Since moving from Opsgenie to Rootly, @pandadoc's stakeholders can ask the Rootly Agent for incident details instead of pulling responders into another conversation. That means people get the context they need while engineers stay focused on resolving the incident. The Rootly AI Scribe captures what’s happening along the way, making communication easier during and after the response. The move is also helping PandaDoc retire custom tooling its SRE team spent years maintaining. Slack agents, slash commands, and on-call payment exporters are giving way to Rootly’s native features and workflows. PandaDoc finished its core setup in under a month. The benefits reach beyond the on-call rotation: easier access to information for stakeholders and less maintenance work for engineers.

How much should you trust an AI agent during a production incident? Join our SF panel with: • @grinich (CEO, @WorkOS) • Vishakha Sadhwani (Sr. Solutions Architect, @NVIDIA) • @niallohiggins (Engineering Lead, @Replit) • @SylvainKalache (Head of AI Labs @rootlyhq) Agents on-call, human judgment & ownership.

When you build an agentic platform, a lot of the engineering work lives in the moments between actions: – Someone answers a question while the agent is preparing a reply. – An approval arrives after the underlying message has changed. – Two triggers fire close together. Each action can make sense on its own. The hard part is reasoning about how they interact. We run into this at Rootly building agents for incident response. Timing, changing context, and human decisions are all part of the system’s behavior. And they’re difficult to review in a static document. So we made our engineering requirements document interactive. Reviewers can replay a timeline, edit a source message, let an approval expire, or click Accept twice. They can explore the transitions and trace the explanation back to the code. That introduces its own engineering challenge: the replay is a separate implementation, so it can drift. Keeping it aligned with production is part of maintaining the document. For teams building agents, making behavior inspectable is becoming part of the engineering work.

Writing down “A config change triggered the incident” is the easy part. But what else had to be true? Which safeguards failed to catch it? What information was missing? What made the responder’s next decision seem reasonable? Asking for a single root cause too early can narrow the investigation before the team understands what happened and steer the conversation toward blame. As Ben McBratney, our Head of Incident Strategy, explains, AI can help reconstruct the incident. The team still needs to do the work of learning from it. That’s why the Rootly Retrospective Template asks for contributing factors, plural. AI helps prepare the draft; responders question it, add context, and put together what they’ve learned.

.@AurorHQ is on a mission to reduce violent retail crime by 50% in five years. When their platform goes down, retail workers can get hurt. That's why Nigel Wright, Director of Platform Engineering, called Rootly when he started building Auror's next stage of reliability. Today, Auror uses Rootly to process every alert source, page the right responder, manage incidents quickly, and write the retrospective. With Rootly, Auror's leadership has seen incidents close much faster across the board.

Dear Afraid to Book the Flight, If time off feels like a favor, you're just making it work. A rotation that works doesn't run on goodwill. From the outside the two look the same, because incidents get handled either way. The difference shows up when someone tries to leave. Try this test. Ask your team, "If two of us were out next week, who covers on-call?" If you have a plan, you're fine. If the answer is "we'd figure it out," you're making it work. Four people is thin. When one person is on vacation and another gets sick, two people are covering everything. My rule of thumb is that a rotation needs at least five people before it has any slack. Google's SRE book says eight. And no, the usual fixes don't solve this. Shift swaps are favors. Stipends don't add a name to the schedule. A quiet pager still means bringing your laptop to your kid's soccer game. If you can't hire, widen the rotation by merging with another team, adding a secondary, or taking low-impact services off the overnight pager. Tired responders miss things. When they do, the fault lies with a system that assumed nobody would ever get tired, not with the person. You should be able to book the flight. —Ben On-Call