When I started Strategy Breakdowns, I worked at Atlassian.

Strategy and Business Operations (”SBO”) in the Jira integrations team.

It was a dope place to work for lots of reasons. One of the big ones was the hardcore commitment to “dogfooding” - using Atlassian’s tools to build Atlassian.

Every single employee lived inside of Atlassian tools. Every project, line of code, strategy doc happened in the same stack we were actually shipping to customers. Jira, Confluence, Bitbucket, Trello, Loom, etc.

It’s not like Atlassian makes accounting software that only finance could use - the company literally makes dozens of tools for enterprise collaboration, so everyone was a target user of some kind.

I’ve wanted to write this piece for a while.

With all the wild AI coding / agentic collaboration / automated development stuff they’ve been shipping recently (I left a few years ago, right as the first AI features were shipping externally), I had to reach out to the team to get a behind-the-scenes deep dive on the books.

Here’s the latest. Enjoy.

— Tom

A normal night on Ploy. Its agents watch your site and your competitors around the clock, then act: fixing SEO, testing landing pages, spotting the enterprise visitors worth chasing and drafting the outreach - all before standup.

We know because we've been testing it on ours - our new media kit shipped in an afternoon. Built by Webflow co-founder Bryant Chou, it already powers growth at Hex and Clay.

Strategy Breakdowns readers: use code Tom-Alder at checkout.

Every hour spent updating fields is an hour not spent selling. Attio flips the ratio: its revenue agents handle the busywork - account research, follow-up drafts, buyer signals - so your team stays on the deals.

It's the agentic CRM built to run alongside your team, not add to their admin.

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

The what: A TLDR explanation of the strategy

AI was supposed to make work faster.

And it definitely has, at an individual level.

But at a team level? It’s complicated…

In Atlassian's State of Teams 2026 report, they found something surprising.

89% of executives said AI had sped up work.

6% could point to real ROI.

Everyone is sprinting → nobody is syncing.

Atlassian calls the gap the “fragmentation tax”:

It’s duplicated work, misaligned priorities and general coordination chaos from rapid individual AI adoption.

It’s worse now that AI floods teams with more output than they can absorb.

Across the Fortune 500, it’s costing $161 billion a year (that’s ~$400M a day)

Plenty of companies are racing to sell the fix. But Atlassian has a unique advantage that allows them to solve it faster than anyone else: 14,000 employees, inside Jira, Confluence etc every day, pushing AI to its limits.

So instead of building for a customer they have to imagine, they build for themselves:

→ Rebuild Jira, Teamwork Collection and the Teamwork Graph around AI agents to give them the same context and coordination Atlassian has spent 24 years giving humans
→ Run it on the hardest enterprise customer on their books, themselves, until the numbers move
→ Ship the proven tools externally, then push the same pattern past engineering into IT, design, ops and marketing

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

“Dogfooding” is a term that comes from "eating your own dog food".

Basically using your own product.

Do it well and you get the fastest feedback loop there is - your own staff hit the bugs, friction, and missing features before a paying customer ever does.

Plenty of companies dogfood.

But for Atlassian, it’s a way of life.

Atlassian sells enterprise collaboration software, and it’s an enterprise itself.

~14,000 employees added over 24 years since its 2002 founding.

That's 24 years of the exact procurement cycles, security reviews, cross-team dependencies, and legacy workflows its own customers wrestle with every day.

A 40-person startup selling enterprise AI has to imagine what enterprise scale feels like.

Atlassian folks can just walk down the virtual hall.

I witnessed this firsthand when I worked there, and it is like clockwork.

And employees aren’t just the power-users. They're the beta testers too.

Every day, new experiments and updates are tested by thousands of employees.

Research teams run internal interviews.

Data scientists track real retention curves.

Engineers go bug-bashing at scale.

The result: by the time external customers touch new features, they’re already proven.

So how does dogfooding solve the ~$400M a day fragmentation tax?

Well, Atlassian's own engineers live in Jira, Confluence and Bitbucket all day.

But even they had their own version of the fragmentation tax, too.

In a DX study Atlassian cites, engineers' AI usage is up 65%, while developer velocity has risen only around 10%.

The gains leak out at the seams:

Their fix was to turn Jira into the system of record that orchestrates AI agents across the entire SDLC, and not just code.

With humans and agents coordinated in one place, agents assigned work straight from a ticket, and every action visible on the board.

And the cool thing is - they actually document their research in public, publish the real data, and explain how the learnings get folded into the customer-facing products:

Here's what makes AI-native Jira an org-wide step-change strategy rather than an isolated dev-centric shift.

Atlassian knew this fragmentation tax - the gap between individual speed and team output - showed up far outside just engineering:

Engineering was simply where the instrumentation exists to prove it first.

But since all types of Atlassians are power-users of Atlassian products, that same dogfooding loop can run across all their non-engineering teams too:

→ Event Marketing ran the entire planning for Team '25, Atlassian's flagship conference, on Jira and Confluence

→ Creative Studio ran its largest-ever internal Jira campaign, with Rovo pulling work items straight from the project plan

→ Community and Learning Ops unified the work of 6 teams around a single source of truth

Customers are living the same story too.

→ Booksy uses Jira Product Discovery to put product, engineering and business teams on one shared view.

→ Procore, on the same setup plus Rovo, saves 23 hours per month per employee.

→ Tempo builds its own agents in Rovo Studio to automate reporting and finish tasks up to 30% faster, without writing a line of code.

The unit of work changes (a campaign or a creative brief instead of a bug ticket or a pull request).

But the graph underneath remains.

Which means Atlassian doesn't have 1 dogfooding loop. It has one per department, all feeding the same products.

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

Atlassian's Teamwork Lab surveyed 12,035 knowledge workers and 172 executives to put a number on why AI isn't paying off at the team level.

→ Where the $161B fragmentation tax actually comes from

→ The 3 pillars that separate the top teams

→ Why only 14% of teams are getting AI right

Atlassian's own blueprint for what comes after AI code generation.

Plus, plenty of videos demo’ing how these things actually work at scale, live. More show, less tell.

This one’s super cool.

A running collection of Atlassian's own teams showing their real setups, and not just sales materials.

→ Event Marketing, Creative Studio, Community and Learning Ops and more

→ Actual workflows, screenshots and lessons learned

It’s build in public meets enterprise, the closest you'll get to sitting inside Atlassian HQ without the swipe-card.

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.