> ## Documentation Index > Fetch the complete documentation index at: https://docs.rootly.com/llms.txt > Use this file to discover all available pages before exploring further. # Related Incidents > Rootly AI ranks past incidents that match the one you are running, grounded in your team's own history, and surfaces them in the Slack incident channel. ## Overview Rootly AI matches your incident against everything your team already solved. Related Incidents in Slack: a proactive summary card in the incident channel, and the full list of matches with per-incident feedback buttons when a responder asks Matches surface right in the incident channel and on demand, whenever a responder asks. ### Acting on a Match Every match card carries the same set of actions: | Button | What it does | | ------------------- | ---------------------------------------------------------------------------------------------------------------------- | | **Mark as related** | Trains the model that this was a good match, so future recommendations within your team's incident set get better. | | **Not related** | Trains the model that this was not a good match, so future recommendations within your team's incident set get better. | | **Channel** | Opens the Slack incident channel for the related incident. | | **View in Web** | Opens the related incident's details page in Rootly on the web. | Feedback only shapes recommendations for your own team — it never trains models for other customers (see [Privacy](#privacy)). *** ## Grounded Answers Ask `what fixed it?` and get an answer you can act on mid-incident: * **On the record.** The documented resolution is relayed. * **Nothing recorded.** Rootly AI says so and offers the retrospective or timeline. Anything beyond your records is labeled inference ("the pattern suggests"), not fact. *** ## Selective by Design * Reads each incident like a responder would: title, summary, severity, services, and more. * Keeps only genuine, concrete relationships. Similar wording alone doesn't qualify. Matching stays live: a new near-duplicate can hit rank #1 within minutes. ### What Qualifies as a Match | Rule | Detail | | ----------------------- | ---------------------------------------------------- | | **Your org only** | Drawn exclusively from your team's incident history. | | **Confirmed incidents** | Excludes cancelled incidents. | | **Recent history** | Looks back 180 days by default. | | **Public incidents** | Only what the whole channel can see. | On first enable, your history processes gradually in the background. New incidents become matchable within minutes. ### Get Better Matches * **Write specific titles and summaries.** Matching reads what you write. * **Fill in the fields.** Every attribute is signal. * **Record how incidents end.** Resolution notes and [retrospectives](/ai/ai-in-retrospectives/overview) answer "what fixed it?" * **Rate the responses.** **Mark as related** / **Not related** tunes suggestions to the way your team works. *** ## Privacy **Enforced in the product, not by prompt instructions:** | Guarantee | Detail | | ---------------------------------- | -------------------------------------------------------------------- | | **Filtered before the AI writes** | The AI never references, or even alludes to, inaccessible incidents. | | **Private incidents stay private** | Excluded from channels, for any asker or setting. | | **Your history only** | Matches come only from your own incidents. | | **No cross-customer training** | Your data never trains models for other customers. | | **Permissions govern the subject** | Refuses questions about incidents you can't read. | For retention and model detail, see [Data Privacy for Rootly AI](/ai/data-privacy-for-rootly-ai). *** ## Setup Related Incidents is part of Rootly AI, behind the [organization-wide opt-in](/ai/ai-settings), and delivered by [Rootly Agent in Slack](/ai/rootly-in-slack/overview). 1. Open **AI & Agents**. 2. Under **Features**, toggle **Related incidents** on. *** ## Frequently Asked Questions Nothing cleared the relevance bar. If just enabled, your history may still be processing. Confirm **Related incidents** and **Rootly Agent in Slack** are both on. Most misses are eligibility: check [What Qualifies as a Match](#what-qualifies-as-a-match). Weak metadata can also keep a real match below the bar. *** ## Related Pages Everything the agent can do in your incident channels. What Rootly AI is and everything it can do. Data handling and model training controls for Rootly AI.