The AI Contradiction Sitting in the Middle of Your Rollout

94% of knowledge workers are using AI at work. The ones who admit it, though, are being rated 10 times lazier than peers doing the same work.

That finding is not from a skeptic. It’s from Atlassian’s own Teamwork Lab, published this week. Conor Donegan, their Senior Principal Researcher, ran a controlled experiment with 961 U.S. workers in March and April of this year. Participants evaluated identical written work. The only variable was whether they were told the person used AI. When AI use was disclosed, evaluators rated the creator as significantly lazier, less principled, and 24 percentage points less likely to be recommended for a high-visibility project.

Atlassian, the company that staked Team ’26 on AI-native teamwork and launched Rovo Studio to general availability last month, published research showing the cultural conditions for AI adoption are broken in most organizations. That took honesty. It’s worth engaging with seriously.

---

The Disclosure Penalty Is Real, and the Fix Is Only Partial

Atlassian’s Teamwork Lab head Molly Sands named it clearly: “Companies are telling the workforce to use AI, but employees are penalizing each other for being honest about it. Until leaders change the culture, you don’t have an AI strategy; you have an AI contradiction.”

And “an AI contradiction” is exactly how I’d put it too.

The research also surfaced what to do about it. Workers who framed their AI use as team-focused (“I used this to support the client” rather than “I used this to save myself time”) were rated 11 points higher on effort and 8 points more likely to be recommended. That’s meaningful. But here’s the part Atlassian buries in the methodology: both groups still ranked well below workers who didn’t mention AI at all. The honest ones, even the strategically honest ones, still paid a penalty. Silence remains the rational choice.

Let that sink in: the workaround doesn’t even work. At least not entirely.

Think about what that actually means as a design problem. The research isn’t offering workers a way to be transparent without penalty. It’s offering a way to be penalized somewhat less. Team-framed disclosure gets you to a 47% recommendation rate for high-visibility projects. Silence gets you to 71%. That’s not parity. That’s still a 24-point gap from the population that said nothing.

The rational actor in this system doesn’t disclose AI use. That’s not cynicism, it’s just math. And it means the 94% usage figure at the top of this post is almost certainly an undercount. That survey asked how many workers use AI. It didn’t account for how many workers are willing to tell a researcher they use AI when they’ve already learned it carries professional cost. The real usage number is probably higher, and the honest disclosure rate is probably lower, than any voluntary survey can capture.

---

Zoom Out, and the Story Is Anxiety

The broader data isn’t more comforting. Mercer surveyed 12,000 workers globally for their 2026 talent trends report and found job loss fears jumped from 28% to 40% in two years. Pew Research found 52% of workers worried about AI in the workplace, against only 36% hopeful. A separate study found 39% of workers say heavy AI use has weakened their own skills, a number that rises to 46% among Gen Z workers.

These are directional findings, not projectable truths. The populations and methodologies differ. But the direction is consistent: at the worker level, the story of AI at work is anxiety, not enthusiasm. The enthusiasm is concentrated at the leadership layer, in keynotes and earnings calls and conference announcements.

Sands said it plainly: “In my world working in tech, if you’re not using AI at all, you’re the one who has to explain yourself. But these findings make clear that my experience is the exception.”

We’ve Been Here Before

This pattern has shown up before. Calculators were once considered cheating. Spell-checkers were a crutch. Spreadsheets were for people who couldn’t do math in their head. Each tool went through a phase where using it marked you as someone who couldn’t do the real work, before the inversion happened and not using it marked you instead. The cultural penalty was real, for a while, and then it wasn’t.

What’s different about AI is the speed. Previous tools took years, sometimes decades, to move from suspicious shortcut to table stakes. AI tools went from experimental to enterprise mandate in roughly two years. The cultural norms haven’t had time to catch up. That gap is what this research is measuring. And unlike the spreadsheet, whose rollout was largely self-paced, the AI rollout is being mandated from the top down while the cultural adjustment is still happening from the bottom up.

---

The Practitioner’s Version of This Problem

For any successful adoption, you need three things in alignment. People, Processes, and Tools. If any one of them is out of sync, the adoption is doomed to fail. I often reach for the fire triangle. For a fire to sustain itself, it needs fuel, oxygen, and heat. Take any one of those away, and the fire dies. Keep that in mind as you look at what is happening here.

Imagine you’re the person being handed the Rovo rollout in your organization. Your executive came back from Team ’26, or read the product announcements, or got the cost-justification slide from sales. They’re excited. The tools and use case look good. Rovo’s agent capabilities are real. And now it’s your job to actually get people using them.

The data says you’re rolling into a cultural environment where a significant portion of your users, if they use AI and say so, will be judged negatively by their peers. Where the rational response to “be transparent about AI use” is to do anything but that. Where the org wants the productivity gains but hasn’t done the work of making it socially safe to pursue them.

That’s not a Jira problem. It’s not a Rovo problem. It’s not something Atlassian can ship a fix for. It’s an organizational culture problem that gets handed to the admin (who is already managing the technical rollout, the governance questions, the permission models, and the user training) without being named as such.

Back to the triangle. The tools are real. The process is moving. It’s the people leg that’s missing, and just like the fire, it doesn’t matter how good the other two are once you pull one away.

There are things you can actually influence here, even if you can’t fix the culture. Frame your rollout announcement around team productivity, not individual efficiency. Surface wins in your internal comms as team outcomes, not time-saved-per-user. Advocate internally for leadership to visibly model AI use before you ask individual contributors to declare theirs. None of that eliminates the penalty. But it shapes the environment the tools land in. And it gives you something more useful to say to your executive than “the tools are ready but the culture isn’t” — which is true, and which they will not want to hear, and which they need to hear anyway.

---

The Real Fix Isn’t a Configuration Setting

The Atlassian research does point at the actual fix: companies where leadership visibly models AI use, where wins get highlighted, where the norm shifts from “AI as shortcut” to “AI as shared tool.” In those environments, the laziness penalty nearly disappears.

But that’s a leadership behavior change, not a configuration setting. And it requires someone to tell leadership that the culture problem exists before they can fix it. That’s a harder conversation than opening a Rovo agent.

Give Atlassian Credit for Publishing This

Let me put this plainly. Atlassian is a public company whose product strategy is built on AI adoption going well. They have spent the last two years constructing a narrative around AI-native teamwork: Rovo, Teamwork Graph, AI-powered automation across their suite, the entire arc of Team ’26. Every keynote, every product announcement, every customer conversation has pointed in the same direction. AI tools are the future of work, and Atlassian is how your organization gets there.

And then their own research lab published a study showing that the cultural conditions for that future aren’t present in most organizations. That the people being asked to adopt the tools are quietly penalized when they do. That silence is still the rational choice in the environments these tools are supposed to thrive in.

That’s not a failure of research integrity. That’s intellectual honesty from a company with every incentive to tell a cleaner story. They could have commissioned research that supported the narrative. They published the thing that complicates it instead.

---

Wrapping Up

Step back and the arc is clean. Almost everyone is using AI at work. The ones who are honest about it get rated lazier, less principled, and less promotable for doing identical work. The only fix on offer gets you penalized slightly less, never to parity, so silence stays the rational play. And the person who inherits that contradiction isn’t the executive who got excited at Team ’26. It’s the admin handed the Rovo rollout, told to drive adoption into a culture that quietly punishes the exact behavior the rollout depends on.

I’ll be honest about my own bias here. This research didn’t make me uncomfortable. It confirmed things I’ve watched play out across my own career, and that’s exactly when I try to be most careful. Information that lines up neatly with what you already believe deserves more scrutiny than information that challenges it, not less, because it’s the stuff you’re primed to wave through without checking. So the question I had to sit with wasn’t whether this confirms what I think. It was whether I’m just reading what I want to see. What holds it up is the source: a finding this inconvenient, published by the company least served by it, isn’t the kind of confirmation you go looking for.

None of that changes the fact that the tools are genuinely capable, or that a Rovo rollout is still the right strategic move for most organizations. But it does mean the rollout is harder than the announcement layer suggests. If you’re the admin managing it, you’re not just implementing software. You’re navigating a cultural contradiction your organization hasn’t acknowledged yet.

If the company with the most to gain from a different answer is telling you this, the real picture in most organizations is probably starker than the study captures. The technical rollout is the easy half. The hard half is the conversation with leadership about the culture these tools are landing in, and that’s the part I’ll keep coming back to here, because it’s only going to matter more as Rovo spreads across your instance.

So if you’re planning a Rovo rollout, read the study. Not for the fix. For the diagnosis.

Until then, this is Rodney, asking: have you updated your Jira issues work items today?