How I built My AI Chief of Staff
By Dušan Šenkypl, CEO at Groupon

Groupon CEO Dušan Šenkypl shares how he built an AI Chief of Staff using Claude Code - and why agentic AI is changing the way leaders plan, research, review, and execute.

Two hours. That’s how long it took Dušan Šenkypl, CEO of Groupon, to build the first version of his AI Chief of Staff.

In his recent LinkedIn article, Dušan explains how he moved beyond using AI as a chat tool and started building a real operating system for work. Within a week, that system was managing more than 20 projects across strategic planning, executive communications, project research, and other domains. It also supported personal projects outside work - from building systems for his hockey club to analysing AHL players and recommending leadership strategies for Europe.

At the centre of the setup is a single file called CLAUDE.md. Dušan describes it as the AI’s constitution - the rules it follows across every session, project, and interaction. His system is designed to plan before acting, re-plan when needed, automatically seek second opinions on important outputs, capture learnings over time, and verify results before marking work complete.

That structure matters because, as Dušan writes, the real challenge of leadership is not just breadth - it is depth. A CEO or founder works across strategy, architecture, hiring, finance, go-to-market, and organisational design. To be a true partner to leaders across the business, you need to go deep enough into their domains to challenge assumptions, spot what is missing, and make decisions that hold up. Traditionally, that has meant hours of reading, fragmented context, and too much reliance on summaries. His aim was to build something more powerful: not a chat window, but an operating system.

The second key piece is connectivity. Through integrations, the AI operates inside Dušan’s real work environment - across project management tools, knowledge bases, email, calendar, design tools, code repositories, and internal systems. The principle is simple: if he has access to it, the AI has access to it too. That is what allows it to function more like a Chief of Staff than a chatbot.

He also built reusable “skills” - structured workflows for recurring tasks such as second-opinion reviews, architecture analysis, brainstorming, debugging, and multi-phase execution. For more complex projects, the system can also orchestrate multiple agents in parallel.

For Dušan, the impact is bigger than productivity alone. His view is that once you adopt an agentic mindset, you do not just move faster - you raise the bar on the quality of thinking, decision-making, and execution.

What started as a personal AI Chief of Staff is now part of a broader operating philosophy: AI not as a side tool, but as a core way of working.

Read the full article on LinkedIn