Did AI take your job or did your product suck?

Atlassian laid off 1,600 people this week. That's 10% of the company. 900 of them were in R&D. The CTO is gone. They're burning $230 million on severance. The official explanation from co-founder Mike Cannon-Brookes is that this is about "rebalancing for the AI era," changing skill sets, and freeing up capital to invest in AI and enterprise sales.

The narrative writes itself. AI is eating software jobs. Even the companies that build developer tools aren't safe. Inevitable, tragic, move along.

I don't think that's the real story. I think the layoffs are a symptom of something Atlassian has been struggling with for years, and AI just made it impossible to ignore: Jira, their core product, is losing its hold on the market. Not because a competitor beat them. Because the product became something nobody actually enjoys using, and for the first time in its history, switching away from it is becoming easy.

The product nobody likes but everyone uses

If you've worked in software, you have opinions about Jira. Probably strong ones. And if you spend any time on Reddit, Hacker News, or the Atlassian Community forums, you'll notice something remarkable: the people who use Jira every day overwhelmingly do not like it.

The complaints are consistent enough to be boring at this point. It's slow. The UI is cluttered. The search is bad. Workflows have too many steps. There are too many mandatory fields. Everything takes one more click than it should. And the customizability, which is supposed to be Jira's strength, routinely becomes its worst feature because every company's Jira instance ends up as a unique snowflake of pain configured by someone who left three years ago.

This isn't just vibes. The 2025 Stack Overflow Developer Survey showed GitHub overtaking Jira as the most desired developer collaboration tool. On Atlassian's own community forum, users called the May 2025 navigation redesign a "colossal failure on every level." The mobile app is so bad that users prefer opening the desktop site on their phones.

And yet, Jira still has over 40% of the project management software market. How? The same way your company still uses that terrifying Excel spreadsheet from 2014: switching costs. Enterprise inertia. Three years of accumulated workflows, integrations, and audit trails that nobody wants to rebuild.

Jira isn't dominant because it's good. It's dominant because leaving is expensive.

How Jira became a bloated mess

There's a story in product development that plays out over and over. A tool starts simple and good at one thing. It gets popular. Customers ask for more features. The company adds them. Each feature makes sense individually. Together, over time, they create something that no single user actually needs.

Jira is the textbook case.

It started as a bug tracker. Then it became the agile board. Then the service management platform. Then the product discovery tool. Then the portfolio tracker. Each expansion made sense in a roadmap review. Each one brought in new customer segments. And each one added complexity that the existing users had to wade through.

This is what people in product circles call the Product Trap: optimizing for feature count and breadth of use cases instead of for the quality of any single experience. Atlassian had a window, probably around 2018-2020, to step back and fundamentally rebuild Jira's experience from scratch. To rethink how the product should deliver value rather than just bolting more value on top. They could have made it fast. They could have made it clean. They could have made it something a VP would voluntarily open on a Monday morning.

They didn't. They kept shipping features.

The result is a product that tries to serve everyone. Engineering teams only need a fraction of what Jira offers. Product managers need a different fraction. Leaders need yet another fraction, and most of them refuse to use it at all. The people who get the most out of Jira are Jira administrators, people whose entire job is making Jira work. That's a very specific audience to anchor a $5 billion business on.

The moat that isn't a moat anymore

For most of Jira's dominance, none of this mattered much. Even if you hated the product, what were you going to do about it?

Building a custom project management tool was a serious engineering effort. Six months minimum, probably more. You'd need a backend, a frontend, database design, auth, permissions, integrations with GitHub or Bitbucket, maybe a mobile app. Nobody was going to burn half a year of engineering capacity just to avoid Jira's clunky UI. It wasn't worth it.

So instead, you bought a Marketplace addon. Needed better reporting? There's a plugin. Wanted OKR tracking? Plugin. Custom dashboards? Plugin. This is exactly how Atlassian's ecosystem grew. The core product was mediocre at many things, but the Marketplace filled the gaps. At its peak, the Marketplace had over 6,000 apps. For Atlassian, it was a beautiful flywheel: more apps meant more lock-in, which meant more customers, which attracted more app developers.

That flywheel is slowing down, and the reason is something Atlassian probably didn't see coming.

Agentic AI changed the math

Two years ago, if your team wanted a custom project tracker with Kanban boards, GitHub integration, automated sprint workflows, and a clean dashboard, you were looking at a real engineering project. Multiple developers, multiple sprints, significant opportunity cost.

Today, a competent engineer with agentic coding tools can build that in days. Not a sketch. Not a prototype that falls apart. A real, functional internal tool, tailored to how the team actually works. No configuration screens. No unused features. No 14-week rollout plan. Just the tool your team needs, the way they need it.

This doesn't replace enterprise Jira for a 5,000-person organization with compliance requirements and cross-functional reporting needs. But it does replace Jira for the thousands of dev teams that were only using 15% of it and hating the rest.

And it absolutely replaces the Marketplace addon model. When buying a third-party Jira plugin for dashboards or process management was the cheapest option, people did it. Now that building your own version is faster and cheaper than evaluating, procuring, and configuring a Marketplace app, why would you?

The numbers hint at this. Atlassian's Marketplace revenue grew 6% last year. Their cloud revenue grew 26%. When the ecosystem around your product is growing at one-quarter the rate of the product itself, it means customers are finding other ways to solve the problems the ecosystem used to solve for them.

What the layoffs really mean

So let's come back to the 1,600 people who lost their jobs this week.

The easy story is that AI replaced Atlassian's programmers. Claude writes code now, so you need fewer engineers. And sure, that dynamic exists at every software company to some degree. AI coding tools are genuinely changing how much output a smaller team can produce. That's real.

But I don't think that's mainly what's happening here.

Atlassian's revenue growth has been decelerating. The full-year FY2024 growth was 23%. FY2025 came in around 18%. The company's stock has been hammered by investor fears about AI disruption in the software tools market. Not fears about AI replacing Atlassian's employees. Fears about AI making Atlassian's products less necessary.

Cannon-Brookes himself said this isn't about replacing people with AI. He said it's about "rebalancing." I actually believe him on the literal point, but I think the rebalancing is a response to a product and market position that's weaker than Atlassian wants to admit publicly.

When your CTO leaves as part of a restructuring, that's not a cost optimization. That's a strategic reset. They're bringing in what they're calling "next generation AI talent" as replacements. That tells you where they think the existential threat is.

The uncomfortable truth

Atlassian's problem isn't unique. It's the problem every enterprise software incumbent is about to face, but Atlassian is one of the first to feel it acutely because developer tools sit right in the blast radius of AI-enabled productivity.

The pattern goes like this. You build a dominant product. You grow it by adding features. The features create complexity. The complexity creates switching costs. The switching costs become your moat. For years, that moat holds. Then something comes along that lowers the cost of building alternatives from "too expensive to bother" to "we can do this next week." And suddenly your moat, which was never product quality but customer inertia, starts to drain.

That's where Atlassian is right now. Not dying. Still a $5 billion revenue company. Still growing. But growing slower, with a product that repels its core users, an ecosystem that's decelerating faster than the core, and a competitive landscape where "just build your own" is a viable answer for the first time.

Those 1,600 people didn't lose their jobs because Claude learned to write code. They lost their jobs because AI made it easy enough for Atlassian's customers to stop needing Atlassian.

That's a much harder problem to restructure your way out of.