When I worked at Atlassian, I did Strategy & BizOps for Jira’s integrations.

Basically, plugging the dev stack (Git, CICD, security, feature flags, analytics etc) into Jira to power these epic first-in-market contextual experiences within Jira’s Board, Tickets, Backlog etc (make GitHub PR status updates, triage new Snyk security vulns, inside Jira).

Most of that job was creating growth models, business cases, and strategy memos to show how the more of a team's work that's connected to Jira, the more useful Jira gets.

Occasionally, those pages get pulled into discussions well above my pay grade 😬

Back then, the reader of all that connected data was a human: an engineering manager squinting at a dashboard.

Today, it's primarily agents reading the data.

Turns out in 2023 we were building context for the coding agents who arrived in 2026.

We just didn't quite know it yet.

Today I got to sit down with the Jira team to close the loop on how the exact stuff I was working on years ago now powers agentic coding inside Jira.

Enjoy!

— Tom

Rankings rarely crash. They slide a few positions at a time, until a competitor’s post takes your place and the demo requests stop.

Ploy’s agents watch your rankings, traffic and content gaps around the clock. When a page slips, they rewrite it, fix what’s broken and test the new version, so the site keeps earning pipeline.

Built by Webflow’s founding CTO and used by teams like Clay and Hex.

(Strategy Breakdowns readers get 50% off their first month - use code Tom-Alder)

Email and SMS work best when someone tunes them every week, and that someone is usually busy with inventory, launches and everything else on the list.

Omnisend AI writes the copy, picks the send time and builds the segments, so campaigns go out on schedule with far less manual setup. If the switch feels like the blocker, the migration team moves the whole setup in 5 days.

It’s part of why 150,000+ brands use Omnisend, averaging $79 for every $1 spent and rating it 4.8/5.

_{Thank you for supporting our sponsors, who keep this newsletter free.}

The what: A TLDR explanation of the strategy

Over half of all code is now written by AI.

Tech companies’ spending on AI is up roughly 28x.

But developer throughput is up only 1.37x.

So the bottleneck has moved.

Code gets cheaper every quarter → deciding what to build (and proving it's safe to ship) does not.

Today's breakdown is in partnership with Jira by Atlassian, and it's a story about the humble ticket, going through its most important transition in decades.

24 years after Jira started life in 2002 as a bug tracker, that ticket is becoming something new: the instruction a coding agent works from.

Atlassian is trusted by 350,000+ companies, from 85% of the Fortune 500 to 60% of the Forbes AI 50.

Millions of teams already write their work down in Jira, with years of history behind every project.

That history is exactly the context an agent needs.

And for teams building real software (not prototypes or hobby projects), shared visibility and context beats every developer prompting and praying in isolation.

Let's unpack the new Jira.

The how: The strategic playbook boiled down to 3x key takeaways

Natural language prompting has shifted the unit of development from coding syntax → human intent.

But whether it’s a human or an agent writing the code, without the right context, you’ll build the wrong thing.

For 24 years, the place teams write down intent has been the Jira ticket.

The requirements, the decisions, who owns what, and the “why”.

So one of Atlassian’s unfair advantages = wiring that history straight into your agents:

→ Jira Planner reads your codebase plus your Jira and Confluence history, then drafts a structured technical spec (spec-driven development, i.e. agree the plan in writing before any code is generated).

→ Code Context, which dropped in August, indexes code across GitHub and Bitbucket repos into the Teamwork Graph (basically Atlassian's map of 150B+ objects and relationships linking people, goals, code and content).

→ The Teamwork Graph CLI and Atlassian MCP hand that context to Claude Code, Cursor or Codex, so an agent doesn't accidentally build something the team decided against 6 months ago.

The payoff of planning more complex work:

→ up to 44% better answer quality, using up to 48% fewer tokens.

Product teams are feeling the same shift too. When Atlassian surveyed 1,000 product professionals, 86% of PMs agree that deciding what to build matters more than ever.

Let’s take it further.

Jira ships its own coding agent (Jira Coding Agent) in every paid plan, but it doesn't make you choose it.

You can assign a work item to Claude Code, Cursor or GitHub Copilot (Codex coming soon) the same way you'd assign it to a human.

The agent picks up the ticket, reads its context, writes the code, and comes back with a pull request.

Then Jira keeps score:

→ Agent sessions puts every running agent in 1 view, sorted by what needs you first: stuck, awaiting review, complete.

→ Coding agent automations route the chores (bug fixes, vulnerability fixes, test generation, doc updates) to agents automatically, then ping an engineer when the PR’s ready.

→ DX AI cost management pulls spend from Claude, Cursor, Copilot and Jira into 1 place and estimates the cost per PR (the unit economics that barely existed 2 years ago, but every CTO needs to know today)

Atlassian runs all of this on itself first (we broke down that dogfooding loop in Jira built Jira).

6,000+ engineers running 1,000,000+ live agent sessions across 20+ products. The summit session shows how they did it.

When the hottest model provider changes every quarter, build the layer every winner has to plug into. Every agent welcome.

More code, faster, creates a new problem: someone has to check it (still).

Honeycomb's Liz Fong-Jones said it best at the summit:

But with so much code to review, the fix is still an engineering problem initially: architecting hard, deterministic rules in the pipeline (linters, tests, policies).

Once all automated checks pass, the human who owns the ticket (and therefore responsible for the outcome) reviews all the context, and merges the PR (or sends back to the agent for more iterations).

Again, Atlassian puts their own security agents to the test before rolling out to customers: remediation time fell from 11.5 days → 6.7 days (42% faster), saving roughly 50 engineering days every week.

Atlassian recently announced "always-on" agents (some in early access, some coming in the next few months):

Again, agents do the work, people own the outcome.

The new flow for engineering teams:

No person named, no merge.

Shoutout to the Jira team for making this breakdown happen.

Every session from the State of AI SDLC summit is streaming on demand, from Mike Cannon-Brookes and Vercel's Guillermo Rauch on why context comes first, to "When Agents Own the Code, Who Owns the Outage?"

→ Learn more about Jira here

The where: 3x high-signal resources to learn more

Atlassian's CEO, Vercel's CEO, Lovable's Head of Growth and the people running 1M agent sessions inside Atlassian, all on 1 stage.

Start with Mike Cannon-Brookes x Guillermo Rauch on why context comes first. Trust me.

Justin Reock is Deputy CTO at DX, the developer intelligence platform that now sits inside Atlassian.

This is his write-up of the whole summit, with the hard numbers:

1.37x more throughput, roughly 28x the AI spend, and why the “innovation ratio” barely moved.

150B+ objects and relationships.

Up to 44% better answer quality.

Up to 48% fewer tokens.

The engine under everything in this piece, and how Atlassian is opening it to any MCP-compatible AI tool.

That's all for today's issue, folks!

If you liked today's edition, you'll love following along on Twitter @tomaldertweets and LinkedIn /in/tom-alder.

Any feedback on today's edition (or the newsletter in general)? Send us a reply.

Thanks for being here.

Our flagship course on how to use free internet data to make better strategic decisions. Contains 5 years of strategy expertise, proven methods, and actionable tactics to accelerate your career with modern-day strategy skills.

We have a growing audience of 150,000+ strategists from top companies like Google, Meta, Atlassian, Stripe, and Netflix. Apply to feature your business in front of Strategy Breakdowns readers.

One of the most common questions we get asked is: “What tools do you use to run Strategy Breakdowns?” So, we’ve open-sourced our tech stack to give you an inside-look at exactly what tools we’re using to power each corner of this operation.