Complete AI Training
Sign inGet my AI kit

Your job's AI kit

Get your AI kit

Tell us who you are and what you do. We show you your kit right away and email you the link: skills, prompts, AI agents, MCP servers and courses for your job.

500+ jobs ready, and we make a kit for any other job. No payment needed to look.

AI agent for business intelligence analysts

Anomaly Watch Agent

Real, explained metric anomalies surfaced the day they happen

Anomaly Watch Agent: what goes in, what the agent does and what you get

What it does

Important metrics like sales, sign-ups or error rates can shift suddenly, and the monthly report is too slow to catch it. Each day this agent checks each watched metric against its normal range, allowing for weekly and seasonal patterns. It reads your calendar of promotions and launches so planned events are not flagged. When a value falls outside the range, it checks the data feed and related metrics to tell a real change from a data problem. For a real anomaly it breaks the metric down by segment, region or channel to find what drove it, and drafts a short alert with the likely explanation. If it cannot tell a real change from a glitch, it says so instead of raising a false alarm. You approve alerts sent to business owners.

How it works

Follow the arrows from top to bottom. The orange dashed arrow is the loop: when a check fails, the agent goes back and tries again.

Start and resultWhat it doesA check on its own workWaits for your OKGoes back and retries
Yes, continueYes, continueApprovedNoNo 1 STARTS WHEN Daily metric check 2 USES A TOOL Pull each watched metric and its normal range 3 CHECKS THE RESULT Is any value outside the normal range? If not: log a normal result and stop. Back to step 2. 4 USES A TOOL Check the data feed and related metrics 5 CHECKS THE RESULT Is it a real change rather than a data glitch orplanned event? If not: report a possible data issue or note the plannedevent. Back to step 4. 6 DOES Gather context and draft the alert 7 YOU APPROVE Analyst approves alerts to business owners 8 RESULT Anomaly log and any approved alerts
Read the steps as a list
  1. Daily metric check
  2. Pull each watched metric and its normal range
  3. Is any value outside the normal range?If not: log a normal result and stop. Back to step 2.
  4. Check the data feed and related metrics
  5. Is it a real change rather than a data glitch or planned event?If not: report a possible data issue or note the planned event. Back to step 4.
  6. Gather context and draft the alert
  7. Analyst approves alerts to business ownersThe agent waits here for your OK.
  8. Anomaly log and any approved alerts

How it decides

It flags values outside the seasonally adjusted normal range and confirms they are real, not data errors or planned events, before alerting.

  • Adjust the normal range for weekly and seasonal patterns
  • Treat a feed problem as a data issue, not an anomaly
  • Skip changes explained by the events calendar

Make it yours

Every agent is a starting point. You choose these settings for your own situation.

  • Metrics to watch and their ranges
  • Sensitivity of the alert threshold
  • Events calendar source
  • Who receives alerts

What keeps you in control

It always asks you first

  • Sending alerts to business owners

Hard limits

  • Does not act on the metrics, only reports
  • No alerts sent without approval

It stops when

  • Done: anomalies explained or none found
  • Stop: metric history is too short for a range

Set it up

We guide you through the set-up, step by step

Members get the full set-up guide for this agent. No technical skills needed: you copy, paste and upload.

10 minto set it up in your AI
5 AIsChatGPT, Claude, Copilot, Gemini, Grok
  • One set of instructions to paste into your AI, with the clicks for ChatGPT, Claude, Microsoft 365 Copilot, Gemini and Grok
  • The agent then walks you through connecting your own data, one source at a time
  • A downloadable copy with the flow chart, the rules and the full guide
Get access to this agent

An example run

What happensOn Tuesday, April 9, sign-ups at Fieldstone Fitness dropped 40%. The feed was healthy and no promotion had ended. The first breakdown by region showed nothing unusual, so the driver check failed. The agent broke it down by device next and found mobile sign-ups near zero. It drafted an alert naming the mobile app. The product manager approved sending it, and a broken button was fixed that afternoon.

More agents for business intelligence analysts