AI & agents feed
Notes on AI and Agent Systems
Where agents actually change the economics of a workflow — and where governance of agent authority has to sit.
You are deciding what an agent is allowed to do on your behalf, and how you prove what it did.
Publication date unverified · 7 min read
The AI autonomy ladder: assisted, enabled, agent-run
A three-rung model for evaluating workflow authority and five questions to resolve before considering autonomous execution.
Read the note →
Publication date unverified · 6 min read
How to measure an AI deployment's return — and defend it.
Agree the baseline before you build, test on a golden set of real cases, track cost per run, and decide in advance when to stop.
Read the note →
Publication date unverified · 5 min read
Assistants that finish the work: from strategy review to verified result.
A voice or text assistant is useful when it carries approved work over days, stops safely and counts a result only when systems and people confirm it.
Read the note →
Next step
Bring the decision this feed is about.
If the topic is live for you right now, a working session gets you further than an article will.