Strategy engine
Learning loop.
Profiles evolve as evidence accumulates. Pull signals from across Chapter Tools, let Claude propose updates, then review and apply.
Proposed updates (0)
All profiles up to date. Run the loop to look for new patterns.
How this works
The learning loop reads recent signals (manual + synced from Newsroom, Newsbeat, Grid, Pulse, Actions), groups them by brand, and asks Claude to propose small, evidence-backed updates to brand profiles, audience personas, or brand×persona links.
Claude never edits a profile directly. Every proposed change comes here for review, with its rationale and citation back to the source signals. Once you approve, the edit is applied and logged.
You can also call POST /api/learn from outside (with the integration token) to run it on a schedule.