How do we use AI?

We sprinkle AI across our product to help you respond quicker and smarter. Here's how it works. Engineering Team avatar
Written by Engineering Team
Updated over a week ago

Why AI?

We are always looking for ways to improve your organisations incident response. Whether that is improving the lives of responders as they react in the heat of the moment or distilling previous incidents into powerful learnings. We want to make sure things are quick, easy, and understandable.

For us, artificial intelligence (AI) is potentially the perfect partner to help take away some of the overhead. It can help when you're trying to digest huge amounts of information, or communicate clearly and quickly to stakeholders. It can even suggest when an incident is similar to another and potentially pull in the relevant people for you.

Although it is early days, we strongly believe it has a pivotal role in helping you remain in the flow and getting on with the nitty-gritty while it focuses on the bigger picture. Think of it like a helpful colleague chipping in or taking tasks off your hands when it can.

Suggested summaries

In the heat of the moment, responders don't always have time to digest lots of information or communicate externally what's going on. Often it's hard to switch context and spend some precious minutes repeatedly updating summaries or statuses.

This is where Suggested Summaries comes in.

We take a collection of information drawn from the incident and use AI to help pull out the important information for you such as the problem, the impact, anything that may have caused it, and the next steps to that will be taken to help resolve it.

This lets you quickly confirm if the summary is accurate and edit it if you want more detail before rapidly getting on with resolving the incident. Overall, this means your incident summaries are more up to date and in a standard format - allowing newcomers to quickly understand where the incident is and help out.

Related Incidents

Sometimes when you first look at an incident, you might realise that you've seen it before, or something very similar. But where? How long ago? Who was ?

With related incidents, you can consider us as a helpful memory tool.

We use previous incidents to help elevate any context or shared learnings and potentially reduce time it might take to understand the problem again. To do this we build a unique signature for each incident using information like the summary, updates, or custom fields and combine it all together in a way that lets us use AI to spot when 2 incidents are similar to one another.

When we spot a related incident we post in channel allowing you to accept or reject. We also provide the option to pull in the incident lead so that they can help provide any additional help should it be required.
Once an incident is linked you can quickly browse the previous incident, see any actions or pull requests that were made, see what conclusions people came to and even look at the post-mortem.

How do our AI solutions work?

Our AI features are powered by OpenAI.

We typically send data to their API when required by a feature, namely on-demand. This data is not stored by OpenAI and is not used by OpenAI for any reason other than to provide these services. This means it is explicitly not used for training purposes.

๐Ÿ”’ We never send data from private incidents.

OpenAI is listed as one of our data sub-processors, so all customers automatically consent during sign up. However, if you'd like for us to stop sending data to OpenAI, and therefore disable all AI features, please reach out to us.

What data is required for AI features?

Note: Detail about data storage is discussed more in our Vanta report here.

For most AI features, the more data and context you can provide to the model the better the result, typically. For both suggested summaries and related incidents, we use summaries, updates, and custom fields.

So, we suggest allowing the storage of all Slack messages so that our AI suggestions for summaries will consider the ongoing responder conversation, which is usually the richest source of information. This drastically improves the performance of certain AI features, such as incident summary suggestions.

By default, we will store all incident channel messages in order to provide you the best results for Suggested Summaries.

If you have disabled storage of incident channel messages, you may find our suggestions lack depth and may not be as useful. We highly recommend you enable the setting for Store incident channel messages to All .

Did this answer your question?