AI For All! (Well, all Atlassian Cloud customers...)
What happened now?
Back in April Atlassian announced that they would begin to deploy AI in their products to help empower users… and now they’re fulfilling that promise. What began as a beta deployment of AI for ~10% of their customers is now becoming available to anyone who wants it (well, and uses Atlassian products!). There’s several features that will be made available now to Premium and Enterprise Cloud subscribers (admins can turn them on here), with others coming up soon:
My Take
There’s several different areas that I think would benefit from this type of AI. All of them impact end users on a regular basis, and having any of them would help reduce friction, and frustration, with the people I work with (links to YouTube shorts with more thoughts):
Knowledge discovery - whether its natural language search in Confluence or a Virtual Agent guiding someone to a knowledge base article shortening the gap between “I need something” and “here it is” is always helpful. For me this is especially true in systems, as many times you need some special knowledge (looking at you JQL) to extract it. I’ve also found that even if you take the time to set up Confluence as “perfectly” as possible, things can still slip through the cracks in search.
Productivity Assistance - I’ve used tools like ChatGPT for a while and found them incredibly helpful in augmenting my productivity and creativity. Atlassian intelligence now provides similar support in Jira and Confluence and allows users to draft Pages and Blogs, but also to fill out tickets for user stories, requirements and more.
Guided support - Virtual support agents are another area I’m interested in exploring with Jira. Many times the support process involves soliciting information, then moving down a decision tree to get to an answer. That answer could be “restart your computer”, “do XYZ” or even “I’ll need to escalate this”, but offloading that work onto a machine allows the human agents to focus on other tasks, like managing escalations, improving process and more.
Query Translation - Lets face it, most folks won’t learn JQL, SQL, CQL or any of the “QL”’s. Or, if they do, they’ll learn juuuust enough to get by. Those are incredibly useful tools in finding things, but they’re usually too technical for the majority of uses. Using tools like AI to help bridge that gap, and allow individuals to write natural language statements that get converted into something Jira understands is a great move.
Personally Speaking
Personally I’ve been able to use a bit of this set of tools and I’ve found it very helpful… in particular the ability to search Confluence with natural language. I put a LOT of time into crafting structured pages that are as user-friendly as possible… and even then folks still have trouble finding them. While this feature is still in beta, it gives my team the ability to search MUCH more effectively than using the general confluence search. Or, what honestly happens, having folks open one big FAQ page and using CMD+F to look for keywords. To me, this is basically magic and would reduce the need for me to act as a human router by quite a bit.
One particular use case that caught my eye was allowing the agent to take some action. For example, in this gif it appears the agent is able to grant Jane access to Figma. This is a great use of automation to help reduce workloads (I imagine before a human would have to perform this), however, I can see some hurdles… especially if the system that is being requested is restricted or requires additional approval.
Take on AI
Any of the things these tools can do can be done by a human - but they take time. Time to learn the basics, and then time to create something new every time it’s needed. Personally I’ve learned, and forgotten, SQL or JQL more times than I can count, and while it’s not too hard to remember or relearn, it does take time and effort. I can easily see having an AI Assistant augment those skills be a huge time saver.
Confluence, for example, can now help setup pages or draft articles. This doesn’t take away from my ability to add new policies or other content, but instead reduces my workload by adding formatting, suggesting things to include and prompting responses. This allows me to focus on the more interesting, and important, parts of adding the actual content.
Summarizing pages is another great use case - not everyone has the time to read a 10 page blog post about a great new update. Having an AT tool read through and summarize it into a single paragraph is immensely useful.
I can easily see this type of feature making Confluence or Jira much more approachable to folks as it removes some of the objections related to “how do I use the system”, “I don’t know where to start” and the like. This in itself is a win as it reduces the distance from “I don’t know what to do” to “I’ve created something”.
Next Steps?
The next thing I’d love to see is if these tools can also reach out to the right team or person to gain approval to add Jane…. although then I wonder how the agent would handle someone ignoring it’s request (maybe shutting off access until there’s a response?….).
Or, for example, having the system realize I’m writing an article on HR policies, and either suggest other folks who should chime in (either by knowing what they’ve written, their status in the organization or other data), or proactively reach out to let those individuals know my article is there. A similar feature could apply to tickets - pulling in resources who can help unblock or progress an issue without me having to ask. A feature like this would help shorten the time it takes to get something approved, and make sure the folks who need to know about the change, or chime in, are looped in.
Overall I feel that tools like these will augment and improve the humans that use them, and I’m excited to see how Atlassian is combining their current tools with these new ones. I’m looking forward to seeing what everyone does with them!