← Blog

AI & Agents · Feb 24, 2026 · 3 min read

IRIS: How I Built My Chief of Staff Out of Prompts and Logic

The origin story of IRIS, from a scattered task list to a fully functional AI Chief of Staff that now runs the daily operation of 20+ companies.

Ajdin Brković

Ajdin Brković

Published Feb 24, 2026

In 2023, I was drowning. I had 8 companies at that point, and they all needed decisions. Every day, I had 200+ emails, 50+ Slack messages, 30+ decisions that needed to be made. I was hiring a COO, but really I was hiring someone to filter the noise so I could focus on the 3 things that actually mattered.

Then I realized: why do I need a human for this? I don't need emotional intelligence. I don't need creative problem-solving. I need someone to read everything, understand context, categorize it, flag what matters, and delete what doesn't. That's a very solvable problem with AI. So I built IRIS.

IRIS isn't a chatbot. It's a system. It connects to my email, Slack, calendar, and project management tools. Every morning at 6 AM, it reads everything from the past 24 hours. It categorizes by urgency and impact. It generates a daily brief with the 5-7 things I actually need to know and decide on. The rest just disappears.

The early versions were terrible. They'd flag random things as important. They'd miss critical emails. They'd categorize things wrong. So I spent months refining the prompt, adding context rules, training the system on historical decisions. What should IRIS escalate? What should it handle independently? What should it wait on?

What's wild is that IRIS became better at understanding the business than I am in some ways. It knows the revenue targets for each company. It knows the cash runway. It knows the key metrics. So when a decision comes through, IRIS doesn't just flag it. It flags it with context. "This customer churn issue impacts Company 7's runway by 2 months. Recommend immediate action." Now I'm making decisions with better information than I had before.

The other breakthrough was delegation. IRIS doesn't just escalate to me. It escalates to the right person. If a marketing question comes in, it routes to the marketing lead. If it's a product question, it routes to the PM. If it requires my decision, it escalates to me. So I went from being a bottleneck to being a decision-maker. My team doesn't wait for me to read Slack. IRIS finds them directly.

A Chief of Staff costs $200K a year and still can't process information as fast as an AI. So I built the AI.

Now, in 2026, IRIS isn't just my chief of staff. It's the chief of staff for my entire portfolio. It knows about all 20 companies, their metrics, their challenges. It recommends which company I should focus on today based on urgency and impact. It's like having a voice in my head that understands the entire business.

This is what I'm excited about. Not AI replacing humans. But AI doing the work that humans hate and are bad at. The filtering. The organizing. The routing. The categorization. So humans can do what humans are good at. Strategic thinking. Creative work. Relationship building. That's the future.

Ajdin Brković

Ajdin Brković

CEO & Founder

Building 20+ companies simultaneously using AI systems and performance marketing. Previously exited 3 ventures.

Weekly insights

Ideas on systems, AI, and building companies. One email per week.