Rootly's Slack AI agent turns incident response into a permissioned workflow
JJ Tang and Quentin Rousseau are extending Rootly from incident coordination into action-taking AI inside the channels responders already use.
By Ryan Merket ยท Published
Why it matters
Rootly is testing the practical boundary for enterprise AI agents: not whether they can chat, but whether teams will let them act inside live operational workflows.
JJ Tang and Quentin Rousseau's Rootly has pushed its incident-management AI deeper into Slack, launching an @Rootly AI Agent in Slack that can read incident context, answer responders' questions and take operational actions from inside an active incident channel.
The launch, dated June 4, 2026 in Rootly's changelog, is not a new-company moment. It is the latest move in a longer founder bet: Tang and Rousseau have spent the past five years trying to turn the messy human layer of outage response into structured software. An Aligned News post on X framed the Rootly AI conversation as a signal about AI deployment and adoption. The sharper point is where Rootly is choosing to deploy the agent: not in a new console, but in Slack, where incident commanders, engineers, customer-facing teams and executives already coordinate during outages.
Rootly says the Slack agent can use Rootly incident data, Slack channel discussion and bridge-call transcripts in real time. Once mentioned in a channel, @Rootly can return a catch-up brief, identify on-call owners, detect action items that were discussed but not logged, change incident severity or status, assign roles, create action items, page responders through Rootly On-Call, draft or publish a status-page update and surface similar past incidents.
The permission model is the product detail that matters. Rootly says every action runs as the requesting user, under that user's existing Rootly permissions, and every change is attributed in the audit trail. That moves the feature out of the category of a generic incident chatbot and into something closer to an operational agent with role-based access, memory and accountability.
The Instacart scar tissue is still the product thesis
Rootly's founder story has always been tied to Instacart's scaling pains. In Rootly's 2021 seed announcement, Rousseau wrote that he had been an early engineer and one of Instacart's first SRE hires in 2015, responsible for cloud infrastructure, CI/CD and site reliability as the business moved from hundreds of orders to millions. The same post says Tang was a senior product manager at Instacart who built enterprise offerings responsible for more than 20% of company GMV and helped lead COVID-19 response efforts.
That matters because Rootly's AI agent is not aimed at the glamorous part of incident response. It is aimed at the coordination debt: creating channels, assigning roles, chasing down owners, updating customers, logging action items and turning the incident into a retrospective after the fire is out. Rootly's seed post described the old workflow as jumping between Slack, PagerDuty, Datadog, Google Docs, Asana, Jira and internal runbooks. The Slack agent is a direct answer to that original pain, except the automation has moved from playbooks and bots into AI that can act in the same channel where the incident is unfolding.
Rousseau has a second founder thread running through the story. Y Combinator lists him as a two-time founder at PopSQL and Rootly. The Org lists prior roles at Canal+, Onefeat, Instacart, PopSQL and Mylk Guys, and describes him as Rootly's CTO, CISO and co-founder. Rootly says he was an early SRE at Instacart, which helps explain why Rootly's AI posture is less about replacing engineers and more about reducing the number of handoffs that break during production incidents.
Why Rootly wants the Slack surface
Rootly is not trying to win the incident market by being another alerting siren. The wedge is the layer after the alert: incident declaration, responder coordination, communication, status pages, retrospectives and on-call workflow. Y Combinator currently describes Rootly as an AI-native on-call and incident management platform used by companies including Dropbox, Lattice, Webflow, Faire, Figma, LinkedIn, NVIDIA and others. TechCrunch reported in 2023 that Rootly customers were managing more than 60,000 incidents a year through the platform, a figure Tang claimed at the time.
That same TechCrunch report said Rootly raised a $12 million Series A led by Renegade Partners, with Google Gradient Ventures and XYZ Ventures participating, bringing total disclosed funding to $15.2 million. Rootly had earlier announced a $3.2 million seed round led by Ross Fubini at XYZ Venture Capital, with 8VC and Y Combinator participating. No valuation has been disclosed in the materials reviewed.
The Slack agent shows how Rootly is spending that product surface. PagerDuty and Opsgenie are associated with alerting and escalation. FireHydrant, incident.io and Blameless compete around incident process and response. Rootly's current push is to collapse incident work into the collaboration layer itself. If the incident channel is where decisions are made, then the AI that can summarize, recommend and modify incident records needs to live there too.
Rootly has been laying that path for more than a year. In February 2025, Tang wrote on Rootly's blog about building a new-generation Slack AI app using Slack's newer AI app framework. In April 2025, Rootly launched Rootly AI Labs, a community and open-source program around AI reliability tooling, including IncidentDiagram, EventOrOutage, model benchmarks and a Rootly MCP Server. The June 2026 Slack agent is the commercial expression of the same direction: AI that is embedded in reliability workflows rather than bolted onto them.
The buying question is trust, not novelty
The market does not lack AI copilots. Engineering teams are already being pitched AI for observability, alert triage, incident summaries, root-cause analysis, customer support and code fixes. The buying test for Rootly is narrower: will a reliability team let an AI agent act during a live incident?
Rootly's answer is to make the agent permissioned and auditable. That is the right constraint for this category. During an outage, a wrong summary is annoying, but a wrong severity change, a mistaken status-page update or an unnecessary page can create real operational cost. Rootly is not claiming the agent has unlimited autonomy. It is saying the agent can do what the user is already authorized to do in Rootly, then leave a record.
Pricing also shows where Rootly is drawing the line between standard incident software and AI-heavy reliability tooling. Rootly's current pricing page lists Essentials at $20 per user per month for Incident Response and $20 per user per month for On-Call. AI SRE is listed as a separate contact-sales product, with Rootly saying it can identify probable root cause, correlate alerts with changes, generate impact analysis, draft remediation steps and integrate with tools such as Datadog, GitHub and Jira.
That packaging is a tell. Rootly is keeping the core incident workflow accessible while reserving the more ambitious AI SRE product for sales-led evaluation. The Slack agent sits between those two worlds. It is easy to understand because it starts with a mention in Slack. It is harder to sell at scale unless Rootly can prove the agent saves time without creating new risk.
What Rootly still has to prove
The missing metric is adoption. Rootly has not disclosed how many customers are actively using @Rootly in Slack, how often the agent takes actions rather than answers questions, or whether incident teams are using it during severe production outages rather than low-risk retrospectives and internal incidents.
Those numbers matter because the first wave of AI incident products will be judged by trust under pressure. Demos make incident agents look obvious. Real incidents are noisy, political and time-sensitive. The responder who is wrong in Slack is visible to everyone. The AI agent that is wrong in Slack will be visible too.
Rootly's advantage is that Tang and Rousseau are not entering reliability from the outside. They built Rootly around the unglamorous operational mess they encountered at Instacart, then funded it with investors who were buying that founder-market fit before AI agents became the default enterprise software pitch. The June Slack launch is the clearest version of that thesis so far: Rootly is betting that the winning AI in incident response will not be the most general agent. It will be the one with the right context, the right permissions and the right place in the room when production breaks.