MongoDB Compass AI Query Bar
Timeline
5 weeks
Team
Product Manager, Design Lead, Designer, UXR Lead, 4 Engineers
TLDR;
MongoDB Compass is a GUI that allows developers to easily explore their data, but writing queries can still be a complex task. As generative AI tools like ChatGPT gained momentum, I led foundational research to understand how users perceive AI-generated queries and define how this interaction should work at an early stage.
My insights directly shaped the strategy for the AI Query Bar, enabling faster querying for 1M+ active Compass users and laying groundwork for later initiatives like the Compass AI Assistant.
Context: As generative AI gained momentum, we explored how MongoDB Compass could make querying data dramatically easier
During my internship, generative AI adoption was accelerating across the industry. Tools like ChatGPT were changing how developers wrote code, searched for answers, and interacted with documentation.
Compass, MongoDB’s GUI for querying and exploring data, saw an opportunity to reduce the friction of writing complex queries. While the tool removed the need for command-line interaction, users still needed to understand MongoDB’s query language to filter, sort, and aggregate data.
With only 1 month before MongoDB .local New York, I had to quickly plan and execute research before preview
During my internship, generative AI adoption was accelerating across the industry. Tools like ChatGPT were changing how developers wrote code, searched for answers, and interacted with documentation.
Compass, MongoDB’s GUI for querying and exploring data, saw an opportunity to reduce the friction of writing complex queries. While the tool removed the need for command-line interaction, users still needed to understand MongoDB’s query language to filter, sort, and aggregate data.
launched survey!! (n=282)
launched survey with my mentor, i used the data to understand segments and recruit for interview/concept testing. different segments big/med/small, atlas/non-atlas,/low/high amount of queries executed —> intform design concepts and interview guide and recruiment
combined moderated interviews +concept testing (n=12)
mix of AI usage, half view concept a first other half not to prevent bias. 6 users per primary segment to get enough qualitative data - couldn't do more because of tight timeline already doing 12 in one week was a lot!
Concept A:
Concept B
I partnered closely with designer to make changes in real time
Kept regular contact with stakeholders to consult on research plans, project pivots, and product strategy via 1:1s, team meetings, Slack threads, and document comments
Invited PMs, designers, and interested engineers to sessions
Left takeaways and notes after each session in a central document
Shared preliminary insights after the last session with the whole team before I complete a more detailed synthesis and recommendations
Wrapped up projects with a presentation and discussion so the team can be up-to-date on what's already happening and next steps
Recommended to go with query bar with current AI functionality (only query generation) and that a side panel would work for more general AI chatting
Kept regular contact with stakeholders to consult on research plans, project pivots, and product strategy via 1:1s, team meetings, Slack threads, and document comments
Invited PMs, designers, and interested engineers to sessions
Left takeaways and notes after each session in a central document
Shared preliminary insights after the last session with the whole team before I complete a more detailed synthesis and recommendations
Wrapped up projects with a presentation and discussion so the team can be up-to-date on what's already happening and next steps
We moved quickly, but the team strategically announced the feature after the official launch at MongoDB .local London 3 months later
I completed foundational research under a tight timeline to inform the planned preview! However, the team decided not to debut the AI Query Bar in New York and instead aligned the announcement with the official release at MongoDB .local London in September.
Even though timelines shifted, because I delivered early insights, I was able to shape strategy and guide the team’s decisions before engineering implementation.
Impact: Research shaped the design strategy for the first generative AI feature on Compass to enable faster querying for 1M+ active users and provided foundation for later launched AI initiatives
I ensured research insights translated directly into product requirements, helping app analysis become one of AMP’s most mature toolsets. Consultants can now focus more on executing modernization — the code changes and implementation work — rather than spending time figuring out how to approach the project in the first place.
After the initial discovery, I also did deep dives across other AMP product areas like data modeling, data migration, and testing to build a more holistic perspective and understand cross-product relationships. The goal is to move the team toward a clearer north star: connecting tools into a cohesive workflow that could eventually support a more integrated, potentially more self-serve, platform.
