S-20 flagship chain (GC email → task → overlay → takeoff → CO)¶
| Stage | Estimated |
|---|---|
| Status | Deferred |
| Design status | In review |
| Estimate | 3w+ (XL) |
| Confidence | Low |
| Linear | PIN-253 ↗ PIN-439 ↗ |
| Linear status | Backlog |
| Design | Design ↗ |
| Linear epic | Cara AI & Automation |
| Module | Cara ↗ |
Priority: Medium
Scope¶
Chained workflow: incoming GC emails are AI-classified as task-worthy (suggested task, accept/reject/edit UI, per-user learning feedback loop), while separately the AI takeoff engine overlays each HVAC system on the drawing in a distinct color (by pressure class for ductwork, by system type for piping) so the estimator visually validates the takeoff instead of reading a raw list. (Source: PIN-253, PIN-439.)
Acceptance criteria¶
- O365 inbox connector reads incoming email per user
- AI classifies each email as task / not-task with confidence + a suggested task summary
- Suggested-task inbox UI: accept (creates Task) / reject (feedback) / edit-then-accept
- Every accept/reject persisted to a per-user feedback corpus that trains their personal classifier (never cross-tenant)
- AI assigns a system/pressure-class label to each drawing element and renders it in a matching color
- Pressure-class variants (medium/low) use shade variants of the same base color; unidentified items render in an alert color
- Color overlay renders as a toggleable canvas layer over the existing PDF viewer
- Audit log on every AI-created task
Conor's comments¶
After Cara workspace. This incorporates Cara AI, taekoff engine, Tasks and projects. Will be an automated workflow.
Notes¶
Marquee demo; post-launch; highest integration risk
Open question for Conor¶
-