After a couple of months of back-and-forth, I’ve decided it’s time to share our journey with the virtual service agent. Why? Because as more of us start using it, the service and development get better for everyone, us included. It’s like a win-win—helping others means we’re growing, too.
Before decided to give a green light, we took a long, honest look at where we were. Without knowing our starting point, it’s pretty impossible to see if you’re actually getting better, right?
Once we had that baseline, we turned on the virtual service agent and started exploring its abilities. And, even it’s still a work in progress, with new features rolling out practically every week. It’s a bit overwhelming, honestly, but exciting, too, to see what this thing can actually do.
One of the fun ways we tested its limits? We hid an "easter egg" in our Confluence knowledge base. We wanted to see if the virtual service agent could recognize that, even though this little “easter egg” text was in our knowledge pages, it wasn’t important. Kudos to the Atlassian team for adding prompt injection detection—so helpful as we play around with the agent’s potential!
As we moved into testing, we started gathering feedback. After a few weeks, though, we noticed something: while the agent was giving answers, they weren’t always on point. That got us asking, “Why?”
Turned out, it wasn’t too mysterious—the data was just coming straight from our knowledge base.
Before we expanded to a bigger audience, we knew our data quality needed a improvement. Even though we activated the virtual service agent not long after its launch, it gave us clear insight into where we had gaps we hadn’t noticed before.
The faster we got feedback, the quicker we could improve our data. Now, we feel ready to open it up to even more users.
The more people use the virtual service agent, the smarter it gets. By getting more folks to try it, we’re creating feedback that strengthens the service for everyone.
Our goals? Simple: help users get support faster, reduce the number of tickets, and let our support team focus on more complex stuff.
Next up? We’re ready to open the virtual service agent to everyone in the company. We’ll keep gathering feedback, making improvements, and pushing it to keep getting better and smarter.
One thing I’m really hoping for? Consistent language support, so users aren’t left scratching their heads over mixed-language answers.
Ready or not, the virtual service agent is here to make support a whole lot smarter
Jana Bohackova
0 comments