AI Agent in Communication Channels

Rootly AI answers incident questions and acts from Slack or Microsoft Teams, with Rootly and connector-specific permission checks.

​

Overview

Ask Rootly AI to:

• “This is now a SEV1. Update the incident severity.”
• “Draft comms to our high-value customers, letting them know the impact of the incident on their product experience.”
• “Page the security team, and convert this to a private incident.”
• “Catch me up. What’s still open, and who owns it?”

Rootly AI brings the parts of Rootly you reach for during an incident into the communication channel where your team already operates. It’s the same agent with the same capabilities in every supported chat tool. Mention @Rootly and ask.

Answers like this one draw on your own tools through Connectors: plug in your observability, code, and docs sources and Rootly AI investigates with them.

Private Agent is an Early Preview for approved customers. When both Private Agent and AI SRE are enabled for the selected team, it lets Rootly AI reach services inside your network. Its capabilities follow your Rootly role and the agent’s local policy, with a separate access boundary from SaaS connectors.

---

​

Supported Communication Channels

Pick your chat tool to turn on Rootly AI and see where you can use it.

Rootly AI in Slack

One toggle and a Slack reconnect. Use it in incident channels, the Slack assistant pane, and DMs.

Rootly AI in Microsoft Teams

One toggle on top of your Teams integration. Use it in incident channels, team channels, and chats.

---

​

How Rootly AI Works

Rootly AI reads from your Rootly data and the channel conversation, including bridge call transcripts when available, to answer questions and take actions on your behalf. Mention @Rootly and ask. It uses the incident’s live context to decide how to respond, and replies in-thread so the channel sees any action taken.

Rootly AI also answers outside incident channels, such as in private conversations with the Rootly app. Changes to an existing incident happen only in that incident’s channel. What else you can do from each surface, such as paging or creating schedule overrides, depends on your chat tool, and a connector can expose separately governed external actions. See Using Rootly AI for the full guide.

Model details, including failover behavior, are documented in Data Privacy for Rootly AI.

---

​

Key Capabilities at a Glance

Capability	What it does	Example ask
Catch me up	Structured brief of the incident’s state, key events, and open work	”@Rootly what’s going on right now?”
Who’s on call	Current and historical on-call lookups	”@Rootly who’s on call for the Security team?”
Page responders	Pages a team, user, or escalation policy via Rootly On-Call	”@Rootly page the Search Experience team.”
Create overrides	Creates a schedule override so someone else covers a shift	”@Rootly have Sam cover my on-call shift on Friday.”
Update an incident	Changes severity, status, roles, and fields, confirming what it did	”@Rootly bump the severity to SEV2.”
Find related incidents	Ranked matches from your own incident history	”@Rootly are there any related incidents involving this service?”
Create a retrospective	Creates the incident’s retrospective from the template you confirm	”@Rootly create the retrospective for this incident.”
Answer product questions	Explains how Rootly features work, based on Rootly’s product documentation	”@Rootly how do I pin something to the incident timeline?”
Draft customer updates	Chat-ready or status-page-ready update copy from current state	”@Rootly write a draft update for leadership and our partners.”

This table is a sample. The full prompt catalog lives in What to Use Rootly AI For.

---

​

Who Can Use Rootly AI

Rootly AI requires an Incident Response role. Without one, Rootly AI asks you to contact your admin, and requests to page or check who is on call are refused.

---

​

How Permissions Work

Rootly AI uses each user’s existing Rootly permissions. If a user can’t see a private incident in the web app, Rootly AI won’t surface it for them in chat. If a role doesn’t allow paging, Rootly AI declines to page when that user asks.

Connector data and actions can add a separate access boundary. A connector guide identifies which users can use its tools and the provider account that scopes the available data and operations. For example, Devin repository reads use the connected service user’s repository access, including for Rootly AI users who cannot manage the connector.

Permissions are managed in Settings → Roles & Permissions. There is no separate AI permission system.

---

​

Frequently Asked Questions

What about private incidents and private channels?

Rootly AI has to be added to a channel before it sees anything there or takes action. User permissions still apply on top.

Can Rootly AI page using vendors outside of Rootly On-Call?

Rootly AI’s native paging action uses Rootly On-Call only and does not fall back to third-party paging integrations configured elsewhere in Rootly. A connector with an explicitly supported paging tool can define a separate permission boundary; see that connector’s guide.

Does Rootly AI work differently in Slack and Microsoft Teams?

It’s the same agent with the same tools, permissions, and confirmation behavior. What differs is setup and where you open it, which each platform’s page covers.

---

​

Related Pages

Using Rootly AI

How clarification cards work and how actions run.

What to Use Rootly AI For

The full catalog of prompts, grouped by use case.

Troubleshooting

Common symptoms and how to resolve them.

Data Privacy for Rootly AI

What Rootly AI sees, retention, and model training controls.