SVP Atlassian (ex-Coinbase, Airbnb, Instagram, Facebook)
San Francisco Bay Area
Joined October 2010
- Tweets1.7K
- Following3.2K
- Followers20.8K
- Likes47.6K
🫶 Agentic work is happening locally on people's machines - but the rest of the team isn't aware of what's happening.
💡 We’re changing that by bringing local agent sessions into Jira. Otherwise, can an agent really be a teammate if no one can see what it’s doing ?
🔥 Now, you can::
➕See active local sessions on your board.
➕Drag a session into a column to create a work item.
➕Bring human and agent work together in one view, with full context.
🤝 This is how the Atlassian's own teams are working - cannot wait to get this out to everybody to try out.
More annoucements 👇
YES to everything @charliedesigned says here. And at Atlassian, we live by this every day.
🎙️ Five quotes from Mike on how we're thinking about design & AI
Normally I do plenty of drafts and rewrites before posting here. Partially just my own pedantic style, but also because I feel like it's important to be as considered as possible. In this case, @mcannonbrookes's interview with @reckless on @verge's "Decoder" podcast was such a thoughtful articulation of how we're thinking about design and AI here at Atlassian...well...this time I'm going to just quote Mike...
These five are just a sampling of what was a fantastic, wide-ranging discussion on enterprise platforms, AI, and the dynamics of our industry.
1/ Design is going to be key for differentiation
"...if the cost of building goes down, the quality and differentiation is on the experience and the design, and that’s not something that’s easy to do. That requires a huge amount of taste and judgment."
2/ That differentiation path is a long term bet
"We have a huge investment in design and experience. Especially in the AI era, that’s getting ever more important...We’ve come a huge distance in the last five years. In the last two years, we’ve been truly world-class at doing it."
3/ Building scalable design systems is key
"...in the role of design, 50% of their job is to make our design system and to keep it scalable, flexible, and buildable. 50% of their job is to solve the hardest 5% of problems in design...they’re unleashing thousands, if not almost tens of thousands of engineers to build better-designed things when they do that"
4/ The design role will continue to be a deep speciality in its' own right
"...designers can evolve that system and the features that are being built will continue to match what customers expect in that congruence and consistency. That gets us huge leverage overall, but it’s a non-trivial thing. It doesn’t mean the designers are going to go away..."
5/ Making well-crafted interfaces is still hard and still vital
"...Our job is to bring all [everything] together into a coordinated surface, to understand your agentic sessions, your human sessions, your work, and to connect all of your enterprise applications into a giant graph and give you phenomenal answers across the whole thing. That requires a lot of interface investment. That is hard. That takes time. That takes great design and experience folk."
As a mate of mine posted yesterday (hat-tip @alexnako ) - 'It's been a while since a CEO interview landed like this one'. I'm glad more people who perhaps don't know us so well, have the chance to see the depth of appreciation and strategic viewpoint that Mike has on design. Yes - there is always more to do, and things are ever-changing. But our direction is clear.
Check it out here ->
theverge.com/podcast/1000914…
Definitely a must listen from @mcannonbrookes !
94% of engineering teams have adopted AI. Only 6% are actually set up to run an AI-native SDLC.
The missing piece? Context.
Our CEO & Co-Founder @mcannonbrookes and @vercel CEO @rauchg discuss the future of software development and how real results come from connecting agents to organizational knowledge and intent.
Full session below. ⬇️
Sanchan S Saxena retweeted
There's a lot of talk about software factories right now. Frankly, most of it is bullshit.
Beware “I do AI better than you" snake oil and think more from first principles. Strip most of them back, and you find a series of automations, and one of them happens to be a coding agent. (Which is also what CI/CD was. You stopped clicking around testing by hand, and a machine did it.)
There’s a lot right with making your idea → ship loop continually faster and smarter, but I’m not sure that’s really a new idea.
Nobody starts with a factory.
A chocolate factory starts as one person and some beans in a kitchen, hand-making chocolate. Then they get a mixer. A bigger fridge. A conveyor belt. Then a packaging machine. Each addition builds scale, consistency and efficiency. A small business becomes a medium business, which becomes a factory.
What changes along the way? Where the person stands. What the person does and the technology they can access. First, they do every step. They buy a machine or two. Then they do some steps and watch the machines. Then they operate and connect the machines. Then they redesign the factory itself.
Software is the same loop. Ideas are dreamed up, feedback and operational bugs come in, someone turns them into a list of work items, someone builds it, someone reviews it, someone builds it, someone deploys it, it ships, something breaks, someone notices, someone files the bug. Rinse. Repeat. You've run that for years with a person at every station.
To improve the factory you already have, upgrade one station today. Add automations and AI agents to code or review. Observe, learn. Then do the next one. Thoughtful, ambitious, patient scale.
Are there roles for humans in that factory? Abso-bloody-lutely. Initiating ideas, observing the operation, reviewing at discrete checkpoints or severities, co-designing both the flowed work and the factory itself, owning exceptions and resolving conflicts, adding context. (Just look at one of our latest Loom action plans demo👇🏻 for an example of human-AI collaboration)
I believe this human judgement and intuition only get more important as factory throughput scales.
In short - don’t believe in the mythical agent month. You already have a software factory. AI just allows you to make more steps repeatable, more flexible - and that lets you move where you stand.
Atlassian used Transcribe 2.0 in @Loom and found it more accurate at capturing user instructions. Dictate your change requests using Loom, then directly export to Cursor to code it up.
🗓️ For years, Jira has been the system of record for how software teams work. Today that foundation starts doing something new: coordinating agents alongside engineers.
🛳️ We’re shipping the system that moves teams from one-off agent sessions to governed execution loops across Jira and DX:
→ Code Context gives agents secure, repo-level intelligence grounded in your real architecture and decisions
→ Agent loops in Jira turn well-defined backlog items into ready-to-review pull requests
→ Standards and AI Review hold every agent to the same quality bar before code reaches a human
→ DX for Agentic Development ties AI spend directly to what you actually ship
🤝 The hard part of scaling agents was never model intelligence. It was organizational context, governance, and proof of impact.
🔨 So we built all three into the tools your teams already use.
🫶 Developers keep the merge button. They just stop being the bottleneck for everything leading up to it.
🤔 What happens when agents don’t have details? They fill in the blanks on their own and move confidently in the wrong direction.
🧙♂️ Jira Planner turns rough ideas into agent-ready specs that are grounded in context.
The difference from doing this in a chat window:
⛱️ Grounded in your real codebase. Planner reads across multi-repo setups and pulls in context from the Teamwork Graph. Your architecture, your ownership structure, your previous decisions. Not generic assumptions.
🤝 Multiplayer, not single-player. Plans publish as Confluence Live Docs. Engineers, PMs, and designers co-edit, leave inline comments, and work through trade-offs together while the plan is still taking shape, not after it's final.
🫶 From plan to work. Planner breaks the finalized spec into Jira work items — sequenced, with dependencies and acceptance criteria — each carrying the full context of the plan. Agent orchestration picks up exactly where planning left off.
Now in early access. Learn more and join the waitlist:
atlassian.com/blog/jira/intr…
🤔 For a decade plus, slides have lived in a separate tool and a separate file, disconnected from the work they’re actually about.
🧍 You rebuild context that already exists somewhere else, and it goes stale the moment you hit present.
🚀 That’s what we set out to fix, and today I’m excited that Confluence Slides is generally available for everyone.
🔥 Now your presentations live right next to your docs and projects, connected to what your team already knows. A few things I love:
📊 Slides live right where your work lives, on the same pages and data your team relies on
🤖 Rovo turns a rough doc into a real deck, so you spend less time formatting and more on the story
🔄 One source of truth, so when the work changes, your slides change with it
A presentation was never meant to be a separate thing you rebuild from scratch. It’s just your team’s knowledge, told well.
Check it out 👇
youtube.com/watch?v=1sjJfS2I…
🤔 Without the right context, every day is day one for your coding agent.
✖️ They don’t know why certain tradeoffs were made
😱 They don’t know which services depend on what
🧐 They don’t know what’s going to break downstream
🚀 So today, we’re introducing Code Context into the Atlassian Teamwork Graph.
💥 Code Context taps into knowledge from the Teamwork Graph to connect to Jira work items, Confluence decisions, Rovo and more.
💡 It gives agents a more complete understanding of the work, so they can produce better outcomes, and can catch cross-repo dependencies before they become shipped errors.
🫶 Any agent (Cursor, OpenAI's Codex or Anthropic's Claude Code) can use it through the Teamwork Graph CLI, and accountability and governance stays with developers through its permission-aware design.
➕ One context layer to reason over. That’s the difference between an agent that ships and one that creates work.
🚀 Here’s how we’re building the future of Atlassian' Confluence: 🤖 the UI is the API.
🤝 Agents can do nearly anything a person can, with no special doorway. Anthropic's Claude, OpenAI's Codex, Gemini, or any tool built on MCP meets your team on the same canvas.
🎨 Because the canvas still matters. Confluence is the shared visual space for knowledge, now shared by humans and agents alike.
Today we shipped three steps toward that future:
💬 @mention an agent in a doc or comment to hand off a task or co-edit in context
👤 agent avatars that show presence when one is actively at work on a page
📊 enterprise-grade analytics on which agents viewed your pages and live docs, for full transparency
♥️ This is just the beginning.
🫶 Most teams already explain complex work by showing it.
💨 A designer walks through a prototype. A PM points out a UX issue. An engineer records the exact bug path.
🫂 Atlassian' Loom video prompts make that same natural workflow useful for AI agents too.
🏹 Record your screen, talk through the task, and Loom turns the video into a structured action plan with transcript, clicks, links, and visual keyframes.
🧙♂️ Share it with an agent like Rovo, Claude Code, or Cursor, or create Jira work items in a few clicks.
💨 This is a faster, more human way to work with AI: stop engineering prompts and start showing your agents what to do.
🚀 Go give your agents the full picture:
🚀 Working in @OpenAI 's ChatGPT? A user asks ChatGPT to review a @Atlassian 's Jira epic, and it finds the gaps, creates the missing tasks, and links everything back together.
💨 No prompting gymnastics. Just an agent doing real work.
🚀 That’s happening at scale right now. The @Atlassian Rovo MCP Server is seeing 5M+ tool calls every working day. Nearly a third are writes.
🔑 Agents aren’t reading your work system. They’re participating in it.
Check out some stats:
💡 44% of users aren’t on software teams. PMs and C-level execs are using MCP just as intensely as engineers, because their entire job is connecting dots across teams.
🏹 More than 50% of monthly active users are enterprise customers. Enterprises almost never move first. They’re moving here because the bigger the org, the more a connected context layer is worth.
🔥 The flywheel is already spinning. Every agent write enriches the Teamwork Graph. A richer graph makes the agent smarter. We’re seeing 44% more accurate answers with 48% fewer tokens.
And its not just one - agents like @claudeai Code by @AnthropicAI , @ChatGPTapp by @OpenAI , @cursor_ai , @figma , and more are actually being used in enterprise work.
🥇 Jira Slack Agent now has new features that uplift your workflow.
🫶 Jira slack agent has existed for more than a year now - and we have millions of monthly active users of that agent in slack.
💥 Based on working with our customers, we have learnt a bunch of new ways in which we can improve this and now we are officially launching it to you all.
🏹 With the new Atlassian Jira for Slack app, you can now turn any conversation into a tracked work item the moment it happens. Just mention @Jira to:
✅ Create & update: Turn threads into structured Jira tasks.
✅ Capture context: Automatically pull in decisions, links, and team knowledge from the thread or channel.
✅ Orchestrate work: Assign tasks to teammates or AI agents directly from chat.
🚀 We just shipped @claudeai Agent for @Atlassian's Jira.
💡 The idea is simple: assign a Jira work item to @claudeai , get a draft pull request back. No context-switching. No copying specs into a chat window. No setting up branches manually.
🏹 Here’s what makes this different: the work item IS the spec.
💨 When Claude picks it up, it’s not starting from scratch. It already carries the requirements, the acceptance criteria, the designs, everything — powered by Teamwork Graph, the connective layer across Atlassian that gives agents real context. That’s something only Atlassian can do.
🫂 The agent reads all of that, codes in a secure sandbox, and opens a PR for your team to review.
💻 This is what we mean by the AI SDLC. Not replacing developers but removing the friction between planning and building.
Three things I love about this:
1️⃣ It works where teams already work. Inside Jira, via the assignee dropdown.
2️⃣ Humans stay in the loop. Every PR is a draft, reviewed by your team.
3️⃣ It’s one of many paths. Use it managed in the cloud, or connect Claude locally via our MCP server.
🚀 Less prompting. More shipping. Take your context to your agent.
🫶 You can now deep-link from any Atlassian's Jira work item directly into your AI coding tool of choice, like OpenAI's Codex, Claude Code, or Rovo Dev CLI.
➕ Rich prompts with clear requirements, contexts, and agent instructions produce better code.
💥 No more copy-pasting context—just stay in your flow and get back to the code.
🔥 We just showed the world what happens when AI agents actually understand your company — not just your code.
🤩 The demo is simple. Two Claude Code agents. Same repo. Same task. One is flying blind. The other has the Teamwork Graph.
🦯 The blind agent? Dove into local files, hallucinated a schema, and confidently shipped a plan that would’ve broken production with duplicate notifications. Classic AI: fast, confident, wrong.
🧠 The agent with the Teamwork Graph? It did something no one expects an AI to do — it stopped and thought first. Pulled up Confluence docs.
Found a hidden dependency in a separate repo. Built a mental model of the system. Then wrote the correct solution, faster and cheaper.
🔑 This is the unlock: AI agents that work the way great engineers work. They don’t just read code — they read the room. The design decisions in your pages. The dependencies across your repos. The product intent buried in your PRDs. All connected, all permissions-aware, all in real time.
🛞 And here’s the flywheel: the more your team works across Atlassian, the richer the graph gets, and the smarter your agents become.
🤝 Context isn’t a nice-to-have. It’s the difference between an agent that ships bugs and one that ships features.
@Atlassian 's TEAM '26 Anaheim - what an amazing event to be with customers, partners, and our team members to announce some of our biggest product launches.
🔥 Orchestrate agents in Jira --> now generally available. Orchestrate, manage, and govern your favorite agents Figma, Canva, GitHub, and more.
🏹 Teamwork graph (your context graph) - 44% increase in quality of output and 48% less token usage by your agent, when you use Atlassian's Teamwork graph.
💥 AI Planner in Jira: Create your spec for your work, define technical designs and provide that as an input to your favorite agent.
💡 Confluence Remix and Slides: Now you can remix any content into visuals, infographics and turn any page into an AI generated slide deck that fully matches your brand guidelines.
🚀 Loom as an input for your agent: Just describe what you want to fix or change in your product. Loom records your voice, clicks and all the necessary data to feed to your favorite coding agent.
🚀 Stop tool-switching and start shipping with agents in Jira. 🚢
🏹 Agents are everywhere in @Atlassian 's Jira - from GitHub Copilot and to many more.
⚡️ See how our AI Product Manager, @matthewcanham , uses the GitHub Copilot agent to kickstart work while staying in the loop to refine and finalize the results.
🫶 Confluence Remix enables anybody to tell a story, in ways that resonate with them.
🏹 No longer are you limited in expressing or consuming content in only one format - your imagination is the limit on how you express your work.
🔥 You can express your work as an infographic, or a visual image, or a flip card - basically, anything you can imagine.
🥇 And we take this same philosophy a step further with the 3P agentic workflows we're building with @Lovable , @Replit , and @GammaApp. Now knowledge doesn't stop at the borders of @Atlassian , as users can go directly from idea to presentation and prototype.