Conversation Intelligence
From a conversation to a decision someone can trace.
The Workspace brings a conversation's transcript, cited findings and themes together, so a decision can be traced back to the line.
A quality lead, Priya Nair, reviews conversations. Beside her is an insights board: pick a theme to see its mentions over the last seven days, and hover or focus a day for its count. Below, a support call is scored against the rubric. Open a finding to see the transcript line it cites and whether that evidence is Exact or Unverified, then set the result and the review state. A person makes the call, and the score updates. The demo plays by itself until you interact.
Themes · last 7 days
38mentions
RisingCustomers ask what a charge is before they dispute it.
Call review · Support
Daniel Okafor with Maya Chen0 of 4 reviewed63A person makes the call. The review state stays with the finding.
Current limits
- Recorded calls are not yet processed automatically after upload.
- Summaries, themes and scores have not been through a quality evaluation on labelled data.
- AI scoring of support calls is not available (see Support quality review).
Illustrative interface — fictional demonstration data.
Call review
Open the call, the score and the line together.
Open a conversation's transcript with its findings cited to the line. On support calls, a manager marks each call not reviewed, in progress or reviewed.
- Transcript with speakers and timestamps
- Findings cited to a line
- Review state on support calls, set by a person
Rubrics
Your rubric, set by your organisation.
Criteria come from a rubric your organisation configures. On support calls a person reviews each behaviour against it; AI scoring of support calls is not available.
- Configurable rubrics
- Behaviours reviewed against the rubric
- Support review done by a person
Themes
Themes you track, with the line under them.
Track the themes that matter and see the utterances behind each count.
- Themes with the utterances beneath them
Theme · last 7 days
Billing clarity Rising
7 mentions on Sun · 38 this week
Customers ask what a charge is before they dispute it.
Validation Queue
A finding waits for a person.
Research findings wait in a queue until a researcher confirms the cited turn.
- Confirm each cited turn
- Findings keep their citation
- The decision stays with the finding
Ask
Ask about a meeting.
Ask a question about one meeting in the desktop app and get an answer from that meeting's transcript. Search across all of your conversations is not available.
- One meeting at a time
- Works in the desktop app
Coaching
Coach from real moments.
Managers prepare coaching from the moments in real conversations, and each person has a scorecard of their own.
- A team view for managers
- A scorecard for each person
What it is built around
Every finding cited
A finding points at the line and the speaker that produced it.
Rubrics you configure
Rubrics are set by your organisation.
A person decides
Support calls are reviewed by a person, and the review state is kept.
Explore Conversation Intelligence.
The rest of the platform.
Common questions
Which conversations can it analyse?
Spoken conversations captured with the desktop app. Recorded calls are not processed automatically after upload.
Who writes the rubric?
Your organisation. Rubrics are configured by the customer, not supplied by Tessavra.
What does Exact or Unverified mean?
On a finding tied to audio, a label shows how well the line was matched to the recording. Unverified means a person should listen before treating it as the record. A finding tied to a text message is located by the message itself.
Have summaries, themes and scores been checked for accuracy?
Not yet. They have had no quality evaluation on labelled data.
Talk to us about this.
Conversation Intelligence is in development. Tell us what you want it to do and we will say plainly where it stands.
Register interest