🌱 How Rovo Deep Research supports our team’s expertise

[#1]

Welcome back to our series! We're so glad you're here. In today's article, Atlassian Learning's Rovo champion, Michelle Cacciapaglia, shares how Rovo’s Deep Research supports her creativity and allows her expertise to shine. We also share helpful information you can apply to your own research.

By: Michelle Cacciapaglia, a Senior Instructional Designer on the Learning Content Design Team

The scene: When an Instructional Designer on the Learning Content Design team plans to create a new course for the Atlassian Learning catalog, they first perform a needs analysis to find out everything they can about the audience for their new training. They research their audience’s roles, core tasks, goals, and challenges. Then, they learn about the Atlassian apps and features that can support their audience and help them overcome challenges. They comb through internal and external user research, feedback, case studies, and more. As you might imagine, a needs analysis takes a lot of time, leads to many dead ends, and can be tedious.

When Atlassian first announced Rovo’s Deep Research capabilities, Michelle, the designer behind our Get the most out of Rovo learning path, saw an opportunity to have AI assist so she could focus on creative thinking and building the best possible learning content.

Read on for an example of human + AI collaboration that elevates an entire team of human expertise and creativity while offloading repetitive, tedious work to a machine designed to effortlessly handle it all.

You'll also see how you can apply the Rovo Deep Research lessons Michelle learned to your own research.

---

Why I needed help with research

Before I begin writing a new course for Atlassian Learning, I spend about a week looking for relevant documentation within Atlassian. I look for details about features and functionality, user research other teams might’ve done to learn more about our customers, and more. But we have 10,000-plus employees documenting in all kinds of ways in various places, and it can be tricky to hunt everything down. I don’t know what I don’t know, essentially.

Even if you don’t perform need analyses at your job, I think the process we go through is a common situation: you’re at work, you’re trying to find information internally, but there’s so much out there, and it’s not necessarily organized logically, or in easy-to-find places, and often people create similar types of things that might overlap or duplicate. Whether you’re using Confluence, SharePoint, or some other type of intranet, trying to find what you need can be super time-consuming and frustrating. Clicking on pages and skimming them, and then clicking on more pages and skimming them, and trying to make sense of it all, feels so monotonous and draining.

When I heard about Rovo’s Deep Research capabilities, I thought it might be able to help.

Refining how I use Rovo Deep Research

First, I started out with a really basic test-run prompt in Rovo Chat, as if I were planning to write a course about Confluence: “You are an expert in instructional design. Help me conduct a training needs analysis to identify what new users need to know when using Confluence.”

I was shocked by how much information Rovo Chat found (86 documents!) and how it analyzed and pulled it all together into a report so quickly. What used to take a week or more took only minutes. Rovo also created a research plan with questions, even though I didn’t give it much information.

The output wasn’t perfect, but I was impressed with Rovo’s capabilities.

Next, I wrote a long, detailed prompt, hoping to get better results. The prompt included direction on Rovo’s persona, as well as my own:

• “You are an expert in instructional design, training needs analysis, and user research. I am an instructional designer assigned to design an instructor-led course for customers onboarding to Confluence.”

Then, I added details about the project and the audience and told Rovo:

• “Your job is to help me conduct a training needs analysis and target audience analysis for my project. You will search existing data, analyze it, and synthesize it.”

I added a bulleted list of key information I needed, including the most important personas to focus on, top use cases, common challenges, and existing case studies and scenarios. I even asked Rovo to make Atlassian subject-matter expert suggestions for me, so I could get their input about the project.

This time, because I had been more specific about what I was looking for, Rovo searched thousands of documents and found 130. They were all more relevant than the first batch it came up with. Then Rovo analyzed and summarized the results, saving me not only time but also mental energy. Because Rovo respects privacy permissions, I knew I wouldn’t be looking at anything I wasn’t approved to see.

Then I began experimenting with the prompt in my work. I asked my team to try it and give feedback, and we made improvements. We added date parameters (so Rovo didn’t go back too far and find outdated documents), and we added a few other specific research questions based on the work we usually perform.

We used the prompt for a few months and were pretty happy with what we had, but I wasn’t quite getting what I needed. Rovo was finding really great documents, but in its analysis, Rovo was making inferences about the data and trying to design the training for me, which I didn’t need or want. What I needed was summarized data that I, as the expert, could then apply to my instructional design process. I also saw how Rovo was addressing every question I asked (the second phase of our prompt included a LOT of questions) and answering them all, whether or not it had good information to support the answer. I also saw a lot of overlap within the analysis, and it wasn’t as deep as I wanted.

I realized Rovo was trying to BE an instructional designer (because I told it to), which is why it was making inferences and designing training rather than focusing on the research. But with Rovo Deep Research, the persona is baked in: it's meant to retrieve information, research, synthesize, and report on what it finds. We just give it tasks and parameters. With Rovo Agents, on the other hand, there’s so much flexibility that giving it a role is important.

Finally, for the third iteration of my Rovo Deep Research prompt, I wrote a much simpler, shorter prompt that doesn’t give Rovo a role or mention training.

See the prompt our team uses below. You can modify it for your own deep research needs.

This refined prompt is about one-third the length and way less complex than the second iteration, and we are so much happier with the results.

How you can use Rovo Deep Research

To try it out, open Rovo Chat, write or paste your prompt, then select Deep Research as the chat mode. For the latest steps and screenshots, see our support doc.

Have you tried Rovo Deep Research? Did you experiment with our prompt? Let us know in the comments!

If you’d like to learn more about Rovo, check out our free, one-hour public Rovo class or our free on-demand Rovo learning path. Or, try a use case or prompt from our library.

AI use transparency: We used a Rovo Agent to organize a Loom interview Julia and Michelle conducted to get ideas for this article flowing (stay tuned for an article about how we did it), and we used AI for a final grammar proof (but we didn’t take all of its suggestions😉).

Keep an eye out for more "Behind the scenes with Atlassian Learning" posts. 🌱 

Read the whole series

[Becky Mueller]

As a member of the Learning team, I can attest to how much Rovo Deep Research has helped with my needs analysis, and Michelle's prompt was incredibly helpful!

[Bill Sheboy]

Hi @Julia Eddington 

Thank you for sharing this information.  Yes, and...thanks again as this is the first (and only thus far) community article I have seen from an Atlassian team member to disclose use of AI tooling to compose the content.

For community members who want learn about such AI-tooling disclosures, please see here and here.

Kind regards,
Bill