Most of my work now runs through a team of AI agents. Not one big chat or multip
By Andrew Bailiff · October 31, 2026 · Curated by George's Blog
Most of my work now runs through a team of AI agents.
Not one big chat or multiple sessions. A small org chart.
↳ One lead session I talk to every day
↳ Leads that each own one project
↳ Workers that each get one narrow job
Every worker writes its full output to a file. Then it sends back it's part: status, where the file is, what it checked, and anything that didn't check out.
Before an agent relies on a note, it checks the note against the live data. If they disagree, it has to say so.
And there are lines none of them cross on their own:
→ Nothing client-facing gets sent without approvals
→ No live campaign changes until authorized
Those wait for me.
How are you splitting up AI work right now: a bunch of sessions, or a team?