Goodwood runs its own delivery on Claude, and builds the same capability for clients. What makes that safe for financial services work is not the model. It is the written rules around it.
The Rules Come First
- Writes to client data are guarded — off by default, one pause switch, a ledger written before each call, caps that refuse, read-back, and a drilled rollback
- A person approves what leaves the building — client-facing sends, production deploys, permanent deletes, and spend always wait for a yes
- Every target is explicit — each deploy and each write names its project or portal, so work cannot land on the wrong client
What This Looks Like for Clients
Connectors, supervised agents, and AI inside client deliverables, such as meeting summaries and transcript processing, all run under the same safeguards as any other client system. The data side is covered by goodwood intelligence, which is coming soon: advisor and fund data with the judgment to act on it.
Want AI on your revenue system, with the rules attached?
Schedule a Call