Atlassian

Designing reasoning primitives for AI summary

Every Atlassian app has Rovo, an AI assistant that can summarize any page. The problem was that the output was a wall of unstructured text. No hierarchy, no context awareness, no usefulness. I designed a generative summary system built on reasoning primitives: a modular reasoning engine, a JSON output schema, and persona-aware logic that adapts what surfaces based on who's reading.

The summary evolved through three versions: unstructured (raw AI output), structured (scannable with a health badge, headline, and key signals), and actionable (interactive next best plays with confidence tags). The system shipped across three enterprise GTM intelligence products with 2,100+ monthly users and 1,255 sellers reading the summary daily.

Atlassian

Designing an AI-powered enterprise sales dashboard

SalesEdge is Atlassian's internal sales intelligence platform. When I joined as Design DRI, there was no single surface where a seller could answer ‘where do I stand this quarter.’ I found this gap during a UX audit, pitched it to the Principal PM, and designed a pipeline dashboard that became the most-used page in the product with 2,440+ daily active users.

After shipping the first version, I kept pushing the dashboard further. I layered AI progressively deeper into the same surface: from intelligent assistance with next best plays, to agentic workflows where sellers delegate complex tasks to a Rovo Agent, to generative UI where sellers create custom charts through natural language. I also stayed in the codebase during QA, merging PRs to fix design system issues myself.

I'm a product designer and design engineer with 6+ years of experience across enterprise SaaS, AI product surfaces, and design systems.

Drop a mail to connect or collaborate at howdy@som.design