Writing
Notes on agentic AI in production.
Essays on what actually works when you put AI agents to work for real businesses — architecture patterns, failure modes, the operational reality behind the demos. Less hype, more receipts.
-
Designing human-in-the-loop approval surfaces for agentic systems.
Why every AI automation I install has an approval step on consequential actions, and what makes the approval interface either trustworthy or theatre. In progress.
More essays landing through Q3 and Q4 2026.
Topics in the pipeline: choosing between Claude Code, n8n, and Zapier for SME automations; what failure modes look like in production agentic systems; the operational economics of running a one-person consulting practice on AI tooling. Want one of them written sooner — or want to suggest one? hello@goozel.io.