Andrey Khusid cofounded Miro, a provider of visual workspaces used by over 100 million people to collaborate on distributed teams, in 2011. In a conversation with McKinsey Senior Partner Martin Harrysson and Associate Partner Ninad Kulkarni, Andrey shares how he grew the company from an online whiteboard into an AI-powered workflow platform used by more than 250,000 global companies. He discusses how AI is transforming developers’ roles, reshaping product road maps, and shifting enterprise operating models.
This interview has been edited for length and clarity.
Martin Harrysson: Could you share a bit about your background and what led you to found Miro?
Andrey Khusid: We started in 2011 with an online whiteboard called RealtimeBoard. Early on, the most popular use cases were in user experience design and agile product development—teams collaborating visually to align on journeys and concepts and to facilitate group work.
Adoption was product-led—people would sign up, start using it, and share it with others. During the pandemic, that adoption scaled across enterprises into a broader set of use cases such as workshops, process design, and creative problem-solving.
As usage scaled, we started to see a clear pattern in how teams worked. Product managers were managing backlogs in one place, designers were wireframing in another, and engineers were doing architecture design elsewhere. We saw the opportunity to connect those workflows end to end on a shared canvas for full teams to collaborate on.
Now, with AI, that connection and collaboration across teams becomes much more powerful. You can move from one step to another with a click while keeping humans in the loop—who are leading and refining rather than creating everything from scratch.
Martin Harrysson: You’ve emphasized the importance of cross-functional collaboration, and now AI is changing how work gets done. How is AI reshaping both technical and nontechnical roles?
Andrey Khusid: I’m observing a shift toward what I would call “makers.” Previously, roles were clearly defined—product managers, designers, engineers—but now, with AI, individuals can start to create, execute on it, and ship it end to end.
Previously, roles were clearly defined—product managers, designers, engineers—but now, with AI, individuals can start to create, execute on it, and ship it end to end.
This personal agency doesn’t eliminate collaboration, but it changes it. In smaller setups, makers can operate more independently, while still needing a way to keep track of the bigger picture. In larger organizations, where there are more dependencies, coordination continues as before—just with the expectation of higher speed and innovation.
At the same time, we’re seeing the rise of agentic workflows. Instead of coordinating only human to human, you now coordinate human to agent and increasingly agent to agent. That adds complexity. You’re managing multiple agents—design, product discovery, code generation—running in parallel. So the coordination doesn’t go away; it expands into multiple dimensions.
That’s why today at Miro, we’re focused on collaborative human–agentic workflows. At the end of the day, the role of humans is to make decisions and take accountability for them, so you need a place where those decisions happen and where that accountability lives. For us, that means creating a single pane of glass where everything comes together—people, workflows, and agents working in a single connected environment.
Ninad Kulkarni: You mentioned connecting workflows and coordinating across roles and agents. How does Miro use its own platform internally to bring product, engineering, and broader teams together?
Andrey Khusid: Internally, we see fewer handoffs between people and more focus on building a shared understanding as work progresses.
I will give you a real example. A product manager who joined recently used Miro Insights, our opportunity-scoping tool, to pull together customer requests and conversations, map them into a data table on a canvas, create a backlog, and prioritize it live for a discussion. Previously, getting that level of input would have meant reaching out across sales and go-to-market teams, waiting for inputs, and then following up—often taking several weeks to come together. This product manager did it in a few days, with almost no historical context of the product.
Another example is how teams move between Miro and AI tools. One product manager started by sketching out an idea on a canvas—using screenshots, notes, and a few sticky elements—and used that as context for a code generation tool. From there, work moved in and out of specific task-based tools, with the canvas as the place where teams came back together to align and make decisions.
Ninad Kulkarni: It sounds like what used to be a fragmented feedback loop—inputs coming through sales, support, and other channels—is becoming streamlined and faster. How is that changing how you think about what to build?
Andrey Khusid: It’s evolving as we speak. The way software is built is changing, and it’s becoming more feasible to go deeper on specific customer needs because you can deliver changes faster.
The way software is built is changing, and it’s becoming more feasible to go deeper on specific customer needs because you can deliver changes faster.
So the question becomes less about building a broad set of features and more about how well you solve a particular problem for a particular customer. If you go the last mile and really solve what the customer needs, that’s where the value is.
Martin Harrysson: Building on that, as feedback loops get faster and teams move more quickly, how are you measuring engineering productivity—and, more broadly, how do you define success in product development?
Andrey Khusid: We look closely at the velocity of value going into production, and we’re seeing that increase significantly. In parallel, we’re redesigning toward an AI-first delivery model—not where AI amplifies output by 10 to 20 percent, but where it delivers multiple-X returns, with AI leading and humans in the loop.
What matters most is whether we move the needle with every investment. That has always been true, but the environment is now much more crowded. Teams can ship features almost overnight. So it’s not about building more features but identifying the right bets and configurations that actually drive the business.
Of course, that’s a much more complex problem to solve. It requires a clear point of view and the ability to digest signals in real time. Ultimately, you need to tie everything back to impact—what is actually moving the business forward.
In a larger organization, many teams own different parts, and each can optimize for its own area. But those local optimizations haven’t always added up to the best outcomes. Before AI, a company could run many local optimizations and still succeed. Now, with the level of competition and how fast things move, it’s more important to focus on a small number of bets that drive the biggest impact.
Martin Harrysson: You’ve mentioned the need for more real-time signals and a clearer connection to impact. How can companies evolve their operating models in response?
Andrey Khusid: My view is that companies that don’t adopt an AI-first operating system will face significant challenges. By AI-first, I mean making AI more proactive—moving from asking questions to AI proactively surfacing insights, patterns, and recommended actions.
When you do this, instead of navigating layers of data and people, you get direct access to signals that matter and can tie them to impact. That changes how companies operate. Humans can focus on decisions and accountability, while AI provides continuous context and recommendations.
Humans can focus on decisions and accountability, while AI provides continuous context and recommendations.
The challenge, of course, is building and scaling that system. A company will need to take a more centralized approach to collaboration. The goal is to empower teams in ways that help them think and act, supported by the global context of what matters, the right data, and agents working together.
Companies, including Miro, are still early on this journey. In most organizations, AI deployments are still in the “all flowers bloom” moment but are not yet coherent systems. The companies that get there first will move faster to completely redefine how work gets done.
