Sentinel is the quality watchtower for Insights. It surfaces the headline performance KPIs (AICSAT score, handle time, first-response time) and the per-agent breakdown against your QA scorecards. Where Drivers and Signals tell you what’s happening with your players, Sentinel tells you how well your support is responding. URL:Documentation Index
Fetch the complete documentation index at: https://setup.cevro.ai/llms.txt
Use this file to discover all available pages before exploring further.
https://app.cevro.ai/insights/sentinel
KPI Cards
Three top-level cards summarize the selected period:| KPI | What it tells you |
|---|---|
| AICSAT Score | Average AI-rated scorecard score plus the evaluation count. Only shown when AICSAT is enabled for your workspace. |
| Average Handling Time | End-to-end conversation handling time per ticket |
| First Response Time | Time from the first inbound message to the first agent reply |
Score Over Time
A line chart showing daily AICSAT score and conversation count across the selected window. Useful for spotting quality drift before it shows up in CSAT. This chart and the per-section table below it are only visible when AICSAT is enabled.Performance by Scorecard Section
A table of every AI agent’s average score against each section of your active scorecard. Section names come from your scorecard configuration in Settings → QA Scoring — common ones include Resolution Clarity, Empathy & Tone, Clarity of Instructions, Professionalism. Cells are color-coded:- Green (
≥80%) — passing - Yellow (
≥60%) — borderline - Red (
<60%) — failing
The table only includes scores from the most recent scorecard version. Archived or older versions are excluded — so rescoring history won’t pollute the average when you update a rubric.
Filters
Standard Insights filters apply:- Brand — narrow to a single brand or “All Brands”
- Date range — preset (last 7/30/60/90 days) or custom
- Advanced filters — Player metadata + Ticket metadata (Status, Agent, Browser, Platform, Language, Channel, Polarity, State, Contact type, CSAT Rating, Signal, Contact driver, Page, Scorecard score ranges)
Export
When AICSAT is enabled, an Export button in the top-right exports the granular QA evaluation data for the current filter selection — handy for offline analysis or sharing with QA managers.Scorecards & Rules
Sentinel doesn’t replace QA Scoring — it consumes it. Define your scorecards once in Settings → Scorecards and Sentinel makes them visible at scale. Scorecards are made up of sections (Accuracy, Soft Skills, Process Compliance, Resolution…) and rules within each section. Every closed conversation is evaluated against every rule, with the AI providing reasoning for each verdict.How Sentinel Plugs into the Rest of Insights
A common workflow:- Sentinel surfaces a per-section score drop on Brand X
- Filter by the failing scorecard section + Brand X
- Note which agents or contact drivers over-index in the failing slice
- Open Drivers filtered to those Needs to see whether the underlying Problem changed
- Decide: train the agent, update the AIP, or escalate to product
Tips
- Start narrow. Pick one scorecard section and one agent group to focus on first — sectional drift is easier to coach than overall score drift.
- Pair with Feedback. When CSAT looks good but a Sentinel section score is low, the player is patching for the agent (“the agent was so nice!” + the scorecard says the agent never resolved the issue).
- Use the export. For monthly QA reviews, exporting granular evaluations beats trying to summarize from the dashboard.
Related
- QA Scoring — how to define the scorecards Sentinel consumes
- Signals — agent-side detection that complements scorecard rules
- Drivers — pair scorecard issues with the Themes that produced them
- Feedback — CSAT and agent-level sentiment breakdown