic conversation analytics for recordings from other systems

These run one recording at a time. Their results stay with the transcript and do not feed the agent dashboards.

01

Transcribe with the Conversational model

Send any recording by URL to batch transcription with Hamsa-Conversational-V1.0, built for calls, meetings and dialogues. It separates speakers, adds punctuation and word timestamps, and detects the Arabic dialect.

02

Summaries and key topics

Turn the transcript into a summary, key topics with bullet points, keywords and an FAQ.

03

Your own write-up

Custom prompts produce the document your team needs from each transcript, in the format it already uses.

Access

Who sees which calls

01

Projects keep calls apart

Each project has its own agents, numbers, call history and members, so one team’s calls stay out of another’s view.

02

Roles decide access

Owner, Admin and User roles, with custom permissions per member, control who reads call history and who can monitor live calls.

03

Recordings by signed URL

The API returns each call’s recording as a signed URL, and access to call history follows the member’s role.

FAQ

Questions about call analytics and QA

Which calls does Hamsa call analytics cover?

+

The dashboards, call history and outcomes cover calls handled by Hamsa agents, on the phone and on the web, and call history also lists chat sessions. For recordings from other systems, batch transcription and AI Docs produce a transcript, summary and key topics per recording, separately from the dashboards.

Do transcripts separate the agent from the customer?

+

Yes. Agent call transcripts identify each speaker and carry timestamps. For other recordings, the Conversational transcription model separates speakers automatically and returns word-level timestamps.

Is sentiment analysis included?

+

Yes, for agent calls, but it is off by default and you turn it on per agent. The dashboard then shows the split of positive, neutral and negative calls. Use it as a directional indicator rather than an exact measurement.

Can we pull our own fields out of each call?

+

Yes, with outcome parameters. You describe the fields as a JSON schema of strings, numbers, booleans, enums, objects and arrays, and each call’s results appear in call history and go to your systems by webhook.

Can we get the data into our own reporting?

+

Yes. Call history exports as CSV, and the dashboards export to Excel with one sheet each for overview, performance and satisfaction. The analytics API returns the same metrics and exports conversations as an Excel file.

Can supervisors listen to live calls?

+

Yes. Team members whose role allows it can join a live call as a silent listener, follow the transcript, send the agent instructions and end the call. The caller is not notified, so make sure your monitoring follows the laws that apply to you, for example with a monitoring disclosure.

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.