> For the complete documentation index, see [llms.txt](https://developer.harness.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://developer.harness.io/continuous-delivery/use-continuous-delivery/verify-deployments/configure-cv/configure-verify-step-with-ai/ai-verify-logs-feedback.md).
# Give feedback on AI Verify (v2) log analysis
Give natural-language feedback on AI Verify (v2) log analysis so Harness remembers your corrections and applies them to the next verification for the same service and environment.
{% hint style="info" %}
This feature is in Beta[^1] and requires enablement. [View all early access features.](/release-notes/features.md)
{% @harness-feature-flag-access/request-feature-flag-access module="continuous-delivery" pagePath="continuous-delivery/use-continuous-delivery/verify-deployments/configure-cv/configure-verify-step-with-ai/ai-verify-logs-feedback" featureFlags="%5B%7B%22name%22%3A%22AI%20Verify%20logs%20feedback%22%2C%22key%22%3A%22CDS\_CV\_AI\_VERIFY\_LOGS\_FEEDBACK\_ENABLED%22%2C%22status%22%3A%22BETA%22%2C%22description%22%3A%22Adds%20a%20feedback%20composer%20and%20management%20panel%20to%20the%20AI%20Verify%20(v2)%20Analysis%20view%2C%20so%20the%20log%20analysis%20flow%20can%20learn%20from%20your%20feedback%20and%20improve%20the%20outcomes%20and%20explanations%20of%20later%20verifications.%22%7D%5D" %}
{% endhint %}
You can tell Harness, in plain language, whether it got a cluster right or wrong, either while the log analysis is still running or after it completes. Harness stores that feedback against the run's service and environment. It never applies to the run you gave the feedback on; the item shows a **Not Applied** label until the next verification for that same service and environment applies it.
This is different from [Event preference or log feedback](/continuous-delivery/use-continuous-delivery/verify-deployments/cv-results/log-feedback.md), which sets a fixed risk level on a log cluster for the AI Verify (v1) step. AI Verify (v2) logs feedback accepts free text instead of a preset risk value, and it only affects the LLM-driven steps, such as anomaly classification and root cause analysis, not the underlying statistical clustering.
***
## What you will learn from this topic
* **Give feedback on a log analysis:** How to attach natural-language feedback to a log cluster, or to a run as a whole, from the Analysis drawer.
* **Manage feedback for a service and environment:** How to review, correct, or archive feedback items from the Feedbacks Management panel.
***
## Before you begin
* **Completed AI Verify (v2) log analysis:** Run a pipeline with an AI Verify (v2) step that analyzes a log health source. Go to [Configure the AI Verify (v2) step](/continuous-delivery/use-continuous-delivery/verify-deployments/configure-cv/configure-verify-step-with-ai/ai-verify.md) to add it.
* **Resolved service and environment:** The executed stage must resolve a single Continuous Delivery service and environment. Stages that do not deploy a service, or multi-service rollup views, do not expose feedback. Go to [Troubleshooting](#troubleshooting) if the composer stays disabled.
* **Pipeline permissions:** Feedback uses the same permissions as AI Verify. View or Execute permissions on the pipeline let you see feedback; whoever can act on AI Verify can also add, correct, or archive it. Go to [RBAC in Harness](https://developer.harness.io/docs/platform/role-based-access-control/rbac-in-harness) to configure roles.
***
### Give feedback on a log analysis
Give feedback from the Analysis drawer of any AI Verify (v2) execution, whether the log analysis is still running or has already completed.
1. In your Harness project, go to **Deployments** -> **Pipelines**, then select the pipeline and execution that ran the AI Verify (v2) step.
2. On the step, click **View Analysis** to open the Analysis drawer.
3. Optionally, select a point on the scatter plot, or click **Give feedback** on a row in the **Events List** tab. Either action attaches that log cluster to your feedback as a removable tag. Skip this step to leave feedback as a general comment on the run rather than on a specific cluster.
4. In the feedback box, enter your correction in natural language, for example "Starting scheduler logs are expected on every boot" or "A 500 on checkout must always be treated as critical."
5. Click the submit arrow.
An inline confirmation reads **Feedback added.** Harness does not apply this feedback to the current run. It takes effect starting with the next verification for the same service and environment.
{% hint style="info" %}
**Feedback only softens a verdict today**
In this release, feedback can only downgrade a cluster's classification, for example from Critical Regression to Acceptable Noise. It cannot raise a classification, so feedback can never push a healthy run into an unhealthy state. Raising a classification through feedback is on the roadmap for a future release.
{% endhint %}
### Manage feedback for a service and environment
Click **Manage Feedbacks** in the feedback composer to open the Feedbacks Management panel. Because feedback is scoped to the service and environment, not to a single pipeline, this panel lists the same items for every run against that service and environment, regardless of which pipeline produced the run.
The panel has two sections:
* **Active Feedback:** The compacted summary of every active feedback item for this service and environment, applied starting with the next verification. A feedback item marked **Not Applied** was submitted after the current execution's analysis started, so it is not part of this run's summary.
* **Feedbacks List:** Every individual feedback item, with its text, any attached cluster, who created it, and how long ago.
Each item in the Feedbacks List has two actions, available from the kebab menu on the row:
* **Not relevant** archives the item. It stops influencing future verifications but remains visible in the audit list, and it never changes a past verdict.
* **Make a correction** replaces the item's text. Harness archives the original and creates a new item that keeps the same attached cluster, so you do not lose that context. There is no in-place edit; a correction is always a new item linked back to the one it replaced.
***
## Troubleshooting
The feedback box is disabled
The composer reads: "Feedback isn't available until this run resolves a service and environment." This appears when the executed stage does not resolve a single Continuous Delivery service and environment, for example, a Custom or CI-only stage, or a multi-service rollup view of the Analysis Summary. Open the Analysis drawer from a single CD deployment stage that ran the AI Verify (v2) step.
A feedback item still shows "Not Applied" after I expect it to apply
"Not Applied" means the item was created or last changed after the current run's analysis already started, so this specific run's summary was generated before that change. The item still applies starting with the next verification for the same service and environment.
***
## Next steps
* [Configure the AI Verify (v2) step](/continuous-delivery/use-continuous-delivery/verify-deployments/configure-cv/configure-verify-step-with-ai/ai-verify.md): Review the agent timeline and scatter plot that feedback is based on.
* [Event preference or log feedback](/continuous-delivery/use-continuous-delivery/verify-deployments/cv-results/log-feedback.md): Set a fixed risk level for a log cluster on the AI Verify (v1) step.
{% @harness-feedback/feedback %}
[^1]: Beta features are feature-complete but still being validated. Available on request.