Administration & GovernanceUsage & Observability

Admin Dashboard

View as Markdown

The Admin Dashboard is the executive and operations overview for AIvis. It helps administrators see whether AI work is being adopted, whether Agent and automation outcomes are succeeding, and where usage, feedback, or coverage needs attention.

Use it as a product and operations surface. For request-level investigation, use Tracing and Query History. For raw model usage and cost breakdowns, use Usage Statistics.

AIvis Admin AI Impact Dashboard design blueprint showing the full-admin entry, executive hero, KPI cards, trends, rankings, attention signals, and data boundaries

Interface Screenshots

The screenshots below show the dashboard with real example data from a Chinese deployment. Use them to locate the main areas; values vary by deployment and selected time range.

AIvis admin dashboard screenshot showing AI impact overview, verified outcomes, activity trend, and feedback quality AIvis admin dashboard screenshot showing top AI capabilities, popular Agents, active employees, and attention items

Access and Entry

ItemGuidance
EntryOpen the AIvis admin console and go to /admin/dashboard.
Required accessThe page is restricted to administrators with full admin-panel access.
Primary audienceExecutives, operations owners, AI platform administrators, and authorized auditors.
Data boundaryThe dashboard is aggregated for oversight. It is not a substitute for per-request audit trails or access-control review.

What the Dashboard Shows

AreaWhat it answers
Adoption overviewHow many eligible users were active and what share of the organization used AI in the selected period.
Verified AI outcomesHow many AI questions were answered, Agent tasks completed, and automation runs succeeded.
Activity trendDaily successful AI outcomes, split across questions, Agent tasks, and automation runs.
Model usageModel calls, success and failure counts, effective tokens, measured cost, token coverage, and pricing coverage.
Feedback qualityPositive and negative rated responses, feedback coverage, positive rate, and whether the sample is large enough.
Top capabilitiesWhich AI capabilities produced the most completed outcomes and how many unique users used them.
Top AgentsWhich Agents completed the most interaction or automation outcomes.
Active usersWhich users were active, when they were last active, and how their activity is distributed.
Needs attentionFailed or cancelled Agent tasks, failed or skipped automation runs, partial outcome coverage, low feedback sample size, incomplete token coverage, or incomplete pricing coverage.

Select a Time Range

The dashboard supports preset ranges for the last 7, 30, or 90 days, plus a custom date range. Custom ranges are interpreted in the browser’s selected timezone and must cover no more than 366 calendar days.

The page writes the selected range into the URL query string:

Query parameterMeaning
range7, 30, 90, or custom.
startISO datetime for the inclusive start of a custom range.
endISO datetime for the exclusive end of a custom range.
timezoneIANA timezone used for date boundaries and display.

The dashboard refreshes automatically while the page is visible and also provides a manual refresh button. During refresh, existing data remains visible while newer data loads.

Read the Numbers Correctly

SignalHow to interpret it
Updated timeThe latest generated timestamp returned by one of the dashboard sections.
Previous-period comparisonHidden or marked unavailable when the previous period is outside tracked coverage.
Partial outcome coverageThe selected range includes time before AI outcome tracking fully started, so totals may be incomplete.
Token coverageSome model calls may not report token metadata. Treat token totals as measured coverage, not guaranteed complete usage.
Pricing coverageSome calls may not have a known price. Unpriced calls are not zero-cost; they are missing priced coverage.
Feedback sampleFeedback quality needs at least 5 rated responses before the sample is treated as sufficient.

Drill Down

From the dashboardFollow-up page
Detailed model usageOpen Usage Statistics. Some deployments expose this only on eligible tiers.
Response quality and individual feedbackOpen Query History when query history is enabled for administrators.
Top AgentsOpen Agents to review ownership, configuration, and sharing scope.
Active usersOpen Users, Groups & Roles to review access and group boundaries.

Backend Sections

The page loads independent dashboard sections. If one section fails, the rest of the dashboard can still render and the failed card can be retried.

SectionBackend route
Overview/api/admin/dashboard/overview
Activity trend/api/admin/dashboard/activity-trend
Feedback/api/admin/dashboard/feedback
Top capabilities/api/admin/dashboard/capabilities
Top Agents/api/admin/dashboard/agents
Active users/api/admin/dashboard/users
LLM usage/api/admin/dashboard/llm-usage

Verification

  • Sign in with an administrator account and open /admin/dashboard.
  • Switch between last 7, 30, and 90 days and confirm the URL and dashboard values update.
  • Apply a custom range and confirm invalid ranges over 366 days are rejected.
  • Use the manual refresh button and confirm existing content remains visible during refresh.
  • Compare totals with Usage Statistics, Tracing, or Query History for a known test request.
  • Use a non-admin account and confirm the page and API routes are not accessible.

Troubleshooting

SymptomPriority checks
Dashboard is inaccessibleConfirm the user has full admin-panel access and is in the correct deployment or workspace.
A card shows unavailable dataRetry that section, then check the corresponding backend route and database metric query. Other cards may still be current.
Previous comparison is unavailableThe previous period may fall before tracking started or outside available coverage.
Token or cost totals look lowCheck token coverage and pricing coverage before treating the displayed totals as complete.
Feedback quality is unavailableConfirm query history and feedback collection are enabled, and that at least 5 rated responses exist in the selected period.
Activity is lower than expectedCheck the selected timezone, time range, and whether upstream bot, Agent, or automation requests reached AIvis successfully.

Security and Maintenance

  • Treat dashboard access as administrative access because it reveals user activity, Agent performance, model usage, and cost patterns.
  • Do not export or share active-user rankings outside approved operational and audit channels.
  • Review dashboard data after enabling new Agent, bot, automation, model, or tracing paths.
  • Use attention items as operational signals, then investigate the underlying requests in tracing or query history.
  • Record known coverage gaps when rolling out the dashboard to an existing deployment.