First look: behind the scenes building an AI incident responder.
Lights, Camera, Action
March 8, 2021
We bring you this update under the rays of some glorious spring sunshine 😎🌼.
One of the problems we hear time and time again is that following up on actions after an incident is harder than it should be. They get written down in slack or a post mortem document, maybe exported to an issue tracker, but more often than not, are then left to languish.
Without a clear link back to the incident that caused them, it's hard to track whether or not they get done, let alone see how long that takes and be able to report on it over time.
As a first step towards solving this, from today you can manage your actions via the incident homepage, and export them directly to Jira. We link them back to the incident that they were created from, and keep the state in sync for you.
Next up, we'll be building tools to help you make sure they get done, and surface insights so you can truly understand how quickly that's happening.
We've started with Jira, but Linear and others may follow soon. If you're keen to see that happen, let us know!
What we shipped
New
- Money! You can now pay for incident.io through a shiny new billing page, and we'd very much appreciate it if you did.
- Asking for help is now less noisy. If you mention
/incident help, help will be returned in a modal that only you can see. - Actions are now exported to your postmortem document.
Improvements
- When pinning a message to the timeline, we now include the name of the person that sent the original message, along with the person who pinned it. How lovely.
- We'll only send nudges when the channel appears to be active. This means we won't send repeated nudges in the channel, which makes us look a bit silly.
- 👷🏽♀️ We improved the healthchecks that happen before we deploy a change to production. Yay, resilience!
- 👷 We cleaned up a few unexpected frontend errors we were periodically seeing in production. They were small, and nobody noticed, but we knew, and we weren't having any of it.
Bug fixes
- If you escalated to the person that installed PagerDuty to incident.io, PagerDuty would automatically acknowledge the page, and not wake the person up. Unfortunately for the installer, they will now be woken up. Sorry, installer.
- We got bitten by a classic "DivByZero" bug in our insights stats calculations. Easily missed, easily fixed 🙈
- We weren't colouring your severities unless they matched the defaults, so if you used SEV1/SEV2/SEV3 they were all an uninspiring shade of grey in the dashboard. Now they're in technicolour, too!