Building a new-generation Slack AI App at Rootly

Slack is launching a new type of app: Slack AI Apps. These new apps let select partners integrate their AI agents with Slack natively, reducing context-switching for users.

That means you can now get contract insights via Box, draft a project report via Cohere, compare internal product roadmaps via Perplexity, and understand your incident trends via Rootly without leaving Slack—all through native AI chat assistants.

In this article, we’ll introduce the new Rootly Slack AI App and outline what it took to build it using the new Slack APIs.

The New Rootly Slack AI App

The Rootly AI App pulls critical incident data from across your systems—observability, tickets, and communications—to quickly summarize conversations, automate root cause analysis, and draft stakeholder updates, so you can focus on fixing the problem.

Powered by Rootly's AI-native on-call and incident management platform, it leverages data from your observability tools and context from your incidents.

Gain Instant Context

Ask the Rootly AI App to craft status updates and summarize key conversations. It pulls together insights from all connected sources to provide a clear picture of the incident, so you're up to speed instantly.

Automate Investigations

The Rootly AI App pulls data from your observability tools, ticket systems, communication bridges, and more to automate investigations and present probable root causes. It leverages data to answer questions like, "When did orders drop in the last hour?" while providing actionable next steps—without having to jump between dashboards to figure it out.

Prepare Detailed Comms on the Fly

The Rootly AI App prepares detailed communications to update and align stakeholders at key points during an incident, freeing you to focus on resolving the issue quickly.

Exclusive Slack AI Partner

Slack chose only 25 out of the 2,600 vendors available in the Slack Marketplace to develop the first AI Apps. Slack selected Rootly as a premier AI partner because of our popularity and innovation track record within their ecosystem.

Building a Slack AI App

Slack AI Apps add an AI assistant, or agent, into your Slack app so you can interact with a third-party AI alongside your regular channels and DMs. You can interact with the AI agent through a conversational interface. For example, you can ask Rootly AI in Slack how the MTTR in your team has evolved over the past few months, and it will give you a breakdown and show you a graph.

To make this experience possible, we had to:

• Aggregate data from Slack, observability tools, meeting notes, and other sources.
• Process the data using machine learning and implement an LLM that can interact with customers.
• Integrate our LLM with the Slack AI App API.

Aggregating Customer-Specific Data

Each Rootly customer uses our platform to manage their incident lifecycle, from alert to retrospective. This gives us a valuable repository of data that can help customers identify patterns and refine their reliability strategy.

That data, however, is scattered across multiple sources: incident-specific Slack channels, various observability tools, incident meeting notes and recordings, and retrospective documents, to name a few.

We use Slack’s new Data API (available only to select partners for now) to get bulk access to Slack data. Unlike the Web API, which is designed for real-time results, the Data API is designed for extracting large datasets using efficient queries.

We never aggregate data from different customers. All data is managed with strict enterprise-grade privacy.

Implementing ML and LLMs

For Slack conversations, meeting transcripts, and post-incident reviews, Rootly uses Natural Language Processing (NLP) techniques to find patterns and insights. To analyze logs and other data types, Rootly is exploring alternatives like using distilled DeepSeek. We’re constantly pushing the limits of what can be done in the incident management and AI space.

Our customer-facing LLM is an agent based on enterprise OpenAI. We’ve honed it over the past year, and it’s been making our customers more productive.

Integrating the Agent with Slack

Slack AI Apps are initially created as regular Slack apps. What makes them different is the App Assistant Flow, the framework that Slack offers to set up interactions with your AI agent.The App Assistant Flow provides entry points in the Slack UI and lets you configure a side-view panel with rich, in-context interactions. It also includes native loading states and a standardized way of displaying suggested prompts.

What’s next

The Rootly AI Slack App is available in the Slack Marketplace. We are heavily investing in making our agent smarter so your SRE team can be more productive and resolve incidents faster. Our mission is to make the world more reliable while giving SREs the sleep they deserve.