# AI Insights

Source: https://docs.qualflare.com/docs/ai/insights · Last updated: 2026-10-27

AI Insights in Qualflare flags flaky and redundant test cases after each uploaded launch. Review each recommendation, then accept or dismiss it.

In Qualflare, **AI Insights** lists recommendations that flag [flaky](https://docs.qualflare.com/docs/reference/glossary#flaky-test) and redundant test cases in a project; Qualflare checks the cases of every launch you upload, and a recommendation never changes or deletes a case.

| Fact          | Detail                                                                                                                                                                                                               |
| ------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Where         | **AI Insights** under **Analysis** in the project sidebar                                                                                                                                                            |
| Plans         | (available on Core, Scale, Enterprise). On other plans the page shows **Upgrade to Unlock**                                                                                                                          |
| Who           | View: (requires the Viewer role or higher in the project). **Accept** and **Dismiss**: (requires the Editor role or higher in the project)                                                                           |
| Allowance     | The redundancy check uses one AI case analysis per uploaded launch; see [Limits and plan availability](https://docs.qualflare.com/docs/ai/insights#limits-and-plan-availability)                       |
| Notifications | **Flaky Case Insights** and **Redundant Case Insights** in the project's **Notifications** settings ([Insight notifications](https://docs.qualflare.com/docs/administration/project-settings#insight-notifications)) |

## How AI Insights works

AI Insights shows what Qualflare's case analysis finds in each launch you upload:

* **Flaky:** flaky detection checks each case's pass/fail history against the project's [Quality settings](https://docs.qualflare.com/docs/administration/project-settings#quality-settings) and uses no AI, so it runs even with project AI off (available on Starter, Core, Scale, Enterprise, Open Source). It also opens a flaky defect ([How defects are created](https://docs.qualflare.com/docs/defects#how-defects-are-created)) and resolves it when the case recovers.
* **Redundant:** the redundancy check compares cases with AI. It needs project AI on and the monthly case analysis allowance.
* **Low Value:** Qualflare doesn't create Low Value recommendations on any plan, so the **Low Value** tab is empty.

On Starter and the Open Source plan, flaky detection still runs and opens flaky defects, but the AI Insights page shows **Upgrade to Unlock**, and no insight notification is sent.

Every 15 minutes, Qualflare sends project members one in-app notification per type for the new recommendations, while that insight notification is on (the default). A redundant-case notification has a one-time **Create Defect** button.

*Figure: AI Insights shows active recommendations by default, and each tab shows how many it holds.*

## Recommendation types

Each AI Insights recommendation carries a **confidence** score and has one of three types:

| Type          | What it flags                                                        | The detail dialog shows                                                       |
| ------------- | -------------------------------------------------------------------- | ----------------------------------------------------------------------------- |
| **Flaky**     | A pass/fail history that looks like flakiness, not a real regression | The **Flakiness score**                                                       |
| **Low Value** | None is created                                                      | Nothing                                                                       |
| **Redundant** | Coverage another case already provides                               | The **Similarity** percentage and the **Similar case**, which opens that case |

*Figure: A Redundant recommendation names the similar case, so you can compare them before you accept or dismiss it.*

## Open AI Insights

The AI Insights page has **All**, **Flaky**, **Low Value** and **Redundant** tabs, and the last three show how many active recommendations they hold. Its columns are **Case**, **Suite**, **Type**, **Confidence**, **Detected** and **Status**; Confidence and Detected sort. It shows **Active** recommendations by default: use the **Status** filter for Accepted, Declined or Stale ones, or search by case name. A row's menu has **View details**, plus **Accept** and **Dismiss** on active rows.

## Review a recommendation

To review an AI Insights recommendation, compare its evidence, then record your decision:

1. Click a row, or choose **View details** from its menu, to open the detail dialog.
2. Compare the evidence: the **Flakiness score** of a Flaky recommendation, or the **Similar case** of a Redundant one.
3. Click **Accept** to record your decision, or **Dismiss** to keep the case as it is. Both need (requires the Editor role or higher in the project); the app shows the buttons to Viewers too, but refuses their decision. You should see `Recommendation accepted.` or `Recommendation declined.`
4. Change the case yourself, because AI Insights doesn't: archive or edit it in **Browse**. For a Redundant pair, keep one case and archive the other.

## Recommendation lifecycle

A recommendation starts **Active** and moves to one of:

* **Accepted**: you clicked **Accept**. This records your decision only. Qualflare doesn't archive or edit the case, so do that yourself. An accepted recommendation turns **Stale** if the finding stops being true.
* **Declined**: you clicked **Dismiss** to keep the case as it is. Qualflare doesn't raise that recommendation again.
* **Stale**: the recommendation no longer applies, for example because the case's run history changed, or because the case (or the case it duplicates) was archived. Unarchiving the case doesn't restore it right away. The next analysis raises it again if it still applies.

Insights never change or delete a test case. You decide with **Accept** or **Dismiss**, and a recommendation can also turn Stale on its own.

## Limits and plan availability

Plans with AI Insights (available on Core, Scale, Enterprise) include these AI case analyses per month: Core 1,000, Scale 4,000, Enterprise 200,000.

When the allowance is used up, the redundancy check is skipped for new launches until the monthly reset, and flaky detection continues. AI Insights comes with the plan, not a setting an admin switches on, and every project shows it in the sidebar. See [AI usage limits](https://docs.qualflare.com/docs/ai/limits#case-analysis).

*Figure: On Starter and the Open Source plan, AI Insights shows an upgrade prompt.*

## Related

* **Guides:** [Failure Clusters](https://docs.qualflare.com/docs/ai/failure-clusters) · [Create, triage and close defects](https://docs.qualflare.com/docs/defects/manage-defects)
* **Reference:** [Quality settings](https://docs.qualflare.com/docs/administration/project-settings#quality-settings) · [Glossary: Flaky test](https://docs.qualflare.com/docs/reference/glossary#flaky-test) · [AI Features settings and limits](https://docs.qualflare.com/docs/ai/limits)
* **CLI/API:** [MCP server reference](https://docs.qualflare.com/docs/api/mcp-reference). AI Insights has no CLI command or MCP tool