Rising Stars are recognized for providing high-quality answers to other users. Rising Stars receive a certificate of achievement and are on the path to becoming Community Leaders.
@Sherif Mansour I have been using the AI powered search in Confluence since beta and can say it's been a game changer for us. Surfaces answers instead pf pages (where you then need to search for answers).
More recently the introduction of Page Summaries in Confluence has improved quality of life for our internal users - especially when the top of the page says its a 17 minute read!
Rising Stars are recognized for providing high-quality answers to other users. Rising Stars receive a certificate of achievement and are on the path to becoming Community Leaders.
This all looks good, we're simplifying search queries in our own products and I've been trying out the Jira AI search which should make life much easier for our non-technical teams.
Hi @Sherif Mansour , thanks for the overview. While I appreciate that AI is useful, I am a little worried about the environmental impact, which isn't really yet understood for how much energy is consumed by AI. Is Atlassian doing anything or thinking about any initiatives to counteract this consumption?
I think the general assumption that energy consumption is very correlated to cost trends is a good one to have as we approach this problem. Zooming out across the industry, it is also fair to say that if we believe AI will change how teams work every day, then in-general there will be more energy required to power these models.
Bringing it closer to Atlassian, we've long been an advocate for climate action. We set science-based targets to limit warming to 1.5Β°C and achieve net-zero emissions by no later than 2040. Reducing emissions remains our top priority, which includes emissions from our cloud and AI model suppliers.
There are broadly two parts of the equation when it comes to AI. (1) Energy consumption when it comes to training these models - often from third-parties, and (2) energy consumption when it relates to our feature execution.
On the former, we will continue to work with model providers to work out how this gets better over time, but it's clear that as the technology is getting better there is lots of solutions here. E.g. Smaller more efficient models or something like Apple who just announced models running on-device for most transactions and for complex queries they could go to a cloud provider etc. We suspect these patterns to start to follow in our space as well.
On the latter, we continue to make sure our AI features are more efficient over time and reduce their consumption. As a practical example, the team that worked on Natural Language to JQL in Jira moved from a general-purpose model to a fit-for-purpose model that uses about 20x tokens every time it's run since it's launched. More and more teams at Atlassian are looking to optimise usage like that to ensure the usage stays as low as possible.
Worth closing in acknowledging that climate change is impossible alone - we invite all companies here, your own, AI and Cloud services to join us and set science-based targets to reduce their emissions.
Exciting updates, Sherif! The new AI-powered features in Atlassian Analytics and Confluence sound like they'll be a huge boost to efficiency and user experience. Can't wait to try the chart insights and content refresh panels.
Great insights on energy consumption and AI! Itβs impressive to see Atlassianβs commitment to sustainability, especially with initiatives like using more efficient models for AI tasks.
Rising Stars are recognized for providing high-quality answers to other users. Rising Stars receive a certificate of achievement and are on the path to becoming Community Leaders.
Getting chart insights literally made my day, this is so amazing and I have been waiting for this for ages. Last night I was analyzing some excel visuals for 2h, and thinking about how user unfriendly it is.
Thank you for making such amazing products and for constantly pushing the boundaries of whatβs possible. I can't wait to dive in and explore these new features!
Where can we find the technical documentation for how AI is used within Jira? What data is the model trained on? When we turn it on, is the model using our data to train itself better? Is our data ever used to train models which would ever be used outside of our instance?
I'm curious where this documentation, the type of documentation that gets an IT security team onboard, lives.
One use case that came up earlier today would be the ability to summaries linked/child stories, bug etc up to an Epic or Initiatives. Management doesn't typically want to be down in the details of the stories etc so allowing a Product Owner to summarise sub issues to a parent would save huge amounts of time.
The AI free Atlassian courses are a wonderfull idea, thanks !
I have a question : Do you have in your plan an AI feature for JSM that will help agents to resolve issue based on Confluence content ? This resolution suggestion could appear in the agent screen.
In my compagny we have a confluence space with simple articles for customers that is linked to the portal and a space with more technical articles for agents that is linked to the agents screen in order to suggest pages related to the issue. I think it could be very usefull to be able to do the same with AI
@Cyril Donnellan RE: SOC 2 etc for AI, soon! My internal Atlas ticket tells me lots of folks working on this trying to get it out soon to customers! I'd expect coming few months π€πΌ
@mari that would be super cool! I know the team is investing a lot more on using the JSM KB to help with things like this. Passing on feedback @philoye
55 comments