Atlassian · Sep 2025–Present
I’m leading the design work to modernise Jira’s work item. The goal is to bring a twenty-year-old product up to date, with a UI that aligns to the Atlassian Design System, new AI features, and a structure organised around what customers are actually trying to get done.
As Product Design Manager, I’m leading the work to modernise Jira’s work item, the screen that millions of teams use every day. The aim is a modern UI that aligns to the Atlassian Design System, is restructured around customers’ jobs-to-be-done, and has AI built in where it genuinely helps, all without disrupting the workflows people rely on.
I set the vision and the quality bar and led six designers, seven PMs and over eighty engineers. I kept everyone aligned through a leadership design sprint, a shared ‘north star’ we used to make day-to-day decisions, and a few simple team rituals. I made the call to only add AI where a job actually stalls, and I designed the highest-stakes journeys myself (quick-create, agent sessions and the developer experience).
The work is live and ongoing. Performance is 30% faster, the work item is back in line with the Atlassian Design System and reorganised around jobs-to-be-done, and AI now earns its place against real customer needs as we work towards ten million agent invocations a month. Six parallel projects and eighty engineers are all pulling towards the same vision.
Jira’s work item is the screen that millions of teams work in every day. My challenge was to modernise a twenty-year-old experience without getting in the way of the workflows people depend on.
As Product Design Manager I set the vision and the quality bar, worked closely with the Group PM and Head of Engineering, and managed the workload and projects of seven designers at various levels, while also designing some of the highest-stakes journeys myself. The result is one modern, consistent experience that aligns to the Atlassian Design System, is reorganised around what customers are trying to do, and shipped as eight core improvements.
Jira is twenty years old and it shows. Two decades of adding features one at a time had stretched its visual language thin and let it drift away from the Atlassian Design System. It was organised around the order features were built rather than what customers actually needed to do, and its speed and performance had dropped below what people now expect.
Then leadership set the ambition:
“Make the experience iconic. A 15/10.”
In practice that meant a modern UI aligned to the Atlassian Design System, reorganised around customers’ jobs-to-be-done, and much faster and simpler to use. The catch was that millions of daily workflows depend on this product, so I had a hard brief to hold. I needed to change everything, but disrupt nothing.
Partway through, the definition of success changed. Atlassian was shifting its business model away from charging for paid seats and towards usage-based pricing tied to AI usage, so the metric that mattered moved from monthly active users to AI agent invocations, with a target of ten million a month. That one change reframed the whole product. The work item was no longer just a place to manage tasks, it became the place where people and AI agents work together, and that shaped every decision about where AI did and didn’t belong.
This was a big, cross-functional stream and I led it through partnership. Six designers reported to me and I deployed each of them to their strengths. They worked alongside seven product managers, including a Group PM I partnered with every day, and more than eighty engineers across the wider work item stream. Beyond the core team I worked closely with the Head of Product, the Head of Engineering, and people in brand, marketing and the other surfaces our work touched.
Most of my leverage came from alignment and deployment. That meant putting the right designer on the right problem and keeping three org leaders pointed at the same outcome. Where the stakes were highest, I stayed hands-on and did the design myself.
Day to day, my job was to get the best work out of six designers at different levels. I ran regular one to ones, gave feedback often, and matched people to the problems that suited their strengths, so they could do their best work and keep growing as they did it. Where I could, I let them lead their own areas and make the calls, rather than routing every decision through me.
I owned the design strategy for the work item and set the bar for what good looked like. I turned the leadership ambition into clear quality principles the team could actually build against, then held the work to them through critique and review. Keeping that bar consistent across six parallel projects was a big part of the job, so that everything we shipped felt like one product rather than six separate efforts.
A project this size only ships if the people around it are aligned. I worked closely with the Heads of Product and the principal PMs on the product strategy, making sure the design direction supported where the product needed to go. I managed the team’s capacity, flagged trade-offs early, and kept design decisions moving so we could hit our delivery dates without dropping the quality bar.
I also contributed beyond my own stream. I took part in regular design reviews with design leadership, sharing our work for critique and picking up what other teams were learning, and I helped keep our practice in step with the standards and patterns used across the wider design function.
At this scale, the process is the design work. Three things set the team up to move quickly without pulling in different directions.
Before anyone opened Figma, I ran a design sprint with the Heads of Product and Engineering to agree what we would actually ship, then turned that into a shared roadmap. At the same time we translated ‘iconic’ into concrete quality bars and experience principles the team could build against, and we grounded the restructure in customer conversations and usage data, so the new information architecture followed real jobs-to-be-done rather than opinion.
We created a single view of where we were heading together with lead designers, product and engineering, so everyone owned it rather than just being handed it. It became a tool we used to make decisions day to day, not a poster on the wall. It settled a year’s worth of trade-offs before they turned into debates and kept six parallel projects and eighty engineers moving towards the same picture.
A few agreed rituals, like critique, design reviews and decision logs, kept the quality and the reasoning visible to everyone. We split the surface into projects that matched each designer’s strengths, then used shared patterns and the north star to keep those parallel efforts adding up to one coherent experience.
With a ten-million-invocation target, the obvious temptation was to put AI everywhere. But there was a cautionary tale playing out at the same time. Microsoft had pushed Copilot into nearly every surface, including Notepad, the Snipping Tool and floating buttons over the canvas, and while engagement went up, so did the backlash. People felt the software no longer belonged to them, and by March 2026 Microsoft was publicly reducing unnecessary Copilot entry points across its core apps.
The lesson was clear. Visibility drives clicks, not value, and an invocations metric is easy to game with clutter. So instead of scattering AI across the interface, we checked every possible entry point against a real customer job, researched the shortlist, and deliberately cut it back. AI now shows up where a job stalls, not everywhere we could put it. We chose discoverability over ubiquity.
Behind those decisions was a shared set of principles for how people and AI should work together. Interactions should be proactive, using what we know about the person, team and company to help at the right moment, and they should stay in the flow so they feel natural and low-friction. They should be dynamic, letting people move between text, audio, video and visual formats, and they should always be inspectable, so anyone can look under the hood to understand a result or get to the detail beneath it. And they should be collective, connecting individual work to team knowledge while respecting permissions and moving smoothly between people and agents, in a way that still feels unmistakably Atlassian.
The work landed as eight improvements that add up to one coherent experience. A new quick-create journey made it faster to capture work. Every element in the work item was visually aligned and made consistent across the whole of Jira, and the content hierarchy was rebuilt around what customers need. We added a cleaner empty state and small moments of delight, applied AI generously but with restraint, and improved load times and performance by 30%.
‘Iconic’ lives in the details, and quality at this scale isn’t one big move, it’s a thousand small ones held to the same standard. Every element sits on one strict grid, the kind of order you feel without noticing, and the information hierarchy was rebuilt from customer need so your eye lands in the right order. White space is treated as structure rather than something left over, which gives a dense tool room to breathe. A single icon language from the Atlassian Design System runs across the work item, the palette stays neutral by default and only uses colour with intent, and illustration is used sparingly, to add warmth to empty states and first-time moments rather than for decoration.
Leading the stream didn’t mean stepping back from the craft. These are the journeys I owned end to end, the highest-stakes calls where I kept my hands on the decisions.
I chose to let people create work in context rather than on a separate screen. You capture the essentials first and add the detail later, so starting a piece of work never pulls you out of what you’re already doing.
Agents work in visible sessions that people can inspect and steer, rather than being a black box. It’s the ‘inspectable’ principle applied to the moments where trust matters most.
Rather than duplicate the development cycle, I designed the work item to meet developers where their work already happens, connecting to the tools and stages they use instead of asking them to come to us.
The work is live and ongoing, and the direction is already clear. Performance is 30% faster and the work item is back on-system, aligned to the Atlassian Design System and consistent across Jira, with its content organised around jobs-to-be-done rather than the order features were built. AI now earns its place through entry points chosen against real customer jobs, working towards ten million agent invocations a month without the clutter that causes backlash. And for a stream this size, the thing I’m most proud of is that six parallel projects and eighty engineers have ended up pointing at a single, shared vision.
Aligning the three org leaders in a sprint before design started paid off again and again. Later trade-offs were quick because the intent was already shared. Treating the north star as a working tool rather than a poster kept the team moving in the same direction, and deploying designers to the problems that suited their strengths lifted quality, ownership and pace over time.
The success metric changing partway through cost us real rework, so next time I’d pressure-test the measures with leadership much earlier. Some early rollouts moved things customers relied on, which taught us to pair every change with proper migration support. And ‘iconic’ stayed vague for too long, so I’d push for a concrete definition of it in week one.
Three things I’ll take into the next problem of this size. Frame the problem before the solution, and do it with the people who own the outcome. A shared vision beats a hundred review meetings. And you can change everything without breaking anyone, because respecting existing workflows is part of the craft, not a limit on it.