AI extraction & Review Extracted
The screen that decides everything downstream. Five tabs, all editable, with cascading edits and a preview before you commit.
Lesson plans, scripts, questions, materials and every analytics number are derived from what you approve here. Ten minutes of correction now saves a regeneration cycle later.
The five tabs
Inline editing
Every cell is editable in place — click, type, move on. Rows can be added, removed and reordered. Changes are held per section and saved together, so you can work through a tab and commit once rather than firing a request per keystroke.
- Editable text for names and descriptions.
- Marks cells that validate numerically and total as you type.
- Matrix toggles for mapping an assessment to an outcome or strand.
- Row controls to insert, delete or reorder.
Cascading edits
Structure is connected, so an edit in one place has consequences elsewhere. The screen shows you those consequences before applying them.
Renaming a strand, deleting an outcome or reordering gates updates the dependent rows — outcomes, lesson plans and quiz groups — in one pass. A confirmation shows the diff first: what will change, and how much.
Read the diff. A cascade that touches far more rows than you expected usually means the extraction mapped something to the wrong parent, and the fix is upstream of the edit you just made.
Flags and confidence
Where extraction was unsure, the row is flagged rather than quietly accepted — a low-confidence outcome, an ambiguous CLO/PLO mapping, a page OCR read poorly. A banner collects these so you can work through them rather than hunting. Flags do not block approval; they mark what deserves a second look.
Approving
Approval moves the course out of review and queues generation. Individual gates can be accepted or rejected; a rejected gate is marked pending_regeneration and rebuilt rather than silently dropped.