← Workflow Recipes
Workers Comp recipe

How do I run the Monday claims deadline sweep?

A verified Monday-morning list of every statutory clock across a claims portfolio: determination timeframes, freeze records, section 57 examinations and review windows, built by AI and checked by a human.

Tested on 13 August 2026 · Platforms: ChatGPT Enterprise, Claude Cowork · Time required: 20 minutes · For: Case managers, Claims team leaders

The trigger

Monday morning, before the week's new allocations land. Every open claim carries at least one statutory clock: 20 calendar days for an injury claim, 60 for a disease claim, 30 for a claimant's reconsideration request, plus the review windows after a determination. The prescribed periods and their freeze rules are exhaustively listed, so the sweep is arithmetic AI does reliably while a human owns every date.

The inputs

  • A claims system export: claim identifier, claim type (injury or disease), and the date the compliant section 54 claim was received
  • Each section 14 claim's freeze ledger: trigger, notice date, recommencement date
  • Open reconsideration requests with receipt dates
  • Section 57 examination referrals in flight, with appointment dates
  • Recent determinations and reviewable decisions, for the review windows

The steps

  1. De-identify the export first: names become [CLAIMANT_NAME], claim numbers [CLAIM_NUMBER], injury dates [INJURY_DATE].
  2. Paste the export into the prompt below to compute days elapsed, excluded days and projected end dates, with working.
  3. Review the flags: unbacked freezes, freezes on reconsideration rows, and section 57 examinations past their appointment date with no outcome.
  4. Route any missed examination to the delegate for a reasonable-excuse assessment; the sweep records the event, the delegate makes the finding.
  5. Verify every AI-listed date against the source claim record before anything is diarised.
  6. Diarise the verified dates, assign the week's actions, and file the sweep record.

The prompts

Prompt
You track dates only. Never decide anything, and never suggest a
freeze or a notice as deadline relief.

Data: [PASTE DE-IDENTIFIED PORTFOLIO EXPORT]
Rows use [CLAIMANT_NAME], [CLAIM_NUMBER] and [INJURY_DATE].
Freezes show the trigger, notice date [DATE_1] and
recommencement date [DATE_2].

For every row:
1. The clock: 20 calendar days for an injury claim, 60 for a
   disease claim, 30 for a reconsideration request, counted
   from receipt.
2. Days elapsed, days excluded by freezes, and the projected
   end date, with working shown.
3. Flag a freeze with no statutory trigger (section 57, 58 or
   71, or the two further-evidence situations), or any freeze
   on a reconsideration row.
4. Flag a section 57 examination past its appointment date
   with no outcome recorded.
Sort by days remaining, fewest first. Every date is a draft
until a human verifies it.

The checks

  • Every date is verified against the source claim record before the list leaves your hands
  • Day one is the date the compliant section 54 claim, with its medical certificate, was received, not the allocation date
  • Every freeze is backed by an actual section 57, 58 or 71 trigger; a phone call does not stop the count
  • No freeze sits on a reconsideration row: the 30 days has no freeze provisions
  • Review windows match the three review tiers: 30 days for reconsideration, 60 days for lodgement of an ART review, 28 days for a Federal Court appeal

Governance controls

  • The team leader owns the sweep; the delegate owns every determination and reasonable-excuse finding. AI tracks dates and flags gaps, never an outcome.
  • De-identification is mandatory: names, claim numbers, dates of birth and diagnoses never leave the claims system unmasked. Placeholders only, re-attached afterwards.
  • The statutory timeframes are firm and the human owns every date. No prompt or report presents a freeze as deadline relief; freezes exist only where further information is needed for the determination.
  • Every non-counted period is recorded in the claims system as the claim moves, and the sweep record notes the AI's role.

The output

A verified weekly sweep list: every clock, its projected end date, the files closest to the line, and the flagged gaps, filed with the claims records to drive the week's diary.

Time it takes

A first run takes about 40 minutes while the export and prompt are set up; a practised run takes about 20 minutes, mostly human date verification.

TheAICommand. Intelligence, At Your Command.

← Back to the recipe library

General information and education only. Not legal, compliance, financial, or professional advice. This recipe describes a way of working with AI tools; the governance controls, including human review, are part of the procedure and are never optional. Confirm obligations against the primary source and your organisation's own policies.