WhitePaper – Autonomous Decision Making Agents
ActorDo started as exploring a new kind of AI assistant for knowledge workers: one that does not only suggest, summarize, or draft, but can also act within safe and clearly defined boundaries.
The practical research focuses on a key question: when should an AI assistant act autonomously, and when should it stop and ask for confirmation?
The paper introduces a practical autonomy model based on risk, reversibility, user control, and trust, showing that low-risk tasks like labeling emails, drafting replies, or suggesting follow-ups are the best starting points.
A core principle is that every autonomous action must be visible, explainable, logged, and reversible. The goal is to build an assistant that reduces daily micro-decisions without removing the user’s control.
While ActorDo evolved and product was built, some aspects in the document were built in a slightly different dirrection.
I’ve published the current whitepaper here:
https://docs.google.com/document/d/1qNYtXkUx9rna8M9_A5Y9gS5pkI4bUSBZ6dgAyHVEBjE/edit?tab=t.0
Feel free to read, share, or use it if it’s useful.
I’d be happy to discuss the topic further, exchange ideas, or provide a reference if it helps with your own work.
Connect on Linkedin Alex Rada

