Deadline Command Centre
Task + reminder engine built specifically for Canadian immigration matters. Upload a PFL, ADR, interview call-in, IAD appeal record, PRRA notice, or processing-delay confirmation and Claude Sonnet 4.6 extracts every operative date with evidence citations you confirm before anything is scheduled. Per-recipient reminder ladders, non-suppressible safety-net escalation on every hard deadline, four-layer reminder resolution, external mirror to Google Calendar or Outlook. Command Centre Landing all-tier; AI intake + scenario rules + external recipients + role dashboards Premium.
Download PDFOverview
The Deadline Command Centre (DCC) is the platform's task + reminder engine built specifically for Canadian immigration matters. Every task carries an assignee, a deadline class (hard or soft), an optional reminder ladder, and a Command Centre entry so no important date falls through the cracks. Command Centre Landing + personal + firm task CRUD + reminders + kill switches are included on every tier as of v1.23.0 (2026-07-02). AI Deadline Intake, the Scenario Rule Library, external recipients (client + third-party reminders with CASL consent + redaction), Owner / RCIC / Assistant role dashboards, and the external calendar mirror are Premium.
What the module does. Personal tasks and firm-wide tasks. Auto-created tasks from Transfer Room outbound / inbound events, from Active File Review AI intake, and from expiry dates on the Transfer Room Information Card (passport, medical, biometrics). AI-extracted deadlines from uploaded IRCC / CBSA / IRB letters. Recurring tasks with DST-correct wall-clock anchoring. Client meeting invites tied to tasks. External calendar mirror to Google Calendar or Outlook.
What it doesn't do. The AI never sets a deadline for you — every extracted date is a proposal you review and approve before anything is scheduled. The platform does not submit IRCC forms on your behalf. Reminders never go silent — if consent is missing on a client / third-party recipient, an internal follow-up task is created so a human closes the loop.
Onboarding checklist
Follow this checklist the first time you open DCC. Everything below is optional — the module works fine out of the box — but doing these makes the platform work harder for you.
Set your firing hour (Reminder Policy Centre). Reminders honour a per-member firing hour so a 09:00 reminder actually fires at 09:00 in your timezone even if the offset math would have picked 02:00 UTC.
Pick your ladder defaults per task type. The system default is 14 / 7 / 3 / 1 / 0 days-before-due for hard deadlines and 7 / 3 / 1 for soft. Customize per-type from your Reminder Policy Centre if you prefer a different cadence.
Set your weekend-shift rule. If a reminder would fire on a Saturday or Sunday, shift to the prior Friday. On by default; turn off from your Reminder Policy Centre if your practice takes weekend reminders in stride.
Turn on the external calendar mirror (Premium). Opt in per member to see your DCC tasks alongside your bookings + meetings on your Google or Outlook calendar. Only busy-blocks are pushed — no attendee list, no client detail.
Set kill switches from admin (superadmin only). Global + per-tenant kill switches on every engine (intake, reminders, auto-sync, Transfer Room reconciliation, meeting invites) let you pause a specific engine without breaking the module.
Test the AI intake pipeline (Premium). Upload a sample PFL or ADR letter and walk through the review + approve flow so you know exactly what you'll see on a real letter.
AI Deadline Intake
The keystone Premium feature. Upload an IRCC / CBSA / IRB letter and Claude Sonnet 4.6 reads it end-to-end + extracts every operative date with a page number + verbatim evidence quote from the source. Nothing is scheduled until you review + approve the extraction.
Supported letter types: PFL (Procedural Fairness Letter), ADR (Additional Documents Request), interview call-in notice, IAD appeal record acknowledgement (AOR), PRRA notice, processing-delay confirmation, and a free-form fallback that carries basic scenario metadata for anything else.
Review UI. The extraction lands on a page where you see the letter preview side-by-side with the highlighted quotes. Every proposed date shows the AI's proposed deadline class (hard or soft), the AI's confidence score, and a 'Jump to source' link that scrolls the preview to the extracted page. You can override the AI's classification per proposal, edit the date, edit the task title, or reject the proposal entirely.
Hard-deadline gate. Approving a hard-deadline proposal requires you to type-confirm the operative date and BOTH an evidence page + evidence quote must be present on the proposal. This is a hard rule — the platform will not let a hard deadline through without your explicit confirmation of the operative date + a citation you can trace back to the letter.
Content-hash idempotency. The extraction is content-hash idempotent, so re-uploading the same letter doesn't duplicate anything. If you accidentally upload the same PDF twice you get 'this letter has already been extracted' + a link to the previous bundle.
Extraction quota. Each tenant has a daily extraction quota (25 letters/day default; superadmin can raise). Failed extractions do not consume quota; successful extractions consume one unit regardless of the letter size.
Sample workflows — the five driving scenarios
Scenario 1 — PFL response. Upload the PFL. AI extracts the response deadline (14 days from the letter date is the IRCC default). You review the page + quote, type-confirm the deadline, and approve. Result: a hard_deadline task with the 14 / 7 / 3 / 1 / 0 reminder ladder + non-suppressible safety-net escalation, a calendar chip appears in the Command Centre, and a client-side reminder ladder is queued (with CASL consent check on the first send).
Scenario 2 — Interview call-in. Upload the interview notice. AI extracts the interview date + suggests prep tasks (client documents review 7 days before, prep call with client 3 days before). You review, approve. Result: multiple tasks land under one bundle — the interview date as a hard deadline, the prep tasks as soft deadlines counted backward from the interview. If you also send a client meeting invite from the interview task, the client can self-schedule a 15 / 30 / 45 / 60-minute working session on your calendar.
Scenario 3 — IAD appeal record (AOR). Upload the AOR. AI extracts the CBSA production window (60 days by default). You approve. Result: an expectation-watch task with blocked_by=cbsa, expected_by = AOR + 60 days, and manual state 'not_received'. Also created: a linked fallback task worded 'Review whether to follow up with, or complain to, the IAD' that ACTIVATES only if the watch task hasn't been marked 'received' by day 60. If you mark the watch task as received earlier, the fallback never activates.
Scenario 4 — Processing delay. Upload the processing-delay confirmation letter. AI extracts the enquiry date and creates a quarterly recurring task (FREQ=MONTHLY;INTERVAL=3). Every three months a fresh occurrence fires with a merge-field payload rendering IRCC's web-form info in a copyable block. Completing one occurrence records a 'web form sent' completion detail; the series continues so the next quarterly occurrence still fires.
Scenario 5 — PRRA two-date. Upload the PRRA notice. AI extracts BOTH the initial submission deadline AND the perfecting deadline as two independent hard_deadline proposals. Approving both creates two independent tasks under one intake_bundle_id, shown as linked siblings. Each carries its own ladder. Completing one does not cancel the other — they are decoupled by design.
Privacy and redaction
DCC's privacy posture is 'never leak client detail to outsiders by default'. Every external send goes through server-side redaction + a CASL consent check. Every kill switch trips silently — an unavailable engine emits an audit event but never surfaces client detail in the failure mode.
Client reminders. Full task detail (title, notes, due date, ladder) is included in the email body for clients on Premium. Client is the subject of the matter, so this is expected. CASL consent is checked per recipient on every send — no consent = suppression + an internal follow-up task on the source assignee's queue so the RCIC closes the loop manually.
Third-party reminders. Redacted server-side to firm name + due date + a public label you supply on the recipient row. No task title, no notes, no source module context. The recipient row's 'public label' field is the ONLY string that shows up in the email body — anything you type there is what the recipient sees.
External calendar mirror. Only busy-blocks are pushed — no attendee list, no client detail, no matter number. The primary calendar receives full detail; the mirror sits in your workspace only. If a family member has access to your Google account, they see the task title (which YOU control) but no client-side detail.
Audit ledger. Every mutation is journaled to an append-only ledger. IP addresses on client-facing endpoints are truncated to a /24 CIDR before logging. Unmasked personal info stays out of the ledger by construction.
PII lifecycle. Uploaded letters are stored temporarily under your tenant DEK during extraction; the raw bytes are purged from Supabase Storage as soon as the bundle is approved or 24h after upload — whichever comes first. Extracted data (dates, evidence quotes) persists on the DCC tables so a completed bundle can still be reconstructed for audit.
Audit export — PDF records + CSV/JSON history
Every task carries a plain-language 'Why am I seeing this?' provenance panel with source, scenario rule, tenant edits, and the responsible member. From the task modal you can export the full history as a PDF for a file handoff or as CSV/JSON for internal review + reconciliation.
PDF export (records). Formatted PDF that includes the task title, due date, class, ladder, recipients, every state transition, every reminder outcome (sent / failed / suppressed with reason), extension requests + decisions, completion basis + reason. Suitable for including in a matter's paper file or handing off to a covering RCIC.
CSV export (history). Row per event: timestamp, actor, event type, before-state, after-state, reason. Suitable for internal audit review, spreadsheet analysis, or handing to an accountant reconciling billable time against completion timestamps.
JSON export (machine-readable). Full task snapshot with every nested field including provenance, scenario_rule_version, intake_bundle metadata, and the full event ledger. Suitable for integrating with a downstream firm-wide analytics system or reconstructing the exact rule text in force at approval time.
