Can we build a podcast-style workflow using Confluence Audio?

[Gabi Mucha]

My team at Atlassian has been leaning more into long-form content lately - research reports, white papers, event session recaps, and customer stories - and the question naturally follows: how do we make this content easier to consume, more conversational, and more modular?

For context, I’ve been testing out Notebook LM which is a Google experimental tool that acts like an AI research assistant - you upload documents like white papers, reports, or transcripts, and it can summarize, answer questions, or even generate new content based on what you give it. But, I’ve been testing it for a more specific use case: podcasts. You basically upload a long-form asset and it can turn that content into a podcast-style conversation — complete with natural-sounding voices and a tone that’s surprisingly easy to follow. It’s pretty cool and it’s a great way to make dense content more accessible and engaging!

So I set out to explore: could we recreate something similar using internal tools we already have access to? Specifically, I spent some time experimenting with Confluence Audio, testing whether we could stitch together a lightweight workflow to turn long-form content into something more digestible like podcast-style snippets (Notebook LM-style). So, here is a quick write-up at what I tried, what worked and what didn’t.


Confluence Audio

The Confluence team recently rolled out this new AI-powered feature called “Listen to briefing”, which lets you listen to quick, auto-generated summaries of any page. It’s designed to work alongside the written content - so you can listen while also browsing the page, checking out charts/graphics, or skimming through other details.

It’s especially useful for auditory learners or anyone who prefers to absorb information by listening rather than reading. More broadly, it supports multimodal workflows - giving people multiple ways to engage with content in Confluence, whether they want to read, listen, or do both at once.

I gave it a try using one of longer blogs and was honestly impressed. Even when I selected the shortest summary length, it still managed to retain the most important takeaways. It’s a nice break from reading through dense Confluence pages, especially when the content includes lots of screenshots and diagrams. In those cases, listening while skimming the visuals actually made for a really solid experience.

You can also tweak the length (short, medium, long) of the audio and choose the tone (like playful, conversational, or professional) which gives you some flexibility depending on the content. That said, the voice still leans a bit robotic. Compared to other tools which sometimes includes natural filler words like “um” or varied pacing, the Confluence voice feels a little flatter - though not in a way that’s distracting or unusable.

That said, I think (at least for now) Confluence Audio seems very much built with internal consumption in mind. There’s no option to export the audio file, and no transcript you can tweak or repurpose. So while it's great for re-engaging with Confluence content in a new way, it's not quite built for podcast-style output or any kind of modular content creation just yet.

However, the feature is still in beta - and given how solid it already is, I’m excited to see where it goes. There’s real potential here for making Confluence content even more accessible and dynamic.

Summing up...

it looks like I’ll still have to set out to explore other tools for podcasts. Confluence Audio is pretty cool and surprisingly effective for quick in-page listening, but its use cases are quite different. It’s built more for consuming content within Confluence rather than creating standalone assets like podcasts.


Would love to hear if anyone else is experimenting with Confluence Audio or found creative ways to use it!

[Josh]

Great article, @Gabi Mucha! Looking forward to this feature being available for my org in the near future.

[Rune Rasmussen]

Some thoughts on the ever increasing desire to summarize information.
Mostly because I think it's an interesting topic, but also a bit because I am frustrated with the state of some things.

TL;DR: Man on internet questions the benefits and impact of increased options to summarize content and dunks on Atlassian for releasing half-baked features.

Are we risking decision being made based on insufficient information because the allure of not having to ingest all the information available and the modern day mentality of "faster, better, cheaper"?

Are we reaching a point where the amount of information available/new information coming out is so great, that even consuming summarized versions is taking up a significant amount of time.
Just consider how much time one can spend trying to keep up with the changes to Jira and Confluence (at least the changes that are announced).

Could this need or desire for summarized content be a symptom of a greater issue?
If we take Jira/Confluence changes as an example again, there are at least seven different channels to keep track of, and many new features are released in a half-baked state and require regular updates.
Having to read one slightly longer blog post about a change if it was fully (or at least more) baked before release would take less time than reading three posts about the initial change and the following updates to it.

I don't necessarily think that summarizing content is inherently bad, but I do think we (all of us) in general could benefit from having discussions about it, 

[Caity Belta]

I, too, question the impacts of summarizing content. It seems to be an on-trend "solution" (to your point, "to what problem?") from a socio-technology standpoint. Can our brains handle an ever-increasing input of information - summarized or not? I don't know. 

However - as someone who has better comprehension when reading and listening to content at the same time, I welcome the option as presented. As a generally curious person, I get the drive to try and fiddle with it.