Rootly automatically detects related past incidents and posts them to Slack

Rootly, an incident and on-call management platform that competes with PagerDuty, Opsgenie and Jira Service Management, has added a new capability to its AI agent: automatic detection of related past incidents, posted directly in the Slack channel where the response to an active incident is being coordinated. Nobody has to ask for it: as soon as an incident is opened, the agent searches the organization's incident history in the background and, if it finds a match it deems genuinely relevant, shares it with the response team on its own initiative.

The mechanism is designed to be unobtrusive and useful at the same time. The agent posts a maximum of three matches per message, and each one comes with the reason for the match (for example, that it affects the same service or that it is a similar type of failure), the severity of that earlier incident, how long it took to resolve and a direct link to the original incident so the full response can be consulted without leaving Slack. The details are grouped into a thread so the main incident channel does not get flooded with information, and the card can be dismissed if it is not useful.

The rationale Rootly gives for the feature points to a classic problem in incident management: almost always someone on the team has already faced a variant of the problem that is happening, but that knowledge tends to be scattered among whoever took part in that response, or buried in a postmortem nobody thinks to look for at three in the morning. On top of that, the person on call that night is rarely the one who resolved the incident the previous time. By inserting the match directly into the channel, the team spends less time trying to work out whether something similar has happened before, freeing up time to focus on mitigating the current problem, the company explains.

Rootly includes in the announcement a testimonial from Mark Shaw, of Faire's engineering team, who says that proactive detection of related incidents has spared him from digging through old postmortems to find out when a similar incident last occurred, saving time and keeping the team's attention on the problem at hand. The feature is enabled from the account's "AI & Agents" settings, and the company points to its help documentation for more details on how it works.

The announcement comes alongside a broader batch of improvements and fixes published in the same changelog. In the on-call area, a schedule can now be assigned to a Slack user group spanning several Slack workspaces at once, and new users added through single sign-on inherit the organization's time zone instead of defaulting to Pacific. In incident response, it is now possible to sync a ServiceNow CMDB with Rootly's catalogs, incident links shared in Slack show a preview, and creating a task from a Slack form now confirms that the action completed successfully.

On the API and developer tooling side, outgoing alert webhooks now include more detail on each alert, an escalation path read through the API can be sent straight back with a PUT request, and on-call listings filtered by user have become noticeably faster, fixing timeouts that previously ran out the clock. The Slack channel selector for escalation policies now also correctly lists the channels of people whose Slack account resides in a secondary workspace. In Terraform, the rootly_services data source now walks through every page of results instead of silently stopping at the first 50 records, a bug that could go unnoticed in large service catalogs. Rootly's MCP server also gets several adjustments: shift date ranges are now split into chunks the API accepts, out-of-range requests are caught before being sent, and additional aliases have been added for parameter names, while keeping compatibility with the previous ones.

This type of feature fits into a broader trend in the systems reliability sector: the arrival of AI agents that not only automate repetitive tasks but proactively provide historical context during incident response, without the team having to ask for it. At a time when outages at AI services and cloud infrastructure have taken center stage, tools that cut diagnosis time by drawing on the accumulated knowledge of past incidents fit into the move toward what Rootly itself calls "AI SRE": site reliability engineering assisted by agents that act autonomously within the team's existing workflows, such as Slack, rather than requiring an additional tool or search step.