Agentic AI is where a lot of my building time goes right now. Not a chatbot that answers a question, but systems of AI agents that plan, use tools, and carry a multi-step task through to a finished result on their own.
What interests me is the engineering around the model, not the model itself: giving agents the right tools, holding their output to a strict standard, and keeping a person accountable for what ships instead of sitting in the loop on every step. Done well, agentic systems take on real, repeatable work and stay reliable without constant supervision.
How I approach it
- Narrow, clear roles for each agent instead of one model trying to do everything.
- Hard, checkable standards an agent's output has to pass before it counts as done.
- A person who owns the quality bar and reviews outcomes, not keystrokes.
My role
I design and build these systems end to end. The part I care about most is what makes them trustworthy: the guardrails, the checks, and the judgment about when automation is actually good enough to run on its own.