> ## Documentation Index
> Fetch the complete documentation index at: https://help.nops.io/llms.txt
> Use this file to discover all available pages before exploring further.

# AI Anomalies

> AWS Cost Anomaly Detection results filtered to AI services, with severity, impact, and root causes, plus Clara's anomaly analysis.

**Anomalies** shows AWS-detected cost anomalies narrowed to AI services — Bedrock, SageMaker, and related services — so an unexpected spike in model spend does not have to be found by scrolling through every anomaly in the account. Below it, **AI Analysis** carries Clara's scheduled anomaly-detection prompts.

<Note>
  The AWS section appears only when you have an active AWS integration. Without one, the page shows the **AI Analysis** section on its own.
</Note>

## Get started

1. Sign in to [nOps](https://clara.nops.io/dashboard).
2. In the sidebar, open **AI → Anomalies**.
3. Sort out what matters using the summary cards, then work the table.

The sidebar entry carries a badge with the number of open anomalies.

## Key features

### Summary cards

| Card              | What it counts                                   |
| ----------------- | ------------------------------------------------ |
| **AI anomalies**  | AWS-detected AI service anomalies in the window. |
| **Total impact**  | Aggregate anomaly impact.                        |
| **High severity** | Critical or high severity anomalies.             |

All three describe active anomalies inside your selected date window — not an account-wide total — so the numbers always match the table below.

### AWS Detected Anomalies

| Column                             | What it shows                   |
| ---------------------------------- | ------------------------------- |
| **Source**                         | Where the anomaly was detected. |
| **Severity**                       | Severity level.                 |
| **Start** / **End** / **Duration** | When it ran.                    |
| **Cost impact**                    | Dollar impact.                  |
| **Impact %**                       | Impact as a percentage.         |
| **Trend**                          | A sparkline of the spend shape. |
| **Service**                        | The AWS service involved.       |
| **Root causes**                    | What AWS attributed it to.      |

Filter the window with **Last 7 days**, **Last 30 days**, **Last 90 days** (the default), or **All time**.

### Active and Ignored

Toggle between **Active** and **Ignored**. The **Ignored** tab carries a count of how many anomalies in the current window you have set aside.

Select rows with the checkboxes to ignore them in bulk, or ignore everything currently in view in one action. Anything ignored can be brought back from the **Ignored** tab. This is how you retire a known, explained spike so it stops competing for attention with new ones.

### AI Analysis

The lower section shows Clara scheduled prompt results for **AI Services** in the **Anomaly Detection** category — narrative analysis to go with the detected rows.

## FAQs

<AccordionGroup>
  <Accordion title="Why is the AWS section missing?">
    It renders only when an AWS integration is connected and active. AI Analysis still shows without one.
  </Accordion>

  <Accordion title="Which services count as AI here?">
    Anomalies from AWS Cost Anomaly Detection filtered to AI services, including Bedrock and SageMaker.
  </Accordion>

  <Accordion title="Does ignoring an anomaly delete it?">
    No. It moves to the **Ignored** tab and stops counting toward the summary cards. You can unignore it at any time.
  </Accordion>

  <Accordion title="Why doesn't the ignored count match what I expect?">
    The count reflects only ignored anomalies inside the selected date window, so widening the range can reveal more.
  </Accordion>
</AccordionGroup>

## Related

* [AI Recommendations](/ai/recommendations)
* [Tracking](/ai/tracking)
* [Insights](/inform/insights)