When most companies talk about AI enablement, they mean rolling out a few tools and hoping developers figure it out. Sergei Liakhovetsky, VP of R&D at monday.com, took a fundamentally different approach. He paused the entire roadmap for 30 days and pointed all 700 technologists at a single goal: become an AI-enabled organization.
The results speak for themselves. Monday Magic launched with 5,000 solutions built in under three months. Monday Vibe generated 40,000 apps in two months. Sidekick logged 150,000 interactions in less than a quarter.
Under the hood, the impact was just as profound. The team accelerated one of their biggest technical debt problems, cutting a 33-month investment down to just five months of work. Customer ticket resolution time dropped from three days to one. Test coverage doubled, and onboarding speed increased by 21%.
These outcomes didn't happen by accident. They emerged from a deliberate strategy that balanced foundational infrastructure work with rapid experimentation and a willingness to make bold organizational bets at exactly the right moment.
Converting 700 engineers into AI shippers
The decision to pause the roadmap for a full AI enablement month didn't happen casually. The company had been building AI features incrementally, treating them as table stakes rather than transformation. Liakhovetsky and his counterpart in product realized that with 700 engineers across the organization, incremental adoption was not going to create the cultural shift they needed.
They needed to engage, inspire, and provide the tools for hands-on experimentation. But engagement alone was not the goal. The team established clear principles from the start.
First, there would be no hackathon-style output. Everything built during the month needed to be production-ready. The goal was to teach teams how to operationalize AI, not just prototype it. Liakhovetsky was clear that they were not going to run it like a hackathon because they wanted to ensure whatever they built actually reached production.
Second, business purpose came first. Work needed to align with the product roadmap and deliver customer value, with exceptions treated as rare. Finally, they had to preserve customer commitments. The leadership team spent two weeks working with individual squads to understand their existing commitments and determine what could still be delivered, often utilizing AI assistance to get it done.
The execution model combined broad activation with hands-on learning. The team ran 17 workshops with 22 speakers and showcased roughly 70 demos during the month. They established a champions program to create peer-to-peer learning channels and ensured that every team understood how to operationalize AI in production.
Cutting tool approvals to one week
One of the most striking aspects of monday.com's AI enablement was their approach to tooling. Rather than standardizing on a single AI coding assistant, they opened the floodgates.
The team started with Cursor, but quickly realized that tool restrictions were creativity restrictions. Limiting people, Liakhovetsky noted, is limiting their imagination and their ability to move fast. So he worked directly with security, procurement, and legal to design a zero-bureaucracy approval process targeting a one-week turnaround from request to access.
Zero bureaucracy didn't mean zero accountability. The system relied on champion ownership. Anyone requesting a new tool had to become its champion, taking responsibility for adoption, internal enablement, and basic risk checks like minimizing PII exposure.
This model created natural selection pressure. Champions who genuinely believed in a tool would invest in adoption and evangelism because they wanted to keep using it. To enforce this, any tool that failed to demonstrate sustained usage after two months was simply removed from the catalog.
This approach turned tool selection into peer-driven change management. People felt like owners and wanted to make sure others used the tools, leading to organic communication and sharing across teams. Critically, the application security team worked side-by-side with development as real builders, ensuring that rapid experimentation remained safe experimentation.
The invisible foundation that made it possible
While the 30-day pause generated the features, none of those features would have survived in production without the massive infrastructural overhaul monday.com completed beforehand. Before they could move fast on AI, they had to build the infrastructure to move fast at all.
As they moved upmarket into enterprise accounts, the team recognized that their existing SQL-based architecture couldn't handle the demands. The system originally dealt with boards containing thousands of items, but it now needed to maintain millions of items in a single entity.
This led to a two-and-a-half-year foundational rebuild called MondayDB. The team migrated from SQL to Cassandra, added in-memory databases like DynamoDB, and fundamentally reimagined how the platform handled scale.
Liakhovetsky views this foundational work as the equivalent of building highways. Building highways takes a significant amount of time, but once you have them, you have unlimited speed later on.
For enterprise customers, trust is built on multiple dimensions. It is not only about uptime; it is also about performance. In the AI era, when customers expect instant responses, high latency feels exactly like an outage. Monday.com operationalizes this trust by tracking hard metrics like core-flow performance alongside soft metrics like customer sentiment, allowing them to catch trust erosion before it shows up in churn numbers.
Stopping noisy neighbors with cell architecture
After modernizing the data layer, monday.com tackled blast-radius reduction through cell architecture. In multi-tenant SaaS systems, a single high-volume customer can degrade performance for everyone else, a phenomenon known as the "noisy neighbor" problem.
Cell architecture addresses this by isolating customer workloads into separate failure domains. If one cell experiences an issue, the others continue operating normally.
AI amplifies the noisy neighbor problem in completely new ways. With increased automated traffic and new interaction patterns from AI agents, the impact of misbehaving workloads becomes much more severe without strong segmentation.
This cell architecture also laid the groundwork for AI-era security requirements. As AI usage increases across enterprise tenants, proper account isolation becomes a security imperative. Cells provide the foundation for enforcing guardrails, ensuring that one customer's AI agents cannot access another customer's data or consume shared resources unfairly.
The power of intentionality
The story of monday.com's AI transformation reveals that speed from decision to action preserves momentum. The two-week window from the decision to the kickoff of the AI month was strategic. When trying to create a cultural shift, process-building becomes the enemy of progress. You must move fast, learn through execution, and fine-tune as you go.
Monday.com did not stumble into AI capability. They designed for it, starting from the foundational infrastructure work that preceded the AI month, to the careful balance of constraints and freedom during it, to the production-first culture that ensured learning translated into durable capability.
As AI becomes table stakes for software companies, the winners won't be those who ship AI features fastest. They will be the ones who build the organizational muscle to continuously learn, adapt, and operationalize new capabilities while maintaining the trust that enterprise customers demand.
To dive deeper into the strategies behind large-scale AI transformations, listen to Sergei Liakhovetsky discuss these ideas in depth on the Dev Interrupted podcast.
