Provenance
AI

AI Insights — Automated Monitoring

AI-powered analysis that surfaces anomalies, patterns, and actionable insights from your interaction data.

AI Insights automatically analyzes your interaction data and surfaces actionable findings — anomalies, patterns, security concerns, and health summaries — without you having to look for them.

Overview

AI Insights runs on a schedule (based on your plan) and examines your recent interaction data across multiple dimensions. It produces categorized insights that appear on your dashboard.

Where to Find Insights

The AI Insights card appears on your dashboard. It shows:

  • Categorized findings grouped by type (Security, Data & Activity, Notifications, etc.)
  • Severity levels — Critical, High, Medium, Low, Info
  • Token usage bar — Shows your monthly AI token consumption
  • Analyze Now button — Manually trigger an analysis (owner/admin only)

Insight Categories

CategoryWhat It Detects
SecurityLogin failures, brute force patterns, unusual access
Fraud DetectionSuspicious transaction patterns, velocity anomalies
AlertingMetric alerts that are currently firing
ComplianceBulk deletions, unusual data removal
PerformanceSlow processing, throughput drops
ReliabilityNotification delivery failures, retry storms
Data & ActivityVolume anomalies, baseline deviations
Capacity & UsageApproaching plan limits, growth rate
NotificationsQueue health, pending backlog
Service HealthOrigins going silent, new origins appearing
General HealthOverall health summaries, positive trends

How It Works

  1. Data gathering — The system collects interaction counts, top actions/types/origins, alert states, queue stats, hourly trends, user activity patterns, and notification delivery rates.
  2. Shorthand encoding — Data is compressed into a token-efficient format to maximize analysis depth within your token budget.
  3. AI analysis — The data is sent to the AI model with a specialized monitoring prompt that understands provenance data patterns.
  4. Insight storage — Results are stored and displayed on your dashboard, grouped by category and sorted by severity.

Analysis Schedule

Analysis frequency depends on your plan:

PlanFrequency
Free / BetaNot available
ProEvery hour
BusinessEvery 15 minutes
EnterpriseEvery 5 minutes

You can also trigger analysis manually with the Analyze Now button.

Token Budget

AI Insights shares the monthly token budget with Provi:

PlanMonthly Tokens
Pro50,000
Business200,000
EnterpriseUnlimited

Each analysis typically uses 500–2,000 tokens depending on data volume. When the budget is exhausted, both AI Insights and Provi pause until the next month.

Managing Insights

  • Dismiss — Hover over an insight and click × to dismiss it. Dismissed insights won't reappear.
  • Expand categories — Click a category header to expand/collapse its findings.
  • Severity sorting — Categories with higher-severity findings appear first automatically.

What Makes a Good Insight

AI Insights is designed to surface genuine anomalies, not noise:

  • Compares current values against 7-day baselines
  • Only flags alerts that are actually firing (not OK alerts)
  • Identifies origins that have gone silent when they're normally active
  • Tracks notification delivery rates and flags drops below 90%
  • If everything looks normal, returns a single "all healthy" summary
  • Maximum 7 insights per analysis to avoid alert fatigue