Waiting Before Saying “We Added AI”

[#1]

When AI started showing up everywhere, it was easy to think “We should add it, too.” But we wanted to pause and ask a more basic question: “Where does this actually help?” Instead of diving in headfirst, we stood back and watched how people worked. We noticed that while everyone knew diagrams were useful, making them felt like a chore. What if users could just say what they needed—and get it?

That meant teaching the tool to handle natural, everyday language. No fancy syntax, no special phrases. We focused on how people really talk: “Give me an activity diagram in PlantUML which explains the process of checking out in a webshop.” Sure, it sounds simple, but these requests can be a bit vague or open-ended. That’s real life, though—ideas start off messy.

We learned a few lessons along the way, which might be helpful if you’re working on something similar:

1. Don’t just add AI because it’s popular. Find a spot where it genuinely reduces hassle. For us, diagrams were a prime candidate—turning a fiddly, time-consuming task into something as natural as typing a sentence.
2. Accept that people don’t speak in perfect instructions. Designing for everyday language is trickier than crafting a strict command set, but it’s worth it. People shouldn’t have to think like a machine to get their work done.
3. Trust matters. If the tool isn’t sure what you mean, it should say so. Better to admit uncertainty than to guess blindly. Users appreciate honesty.
4. Integration can be messy. Working inside a platform like Confluence comes with its own challenges. We made sure everything works quietly in the background so all you notice is how smooth it feels.

By taking our time and focusing on a specific problem—making diagrams appear from plain descriptions—we ended up with something more useful than a random “AI” sticker. We’re not here to say our approach is the best or to put down anyone else’s strategy. If anything, the takeaway is that careful thought pays off.

If you’re exploring ways to incorporate AI, consider where it really makes a difference, how people naturally communicate, and how to handle those inevitable gray areas. We didn’t rush in waving a banner; we took slow steps toward something that feels more natural, both for the user and for the tool itself. In the end, it’s about making things just a bit easier—without making a big fuss about how it’s done.

In the end, we aimed for something that feels more natural and genuinely helpful, rather than slapping “AI” on the label. This thinking shaped the new Visualizer for Confluence, now able to take your everyday words and turn them into useful diagrams without a ton of extra steps.

If that approach sounds like something that could help you, feel free to check it out on the Atlassian Marketplace. Check out our documentation for more information or contact us via our support channel. We hope it makes your day-to-day just a bit easier.