CONTEXT(CLONE)
Sign in

Review closed claim files against the handling guideline

For: Claims quality assurance lead running periodic closed-file reviews

Pattern: Specialist panelNeeds a connectorDesigned for 25 to 1,000 agents
Get early accessSign in to use

The pain today

Closed-file reviews sample a handful of files because each takes so long to read. Process gaps such as undocumented coverage checks or missed recovery steps go unseen across the rest.

The ask

“Using the closed files exported from our claims system and our handling guideline, check each file for whether the guideline steps are documented: coverage check, reserve notes, recovery considered, customer updates and complaint handling. Show gaps by step with the file passages.”

Plain words, as you would say it to a colleague. Edit it to fit your case before you send it.

What you attach or connect

  • Closed claim files with notes and correspondence
  • Claims handling guideline
  • Complaint handling procedure

The unit of work

One worker task per one closed file per guideline brief.

Why a swarm fits

Five briefs read the same file against different guideline chapters, and files are independent. Each reviewer needs one file and one chapter, so the whole closed book can be read rather than a sample.

Not for

Rating handlers or reopening claim outcomes. Findings are reported by guideline step, not by person.

The decision tree

6 typed decisions, each with an action for every answer

At fixed moments in a run, the engine puts one narrow question to a decision model. The decision model never writes text: it answers yes or no with a probability, picks from listed options, or gives a score, about a small slice of the material. The engine then does exactly what this tree says, which is what makes the run auditable. The thresholds are the template's design values, not measured results.

  1. Before workers, before a task runs

    Small worker or strong modelA choice among options

    Before a task runs

    Does the guideline chapter for this brief apply to a file of this product line and claim type at all?

    Sees only: The file's product line and claim type, and the chapter's scope paragraph

    Why: Recovery or complaint briefs are not run on files where the guideline never required them.

    • AppliesthenAccept
    • Does not applythenSkip this unit
    • Applies with exceptionsthenEscalate to a strong model

  2. After workers, the judge checks

    Evidence checkYes or no, with a probability

    After a worker answers

    Does the quoted file note or letter record that the guideline step was carried out, rather than merely mention the subject?

    Sees only: One guideline step as written and one quoted file note or letter

    Why: A step counts as documented only on real evidence, so the gap figures can be trusted.

    • Yes: 0.85 or higherthenAccept
    • Unsure: 0.50 up to 0.85thenEscalate to a strong model
    • No: below 0.50thenReject and retry

  3. Evidence checkYes or no, with a probability

    After a worker answers

    Across the file's full index of notes and correspondence, is there no entry that records this guideline step?

    Sees only: One guideline step and the file's index with the entries searched

    Why: A gap is a criticism of process, so it must survive a full reread before it is reported.

    • Yes: 0.85 or higherthenAccept
    • Unsure: 0.50 up to 0.85thenMark unresolved
    • No: below 0.50thenReject and retry

  4. Evidence checkYes or no, with a probability

    After a worker answers

    Is this finding written about the guideline step and the file text only, with no handler named or described?

    Sees only: One drafted finding

    Why: The review is about process by step, never an assessment of a person.

    • Yes: 0.90 or higherthenAccept
    • Unsure: 0.60 up to 0.90thenReject and retry
    • No: below 0.60thenReject and retry

  5. Reconciler, while merging

    Conflict checkYes or no, with a probability

    While reconciling

    Do files that documented this step differently each have a reasonable basis in the guideline's wording?

    Sees only: The guideline step's text and two differing documented examples

    Why: Points the quality lead at unclear guidance instead of at files.

    • Yes: 0.70 or higherthenMark unresolved
    • Unsure: 0.30 up to 0.70thenMark unresolved
    • No: below 0.30thenAccept

  6. Accountable person, before anything is settled

    Person decidesYes or no, with a probability

    Before anything is reported as settled

    Does this finding suggest a customer may not have been updated, a complaint not handled, or a recovery not considered on this file?

    Sees only: One file's accepted gaps with their guideline steps

    Why: Anything that might touch a customer goes to the claims quality lead for a human read of the file.

    Accountable: The claims quality lead reviews flagged files and decides on training or guideline changes.

    • Yes: 0.20 or higherthenAsk a person
    • Unsure: 0.05 up to 0.20thenAsk a person
    • No: below 0.05thenAccept

The fleet: who does what

Model tiers by role, not brands: you choose the models. Strong reasoning models plan and reconcile, small fast models do the wide work, and the judge is a decision model from a different family, so it does not share the workers' blind spots.

  1. Planner

    A strong reasoning model turns each guideline chapter into a brief of documented steps to look for.

  2. Workers

    Small fast workers from mixed families, one per file and brief, quoting the note or letter that evidences a step.

    Designed for 25 to 1,000 agents, one worker task per one closed file per guideline brief. Each worker receives only its own unit.

    Decisions here:1. Small worker or strong model

  3. Judge, from a different model family

    A decision model from a different family checks that the quoted note really evidences the guideline step.

    Decisions here:2. Evidence check3. Evidence check4. Evidence check

  4. Reconciler

    A strong reasoning model reports gaps by guideline step and product line, never by handler.

    Decisions here:5. Conflict check

  5. Accountable person

    The claims quality lead reviews flagged files and decides on training or guideline changes.

    Decisions here:6. Person decides

Checked before anything is accepted

  • A step counts as documented only with a quoted note or letter
  • A gap is reported only after a second worker rereads the full file
  • Findings are aggregated by step; handler names are excluded from outputs
  • The run comments on process evidence, never on the claim outcome

What comes back

  • Gaps by guideline step and product line
  • Files flagged for human review, with passages
  • Guideline steps that files show to be unclear
  • Files that could not be read in full

What to measure

  • Share of flagged gaps the quality lead confirms
  • Coverage of the closed book compared with manual sampling
  • Reviewer hours per confirmed gap
  • Cost per file reviewed

Names of measures only. No result is claimed for this template.

Templates open in the workspace chat with the ask filled in. Nothing runs until you send it.

Get early accessSign in to use

More in Insurance and claims

Map one exposure across the whole wording portfolio

For: Wordings manager or portfolio underwriter responsible for product wordings

When a new exposure appears, nobody knows which wordings, endorsements and legacy versions cover it, exclude it or say nothing.

Pattern: Map, verify, reduceNeeds scale6 decisionsDesigned for 40 to 800 agents

Prepare a coverage brief with clauses for and against

For: Claims handler or coverage specialist preparing a position on a complex commercial loss

The schedule, wording, endorsements and loss report each run long.

Pattern: Cross-examinationRuns today7 decisionsDesigned for 4 to 32 agents

Watch wording changes and recheck only what depends on them

For: Product governance lead or wordings manager at an insurer or managing agent

A clause edit in a base wording silently affects endorsements, customer summaries, claims guidance and broker templates built on it.

Pattern: WatchtowerNeeds a connector5 decisionsDesigned for 20 to 500 agents