What building a company-wide AI program taught me about culture, leadership, and the only thing that actually drives change.
Over the past 18 months, I went through the UC Berkeley CEO Program. One of its core requirements is a Capstone Project — a real strategic challenge from your own business, worked through with the rigor of an academic framework and the pressure of actually having to implement it.
I chose AI.
Not because it was trendy. Because it was the most urgent unsolved problem in my company, and I didn’t yet have the right tools to think about it clearly.
What followed was one of the most instructive failures — and recoveries — of my career as a CEO.
The Goal
Mobio Group is a performance marketing agency. We operate in a fast-moving, data-intensive environment where productivity and speed are existential. The goal of my Capstone Project was specific: design and implement an internal AI Enablement Program that would increase productivity by at least 30% per employee — not through a specialist team working in isolation, but by embedding AI literacy and usage across every department of the business.
In other words: not an AI team. An AI company.
What We Built – and What Broke
We moved fast. We created a dedicated AI Committee, positioned horizontally across all business units — acting as an enabler and bridge, not a siloed tech function. Within 2025, the team delivered three internal AI-powered solutions: a fraud monitoring tool, an App Store optimization system, and an AI-assisted candidate sourcing and hiring pipeline.
And then we hit real walls. Three of them.
Wall one: employee resistance — not opposition, indifference. Most frontline employees were simply not motivated to change how they worked. They weren’t actively opposed to AI. They just didn’t see a problem. And that turned out to be harder to solve than outright resistance. Formulating meaningful requirements for the AI team was even harder than adopting new tools. When people don’t volunteer their problems, the AI Committee can’t find valuable solutions. The initiative was stuck at the surface.
Wall two: the pace of change broke our knowledge transfer cycle. At the start of 2025 we were rolling out ChatGPT and Midjourney for the creative team. By mid-year we had shifted to Claude and Google’s tools. By early 2026, we were already working with AI agents and advanced coding capabilities. The pace meant that by the time the AI team had mastered a technology well enough to teach it, a better system had already emerged. We were always behind.
Wall three: a structural gap between AI capability and business context. The AI Committee lacked deep enough understanding of the actual business problems in each department to design truly valuable solutions on their own. And the department heads, even when engaged, offered only surface-level tasks. Public recognition, internal champions, department AI ambassadors — all of these helped, but not enough. The initiative kept living at the top of the org chart instead of flowing through it.
Two Interventions That Changed the Dynamic
After several months of incremental progress and honest assessment, two things shifted the trajectory.
First: leadership by example — with teeth.
Mobio’s Executive Director built a Client Service Radar using AI coding tools: an agent that monitors activity across Asana, Slack, Notion, and HubSpot and generates a comprehensive real-time brief on each key client. He then challenged every senior manager and department head to apply the same approach to their own teams’ challenges. This was not a suggestion — it was an assignment.
Leaders showing what’s possible, then requiring others to follow, shifted the culture in a way that ambassadors and rewards had not. The key word is requiring. Inspiration alone is not enough. At some point, the standard has to change.
Second: mandatory AI certification — with a soft but firm incentive structure.
We piloted this in the Performance Marketing department: employees must pass a practical AI certification to unlock full promotion and salary review opportunities. It’s technically voluntary. But without the certificate, your development trajectory is capped.
This created real motivation to engage — not out of enthusiasm, but out of self-interest. And commitment, it turns out, produces understanding. Once people actually work with the tools to solve their own tasks, they start seeing the possibilities themselves. The sequence matters: commitment first, conviction second.
The Honest Takeaway
AI adoption is not a technology problem. It is a culture and leadership problem.
The tools are accessible. The bottleneck is always human — resistance, inertia, the gap between where people are and where they need to go. The frameworks that proved most useful weren’t technical ones. They were the frameworks I was studying at Berkeley: culture, incentives, leadership styles. They turned out to be exactly the right lens for solving what looked, on the surface, like a technology implementation challenge.
We haven’t reached that 30% yet. But we’re closer than we were — and more importantly, we now understand why the first approach wasn’t going to get us there. The target didn’t change. Our understanding of how to get there did.
There is one more thing I learned that I didn’t fully appreciate at the start: you cannot understand a technology’s potential without personally wrestling with it on a real problem. Watching a demo, sitting through a keynote, attending a workshop — none of that is enough. Understanding only comes from doing. From taking an actual task you own, applying the tool yourself, hitting the walls, and getting to the output.
Only after that experience does a person genuinely grasp what these technologies can and cannot do — and more importantly, where they could change the way their own work gets done.
This is why our certification program is built around hands-on problem-solving, not passive learning. It is the only path to real adoption.
Sergey Konovalov is CEO of Mobio Group and a graduate of the UC Berkeley CEO Program (Class of 2026).
