Assaf Elovic, Head of AI at monday.com – Interview Series

You’ve led AI efforts at some of the most dynamic companies in tech, including Wix and now monday.com—what first drew you personally to the challenge of building intelligent systems?

My journey into AI began during the chatbot boom of 2015. I had an interaction with an AI bot that could truly understand intent; it felt like magic. This wasn’t just a novelty; it was solving real problems like booking appointments and answering complex questions. That moment sparked my curiosity about how these systems worked.

What really drew me in was realizing how accessible AI had become. Some tools and APIs made it possible for developers to build robust applications without needing a PhD in machine learning. The endless possibilities were exciting, and I knew I wanted to contribute to this transformation. Since then, I’ve been dedicated to building AI products that solve real-world challenges and improve people’s lives.

The challenge of building intelligent systems appeals to me because it sits at the intersection of creativity and cutting-edge technology. Every project is like solving a new puzzle; you have to understand not just the technical capabilities but also how people actually work and what they need.

Before joining monday.com, you created open-source tools like GPT Researcher that resonated with developers and researchers alike. How have those grassroots, community-driven experiences shaped your approach to building enterprise AI products today?

The open-source experience taught me invaluable lessons about building for real user needs rather than theoretical ones. When you’re building in the open, you get immediate, unfiltered feedback from developers who are actually trying to solve problems. This taught me to focus on practical utility over impressive demos.

Working with the community also reinforced the importance of making AI accessible. Many of the developers using these tools weren’t AI specialists—they were building applications and needed AI capabilities that were reliable and easy to integrate. This experience directly influences how we approach AI Blocks at monday.com: making powerful AI capabilities available to non-technical customers through intuitive interfaces.

Earlier this year, monday.com recently unveiled a bold new AI vision with three pillars: AI Blocks, Product Power-ups, and a Digital Workforce. How did this framework come together, and what gap in the market are you trying to fill?

Our AI vision emerged from observing a fundamental challenge: organizations of all sizes want to leverage AI, but most solutions require significant technical expertise or are too rigid for diverse business needs. We saw that people weren’t just looking for another AI assistant; they needed AI that could seamlessly integrate into their existing workflows and adapt to their specific processes. Lastly, we are now focusing on helping people get work done with AI, a shift from helping people manage work.

The gap we’re filling is the space between simple AI tools and complex enterprise solutions. Many businesses fall into a middle ground where they need more than basic automation but can’t justify or implement heavyweight AI systems. Our three-pillar approach gives organizations the flexibility to start simple with blocks, enhance their products with power-ups, and eventually build sophisticated digital workforces.

Since the launch we’ve been pushing strong across all verticals with significant growth in adoption and paying users.

We’ve also introduced “vibe coding” products that aim at our mission to democratize software. With the latest advancements in AI it’s never been easier to build full applications with simple natural language. Our latest products like monday vibe and magic can enable any non technical user to leverage the Monday ecosystem to build customized applications for work.

Can you walk us through how AI Blocks work in practice? What’s the learning curve for non-technical users trying to integrate these tools into their daily workflows?

AI Blocks are designed to be as intuitive as building blocks—hence the name. In practice, a user might drag an “extract deadlines” block into their project management workflow, or add a “summarize meeting notes” block to their weekly review process. The blocks handle the AI complexity behind the scenes while presenting customers with simple, familiar interfaces.

The learning curve is intentionally minimal. We’ve seen teams successfully implement AI Blocks within their first session. For example, a marketing team might create a workflow where social media mentions are automatically analyzed for sentiment and key themes are extracted, all without writing a single line of code.

The key insight is that people don’t need to understand how AI works to benefit from it. They just need to understand their own processes well enough to identify where automation would help. We’ve designed the blocks to match the mental models people already have about their workflows.

You’ve recently launched a suite of AI-powered tools including monday magic, monday vibe, and monday sidekick. What makes these products different from traditional assistants or copilots, and what role do you envision them playing across industries?

Our latest releases represent a comprehensive approach to workplace AI that goes beyond traditional assistants. Each capability serves a distinct purpose while working together as an integrated ecosystem that fundamentally transforms how teams operate, solidifying our shift from work management to work execution for our customers.

monday magic brings intelligent automation to workflows, using AI to predict needs and automate complex processes before users even realize they need them. monday vibe is a vibe coding platform that enables anyone to build secure, custom business apps tailored to their team’s exact needs. And monday sidekick serves as your contextual AI companion, understanding your specific work patterns and providing proactive assistance tailored to your role and responsibilities.

Together, these capabilities move our customers beyond simply managing and tracking work to actually executing it more intelligently. Instead of just organizing tasks and monitoring progress, teams can now rely on AI to optimize performance, anticipate challenges, and take action automatically. This shift from passive management to active execution is transformative; it means less time spent on administrative overhead and more time focused on high-value work that drives results.

What makes these different from traditional assistants is their deep integration with actual work context and their focus on proactive rather than reactive support. While most AI assistants wait for you to ask questions, our suite observes patterns, anticipates needs, and takes action within your established workflows and permissions.

monday.com emphasizes explainability and user experience, not just raw model performance. What does that look like behind the scenes, and how do you balance transparency with power?

Explainability is fundamental to building trust, especially in enterprise environments where decisions have real consequences. Behind the scenes, we invest heavily in making our AI’s reasoning transparent. When our Risk Analyzer flags a potential project delay, it doesn’t just raise an alert; it shows exactly which factors contributed to that assessment and how confident it is in the prediction.

This focus came from experience. Early AI systems often felt like black boxes, which made customers hesitant to rely on them for important decisions. We learned that customers need to understand not just what the AI is suggesting, but why it’s making that suggestion.

The balance between transparency and power comes down to layered disclosure. We provide immediate, actionable insights at the surface level, but customers can drill down to see the detailed reasoning when they need it. This approach builds confidence while maintaining usability—customers trust the system more when they understand it, which paradoxically makes them more willing to leverage its full capabilities.

With over 46 million AI actions now performed on the platform, what are some of the most surprising or creative ways customers have used AI?

The creativity of our customers constantly amazes me. We’ve seen a wedding planner use AI Blocks to automatically categorize vendor responses and extract key details like pricing and availability dates. A research team created a workflow that analyzes academic papers and automatically populates a database with key findings and methodology notes.

One particularly creative use case was a restaurant chain that used our AI to analyze customer feedback across locations and automatically flag potential food safety concerns by detecting patterns in complaints. They essentially created an early warning system for operational issues.

What’s surprising is how customers combine simple blocks in sophisticated ways. They’re not just automating individual tasks; they’re redesigning entire processes around AI capabilities we never explicitly designed for their specific use cases.

You also serve as a scout for Sequoia Capital, investing in early-stage AI startups. From that vantage point, what common mistakes do founders make when building AI-first products?

The most common mistake I see is founders getting seduced by the technical possibilities of AI without deeply understanding the user’s actual workflow and pain points. They build impressive demos that showcase AI capabilities but fail to solve real problems in the way people actually work.

Another frequent issue is over-promising on AI autonomy too early. Many founders want to build fully autonomous systems when customers actually need collaborative tools. People want AI to augment their capabilities, not replace their judgment, especially in high-stakes business decisions.

There’s also a tendency to underestimate the importance of trust and explainability. Founders often focus on accuracy metrics but neglect the user experience in handling uncertainty and errors. In enterprise contexts, especially, customers need to understand when and why to trust AI recommendations.

Finally, many AI-first startups struggle with distribution. Having great AI technology isn’t enough; you need to understand how to integrate it into existing workflows and demonstrate clear ROI to decision-makers who may be skeptical of AI hype.

How do you think AI agents will evolve over the next few years—will they become more autonomous, more specialized, or something else entirely?

We’ll see AI agents evolve toward contextual collaboration rather than pure autonomy. The future isn’t fully autonomous agents making independent decisions, but rather agents that deeply understand context and can take appropriate levels of action based on the situation and user preferences.

We’re moving toward agents that can handle routine decisions autonomously while seamlessly escalating complex or ambiguous situations to humans. This requires a sophisticated understanding of context, risk assessment, and user intent, capabilities that are rapidly improving.

I also expect significant evolution in multi-agent coordination. Instead of monolithic AI assistants, we’ll see specialized agents that collaborate with each other and with humans in dynamic teams. Your research agent might work with your scheduling agent and your communication agent to coordinate a complex project launch.

The key evolution will be in the human-AI interface. Agents will become better at communicating their reasoning, expressing uncertainty, and adapting to individual working styles. The goal is seamless collaboration where the boundaries between human and AI contributions become less important than the collective outcome.

Internally, how do you structure collaboration between your AI, product, design, and GTM teams to ensure AI is seamlessly embedded in the user experience?

Successful AI product development requires breaking down traditional silos and creating shared understanding across all teams. We’ve found that the key is establishing a common language around AI capabilities and limitations that everyone can use, from engineers to designers to marketers.

Our process starts with cross-functional discovery sessions where we explore user problems together before discussing technical solutions. This prevents the common mistake of leading with AI capabilities and then looking for problems to solve.

We also invest heavily in prototyping and user testing throughout the development process. Design and product teams work closely with AI engineers to understand what’s possible, while AI teams learn about real user constraints and preferences. This bi-directional learning is crucial for creating AI features that feel natural rather than bolted-on.

From a GTM perspective, our teams are embedded in the development process from day one. They help us understand not just what customers want, but how they think about AI, what concerns they have, and how they prefer to learn about new capabilities. This insight directly influences both the product design and the technical implementation.

Finally, as someone who bridges open source, enterprise AI, and VC, where do you think the next big AI breakthrough will happen—in tools, infrastructure, or something we’re not even looking at yet?

The next breakthrough will likely happen at the intersection of human-AI collaboration interfaces. We’ve made incredible progress in model capabilities, but we’re still in the early stages of figuring out how humans and AI systems can work together most effectively.

The breakthrough won’t be in making AI more autonomous, but in making human-AI collaboration more fluid and natural. This includes advances in how AI systems communicate uncertainty, how they adapt to individual working styles, and how they coordinate with multiple humans and other AI systems simultaneously.

From an infrastructure perspective, I’m watching developments in real-time, contextual AI that can understand and act on dynamic information streams. The ability to build AI systems that maintain context across long time horizons and multiple interaction types will enable entirely new categories of applications.

But honestly, the most exciting breakthroughs might come from unexpected directions. Just as transformers emerged from attention mechanisms in neural machine translation, the next significant advance might come from solving a seemingly narrow problem that has broad applications. The key is maintaining that beginner’s mindset and staying open to possibilities we haven’t imagined yet.

Thank you for the great interview, readers who wish to learn more should visit monday.com.<