Manager 2.0: From Machine Learning to Autonomous Agents – Building Own AI Team

Before the surge in hardware prices, I built a local GPU server to host my own privacy-first “AI Workforce” using n8n and Ollama. Here is my AI journey from the past to the present, and a look at the operational assistants currently working for me behind the scenes.

(The Genesis: My Journey Before ChatGPT)

My introduction to artificial intelligence didn’t start with ChatGPT. Long before that, it began with online courses and training anomaly detection models in preparation for my Master’s thesis.

When Generative AI (GenAI) entered our lives, I integrated a chatbot into my blog and developed https://kids.sysarticles.com—a project that generates AI coloring pages for children. I personally witnessed how those initial distorted, irregular drawings evolved into today’s eye-catching visuals. I didn’t stop there; I built bots that automatically generated content for Sysarticles and Twitter, a personal planning application, and a mobile app that seamlessly extracted data from bulk photos of business cards and books into my database.

However, all of these still involved operational processes. It was time for me to hand over the operational work and become the manager of a 24/7 autonomous team working in the background.

(The Turning Point: Local Server and Architecture)

About 3-4 months ago, I foresaw a rapid global increase in RAM and hardware prices. Within my budget, I quickly assembled a GPU-powered system. I designed this system not for gaming, but to build an AI base running in my home, independent of cloud companies, ensuring my data never leaks.

• Orchestration: All workflows are designed on n8n running on my local network.
• Memory (RAG): I use Qdrant Vector Database and PostgreSQL so my assistants don’t forget the past.

(Current Status: Assistants in the Field)

My ultimate goal is to build a massive 7-person team. Right now, the brain of this team, “Digitali,” is busy reading, indexing (RAG), and learning from my years of file archives. However, two critical members of the team are already deployed in the field:

1. Haydar (Assistant and Planner): Follows my personal “to-do” lists and personal emails. It provides reminders, analyzes information, and delivers evaluated insights.
2. Servet (Financial Advisor): Since we are in an era where economics is crucial, this is the assistant I continue to focus on. Servet currently has two main tasks. First, it wakes up when a new video is uploaded to the niche YouTube channels I follow, watches the video (transcript analysis), and directly emails me the financial and technical breakdowns since they are too detailed for Telegram. Second, it manages my budget. It reminds me of my daily payments. When I’m out, I simply send a text message via Telegram saying, “I spent X amount, I made Y investment.” It understands this, logs it into our shared Excel file, provides ideas on how to evaluate my remaining budget, and conducts investment vehicle analysis if requested.

(Future Assistants in the Pipeline)

• Mimar (Software Architect): Design and coding assistant.
• Çelebi (Content Creator): An assistant that conducts research, produces reports, and generates content when I provide a topic.
• Emin (Family Counselor): Keeps track of the kids’ school calendars, upcoming holidays, and interests. Checks the weather for holidays, creates activity routes, and offers personal development suggestions.
• Kanuni (Legal and Compliance Expert): Follows Kamu SM or Official Gazette RSS feeds and alerts me when there is a new legal change related to the sector. Conducts research and prepares reports when given a task.

(Conclusion)

I highly recommend creating assistants—not just using AI as a chatbot—to save time, spend your time more efficiently and enjoyably, increase the quality of your work, find solutions to your problems, or speed up your tasks.