Analytics answers and opening briefings¶
Status: Current | Last reviewed: 2026-10-02
Open an Analytics Agent under Agents, then select Analysis. You can also configure response components and the opening briefing under Experience. Both places edit the same settings. Changes save collaboratively to the draft; use Preview to test, then Publish to make them available to readers.
Choose the visuals¶
Under Charts and cards, select Choose components & composition. The wide editor lets you choose:
- Built-in charts, tables, single metrics, metric groups, conversation cards, search results, trend reports and dataset summaries.
- The chart styles the analyst may use.
- Published component catalogues and specific components within them.
- Answer length, maximum visual count and whether report details start expanded.
- A composition prompt describing when to choose each component and how to arrange the answer.
For example, select a summary card from a published catalogue, then reference it from the composition prompt: “Use this card for the headline finding, followed by one chart and a short explanation.” Type @ in the prompt editor to select enabled tools or selected components.
Catalogue choices retain their published version. If a newer version becomes available, the picker offers an explicit upgrade while keeping matching component names. Selecting a visual does not enable that catalogue's business actions or data integrations. Choose components that display supplied evidence.
The default built-in components preserve the original CallSight presentation. Disabling a card still allows the analyst to retrieve its underlying evidence and explain it in prose. Keep report answers concise avoids repeating a complete report in the surrounding message; the component expansion setting controls whether the report itself starts open.
Configure the analytical capabilities¶
Prepared insights under Analysis controls which trend levels and summary types chat can read, the number listed per request and whether to prefer prepared answers. To change how reports are built, use Preparation → Prepared insights: that is where the hierarchy, coverage, generation models and summary prompts live.
Intent drift offers two analyses:
- Reported vs inferred classification finds disagreement between source intent labels and transcript annotations.
- Intent share over time compares adjacent windows ending today. Choose the intent lens, window duration, minimum contact count and minimum percentage-point change.
The result limit applies after calculating population shares. Missing windows are reported as insufficient data. Drift does not automatically add new intents to your hierarchy.
What-if scenarios lets you choose the available scenario types, historical window, cost per minute, currency and containment cap. The question supplies its target, such as “What if we deflected 15% of refund contacts?” The analyst asks for missing targets. No billing category or location is preselected.
Process-change reversal needs an area and date, plus measured handling time on
both sides of that date. Area uses the operational hub field. Scenario answers
disclose their assumptions and are illustrative estimates.
Configure the opening briefing¶
Enable Opening briefing in Analysis or Experience.
| Setting | What it does |
|---|---|
| Briefing generation | Pre-generate published agents on a schedule, or generate when opened if the cached briefing has expired. |
| Refresh interval | Choose how often to regenerate; this is an interval, not a specific local time. |
| Operating metric cards | Always include chosen measured KPIs, let the prompt choose, or omit the fixed metric section. |
| Measured KPIs | Select volume, containment, average handling time and/or transfer/escalation rate. Unknown values are omitted. |
| Metric window | Use a completed report period or a rolling window ending today. |
| Maximum secondary findings | Choose zero through six additional findings. |
| Expand secondary findings initially | Control their initial open/closed state. |
| Separate briefing model | Choose a provider, service, integration and model; otherwise use the live analyst model. |
| Briefing prompt | Edit the full briefing task and select available multimodal components with @. |
Expand editor opens a larger prompt workspace. The briefing uses your selected response components and composition guidance. Choose prompt-selected metrics when you want the model to decide which cards best explain the current data instead of always showing the same KPI group.
Published scheduled briefings are generated by the Analytics worker. Draft previews generate when opened. A ready briefing is reused within the refresh window; configuration or selected-data changes create a new one. This schedule does not upload new files or rebuild prepared reports.
Use Retry briefing after correcting a failed model or data binding. Existing chat sessions retain their original configuration and dataset. Start a new chat to test a changed published configuration.