Skip to content

Conversation Intelligence

Conversation Intelligence brings together every conversation the assistant has had with prospects, applicants and students across channels. It opens from the Enrollment Overview and the Enrollment Agent page, and its tabs are kept in the page address.

  • You need view conversation logs. Deciding on queued replies and escalations needs manage review queue. Seeing unmasked contact details needs unmask pii (Admin only by default).

Search by participant or summary, filter by channel and by source (Channels, Student Success, Enrollment), and Load more. Each transcript shows which tools the agent used. Contact details are masked for roles without unmask pii, and a banner says so.

Assistant replies waiting for a person, with counts of pending, held and deferred items. Nothing here has reached the student yet. See Review Queue for how to decide.

A card per agent: deployed channels, enabled skills, total conversations, average accuracy, resolution rate and active sessions.

A read-only library of the skills agents can use, filtered by category and whether they are in use, with the agents that use each and its use cases. Skills are switched on in the Orchestration Hub.

Conversations an agent flagged for a person, in Needs a decision, In review and Resolved. Sort by priority or newest, and search the queue. Each row shows who, why it escalated, the recommended action, priority and source; some are marked Auto-escalated.

The drawer shows the recommended action and the transcript, and offers Assign to (a name or email), Mark in review, Resolve and Reopen, with Previous and Next (also the left and right arrow keys). A Handled today strip lets you Release an item back to the queue.

Aggregated analytics across all agents and channels. Generate Report writes a conversation analytics report from the figures. Manage Cards chooses and orders the cards: health metrics, top themes, most invoked skills, peak hours, channel usage and time-of-day distribution.

Enrollment → Enrollment Agent lists the agents that answer prospects and applicants and where they are deployed, with a search and category chips. Each agent’s page has a real Chat with Agent test (optionally with an applicant’s email to look up their application), a QA Test, and a conversations table with a transcript drawer showing sentiment, detected intents, extracted entities and whether a review is required.