LoginSign Up
← Back to all sparks
R

Rootly

INFRA · APIS
Velocity6.3

Rootly's AI agent is going proactive — surfacing incident context before responders think to ask for it.

incident-managementai-agentson-callobservabilitysredevtools
◆Current state
Rootly has shipped a concentrated run of AI features that push the incident workflow from reactive tooling to ambient assistance. The agent now automatically posts relevant past incidents into the incident channel, transcribes calls with speaker labels in real time, operates on mobile without a laptop, and can sync service catalog data from GitHub and Backstage via CLI. Custom fields on action items unify the data model between incidents and the tasks generated from them.
◆Where it's heading
The architectural direction is clear: Rootly is building an AI layer that works in the background throughout an incident — gathering evidence, matching patterns, transcribing conversation, surfacing historical context — so responders can focus on diagnosis rather than context assembly. The late-July observability stack connector (read-only access to code, infrastructure, feature flags, docs) is the data foundation; each successive release extends what the agent can observe and act on without prompting.
◆Prediction
The logical next step is the agent moving from observation to recommendation — suggesting remediation steps or escalation paths based on the pattern-matched past incidents it's already surfacing. Automated post-incident summaries generated from Scribe transcripts and action item completion data are also within reach.

◆Recent moves

  1. 7d ago

    Never lose the call transcript.

    Scribe now joins incident bridges, transcribes in real time with speaker labels, and stores the recording as part of the incident record. Closes the documentation gap that typically opens when responders are too occupied with the incident to take notes during the call.

  2. 13d ago

    Related Incidents detected automatically.

    ⚡ SPARK

    Rootly's agent now monitors new incidents and proactively posts the most relevant past incidents into the incident channel — before anyone searches. This is the most visible shift in the product's AI posture: from on-demand retrieval to unsolicited context injection.

  3. 20d ago

    The Rootly Agent is mobile.

    The Rootly Agent is now accessible from the mobile app, letting on-call engineers gather incident context and take action during an off-hours page without opening a laptop. Extends the AI layer to the most common first-response scenario.

  4. 28d ago

    Custom fields on Action items.

    Action items can now carry the same custom fields as incidents, enabling consistent filtering, reporting, and automation across both objects. Bridges a metadata gap that previously broke cross-object analysis and forced workarounds for teams trying to track follow-through on post-incident remediation.

  5. 1mo ago

    Meeting Scribe speaks your language.

    Meeting Scribe now auto-detects spoken language and transcribes accordingly, removing manual configuration for non-English incident teams. Extends the feature's reach to multilingual organizations without changing the underlying architecture.

  6. 1mo ago

    Sync your catalog, and mute pages during Alert storms.

    A new CLI syncs service catalog data from GitHub, Backstage, and similar sources, keeping Rootly's context layer current without manual updates. Mobile alert muting lets on-call engineers silence non-critical pages during alert storms. A new Upwind integration rounds out the release.