In a recent study conducted by Atlassian, 89% of executives said AI is accelerating the speed of work.

But they’re struggling to prove it: Just 6% felt confident they could demonstrate AI’s ROI across the business .

Forbes spoke with Avani Solanki Prabhakar, chief people and AI enablement officer at Atlassian , an enterprise software company, to understand why capturing AI ROI remains a challenge — and how organizational change can drive lasting, measurable impact.

This interview has been lightly edited for clarity and length.

Why is it difficult for organizations to measure AI’s value?

Most AI strategies have a team-sized blind spot. Leaders are creating a workforce that has access to AI but doesn’t know how to use it together. AI makes individuals faster, but faster individual work doesn’t automatically mean better business outcomes. In many companies, it creates a coordination problem, pushing more work through systems that weren't designed for this pace.

The goal should be innovation over efficiency. If you only measure speed, usage or tokens, you’re missing the real question: Did the work improve? Did the team make a better decision? Did the customer issue get resolved faster?

At Atlassian, we’ve had to mature our own thinking. We started with adoption because it’s easiest to see. Then we looked at depth of usage and now work outcomes like cycle time, quality and business value. The work, not the user, has to be the unit of measurement, especially as agents start doing more.

What are leaders overlooking as they try to capture AI ROI?

If you start your AI transformation with a goal of efficiency, you've already lost the battle. Stop asking, "Are people using AI?" and start asking, "Is AI improving how they work?”

Another mistake is over-indexing on personal productivity. If everyone is moving faster but the team is still misaligned, you’ve just accelerated the chaos. You’re revving the execution engine while the coordination engine is sputtering. That’s where ROI gets lost.

And finally, measuring token usage is common, but it ignores whether AI is actually embedded in how work gets done and tells you nothing about whether people are innovating.

It’s also important to acknowledge that a healthy AI culture looks different in engineering than it does in legal or HR, so using one metric for everyone can reward volume instead of transformation. We learned this the hard way. Now, we look at usage by craft and study what our "superusers" do differently.

“If you only measure speed, usage or tokens, you’re missing the real question: Did the work improve?” Avani Solanki Prabhakar, Chief People and AI Enablement Officer, Atlassian

What’s at stake when businesses approach AI as a productivity experiment?

They risk mistaking motion for progress. You can have thousands of people using AI and still have no durable change in how the company actually works.

AI is as much a cultural transformation as it is a technological one. The biggest barrier isn’t access to models or tools; it’s mindset, trust, behaviors and confidence. Teams need to know where AI is appropriate and where human judgment needs to stay primary.

If leaders don’t build that culture, AI becomes a side project or a compliance exercise. When AI becomes part of how teams plan, decide and reflect, it stops being an experiment and starts becoming a new way of operating.

Why is AI fundamentally an organizational challenge?

It’s a complete rethink of how work is done. Technology alone won't get you there; you need people and technology priorities moving together.

This is why my role brings people and AI enablement together. If there’s one craft sitting at the center of this transformation, it’s the people function. HR has that enterprise-wide view of capability and culture. If AI sits only with a technical team, you miss the behaviors and norms that make adoption stick.

What shifts when AI is embedded in how teams function?

It looks less like “everyone has a chatbot” and more like the operating rhythm of the team has changed. AI shows up in the rituals teams already run: planning, drafting, onboarding, retros.

You start to see context, workflows and culture changing together. Teams are clearer about what they want AI to do, where human judgment matters and how they’ll review and learn from the output.

A simple example is onboarding. Our People team built NORA (Newlassian Onboarding Rovo Agent) to help managers onboard new hires. It started with a practical question: How do we scale onboarding without burning out our teams? That’s the pattern I care about. Start with a real workflow, ground the agent in trusted knowledge and let the people closest to the work iterate. The best teams are using AI to think better together.

How has your thinking around the HR craft evolved?

As AI changes the cost and availability of intelligence, HR leaders are being handed one of the most consequential mandates: to lead the redesign of how organizations work. I have a few predictions for how we’ll continue to change:

  • HR will become the architect of organizational capacity, moving from managing “headcount” to managing “capacity,” acknowledging the mix of human and agentic capability.
  • Organizational design will revolve around business outcomes versus specific functions.
  • The function will increasingly rely on rich context so AI can operate at its highest capacity, understanding business rules, relationship graphs and situational knowledge.
  • Talent programs will move from fixed jobs to fluid skills and pathways. We’re already seeing roles blurring: Engineers can build prototypes and designers can fix bugs in code. As such, we’ll continue to see roles expanding and evolving.
  • HR isn’t going away — it’s transforming. I believe humans will continue to sit at the center of it, building trust, reading the room and coaching in a way AI simply can’t.

As Atlassian studies its highest-performing AI teams, what patterns do you see?

The strongest teams are disciplined about the problem they’re trying to solve. When we studied commonalities across six of our most AI-advanced teams, we learned:

  • The highest-performing teams frame AI as a tool to solve a real problem, not as something to adopt for its own sake.
  • They treat clarity as a superpower, knowing that AI fails with ambiguity.
  • They mind the 30% quality gap. AI gets you most of the way there, but review, refinement and human judgement are critical.
  • They know roles have changed, and it’s okay to just pitch in.
  • They ran time-bound experiments, identifying specific constraints and setting real deadlines.

Your research also explores AI “superusers.” How do you define them?

Today, we identify superusers by function because each craft — engineering, design, product management, marketing, legal, sales, finance, people — uses AI differently. Each has its own 90th-percentile baseline of weekly AI interactions, and the bar moves with the work.

What matters is how they use it. Superusers are great at creating context and writing precise prompts. They use AI for planning and pressure-testing, not just execution. Most importantly, they build reusable patterns, like agents or templates, that help the whole team. They’re often the team’s connective tissue.

How can leaders identify and foster those superuser behaviors?

Look for the people who are already changing how work gets done. Who’s building agents that others are copying? Who’s sharing what they’ve learned with colleagues?

Then, give them a platform. At Atlassian, we create spaces for employees to share their stories, whether Slack channels or internal events. We’ve found that when leaders demo their own AI use cases, even the messy ones, their teams are more likely to follow suit.

One of our engineering teams piloted a fully AI-driven workflow, shipping four production features with zero manually written code, accelerating their pace by 15x. Their internal blog post detailing the process spread organically and was read by over 1,500 Atlassians, a testament to the culture we’re building. You can't mandate your way to transformation; you have to create curiosity and safe spaces to share and fail.

“When AI becomes part of how teams plan, decide and reflect, it stops being an experiment and starts becoming a new way of operating.” Avani Solanki Prabhakar, Chief People and AI Enablement Officer, Atlassian

Five years from now, what will distinguish companies that successfully scaled AI?

The winners will be the ones that stopped treating AI as a tool rollout and embraced it as a new operating model. They will have built the context layer and redesigned their workflows.

I also think they’ll be better at coordination. AI makes execution cheap, but coordination becomes the bottleneck. The companies that scale will be helping teams stay aligned as the pace accelerates. They’ll move from pockets of experimentation to repeatable patterns across the company.

What organizational foundation must exist to achieve that?

Context is everything. Context means AI understands your goals and trade-offs. Without that, it’s just guessing faster.

Businesses need agent-ready data and a culture of being open by default. To combat AI slop, connect all your apps and tools, build a culture of open data, tie all projects to a goal and treat AI as a consumer of what you know.

You also need guardrails. Responsible AI can’t just be a poster on the wall. Teams need clear guidance on where human judgment must remain primary. Good guardrails don't slow people down; they remove the guesswork so teams can move with confidence.

What are the steps leaders can focus on first?

  • Start small, but with real work. Pick a high-friction workflow, map the steps and ask where AI can take the manual load while humans provide the "taste" and judgment.
  • Get your knowledge house in order. You don’t have to clean up every legacy document, but you do need better norms going forward: Capture decisions where the work lives and make context discoverable.
  • Build fluency through practice. Run workshops on real problems and have leaders demo their own use cases. At Atlassian, we “dogfood” our own technology to truly understand how AI transforms the way we work.

We also just released a new report, Leading with Context: Lessons from Atlassian’s AI Journey , which delves into Atlassian’s own AI transformation, including much of what we’ve discussed today, like what separates teams that see real ROI from those that don't and why we believe context — not models — is the true differentiator in enterprise AI. I’d recommend this report as a resource for anyone looking to start or continue this work.