Build Log #1 — what my AI can and can’t do yet

This is the opening entry of a rebuild in the open: a plain inventory of what the named AI agents at Big Ben Suite draft today, and where a named human still decides everything irreversible. The full log is on LinkedIn and on the Big Ben Suite journey page — this note is the short version, with the link out.

Build Log #1 — what my AI can and can’t do yet.

I said I’d document this rebuild honestly. So here’s the starting inventory, no varnish.

The short version

What the agents do today.They draft — pricing reviews, content, competitor scans, weekly plans — and every draft waits for a named human before it goes anywhere. They test before shipping: nothing reaches the live sites without passing the full test gate first. And they get audited: in July we ran an external-style audit of our own websites and removed numbers we couldn’t evidence — the corrections are logged in public.

What they can’t do yet.They can’t talk to customers — the customer-facing engine exists in code, deny-by-default and tested, but it isn’t connected to live channels; until it is, it says “in rollout,” not “live.” And they can’t run without me: every irreversible decision still crosses my desk. That’s the design, not an excuse.

From the next entry, the log takes a fixed shape — Added, Removed, Broke, Learned. If a week is missed, the log says why.