CONTEXT(CLONE)
Sign in

Stress-test a delay claim against the project record

For: Contracts manager or claims consultant preparing or defending an extension of time claim

Pattern: Adversarial reviewNeeds live modelsDesigned for 10 to 250 agents
Get early accessSign in to use

The pain today

The record of a delay is spread over site diaries, information requests, instructions, meeting minutes and letters. A claim built from the favourable part of that record fails when the other side produces the rest.

The ask

“I attached our draft delay claim, the contract notice provisions, and the site diaries, information requests, instructions, minutes and correspondence for the period. For each delay event, build the chronology that supports us, then attack it the way the other side would. Tell me which events stand, which notices were late, and what evidence is missing.”

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

  • Draft claim narrative with delay events
  • Contract conditions on notice and time
  • Site diaries and meeting minutes
  • Information requests, instructions and change orders
  • Project correspondence

The unit of work

One worker task per one delay event with its dated records.

Why a swarm fits

Each delay event draws on its own thin slice of a very large record. Events are independent, and an opposed team per event finds the adverse documents a drafter tends to skip.

Not for

Critical path or delay analysis on the programme, or valuing the claim. It tests the documentary record, not the schedule logic.

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. Planner, while planning

    Scope checkYes or no, with a probability

    Before work starts on a unit

    Is this record dated within the period of the delay event and does it mention the work area, drawing or instruction the event concerns?

    Sees only: One record's date and subject line with the event's description

    Why: Gives each event a thin, relevant slice of a very large project record.

    • Yes: 0.60 or higherthenAccept
    • Unsure: 0.30 up to 0.60thenEscalate to a strong model
    • No: below 0.30thenSkip this unit

  2. After workers, the judge checks

    Evidence checkYes or no, with a probability

    After a worker answers

    Does the quoted record carry the date and author that the chronology entry gives it?

    Sees only: One chronology entry and the header of the record it cites

    Why: Every line of the chronology must survive a check against the document it relies on.

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

  3. Evidence checkA choice among options

    After a worker answers

    Does the quoted notice letter's date fall within the period the quoted contract clause allows after the event date in the chronology?

    Sees only: The notice clause, the notice letter's date line and the event date

    Why: Late notice is the first point the other side will take, so it is tested from dates on paper.

    • Within the periodthenAccept
    • After the periodthenMark unresolved
    • No notice found in the recordthenMark unresolved
    • Clause period depends on awarenessthenAsk a person

  4. Evidence checkYes or no, with a probability

    After a worker answers

    Does the quoted record attribute the same delay to a cause within our own control, such as labour, plant or late submittals?

    Sees only: One delay event and one record offered by the challengers

    Why: The challengers' adverse documents are held to the same quoting standard as the supporting ones.

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

  5. Run control, between rounds

    Another round?Yes or no, with a probability

    Between rounds

    Did the latest challenge round produce no adverse record that was not already listed for this event?

    Sees only: The adverse record list before and after the round

    Why: Stops searching the record for an event once the challenge has run dry.

    • Yes: 0.70 or higherthenStop
    • Unsure: 0.40 up to 0.70thenContinue
    • No: below 0.40thenContinue

  6. Accountable person, before anything is settled

    Person decidesYes or no, with a probability

    Before anything is reported as settled

    Does the finding weaken an event, show a late or missing notice, or surface a record adverse to the claim?

    Sees only: One event's verdict with supporting and adverse quotes

    Why: The contracts manager and counsel decide what is claimed, conceded and disclosed.

    Accountable: The contracts manager and legal counsel own what is claimed, what is conceded and what is disclosed.

    • Yes: 0.40 or higherthenAsk a person
    • Unsure: 0.15 up to 0.40thenAsk a person
    • No: below 0.15thenAccept

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 splits the claim into events and indexes the record by date, author and subject.

    Decisions here:1. Scope check

  2. Workers

    Small fast workers from an open-weight family build the dated chronology for one event with quoted records.

    Designed for 10 to 250 agents, one worker task per one delay event with its dated records. Each worker receives only its own unit.

  3. Judge, from a different model family

    Challengers and a decision model from a different family search the same record for contradicting entries and late notices.

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

  4. Reconciler

    A strong reasoning model reports per event what survives, what is contested and which records cut against the claim.

    Decisions here:5. Another round?

  5. Accountable person

    The contracts manager and legal counsel own what is claimed, what is conceded and what is disclosed.

    Decisions here:6. Person decides

Checked before anything is accepted

  • Every chronology entry cites a dated document
  • Notice timing is checked against the quoted contract clause
  • Adverse records are reported with the same prominence as supporting ones

What comes back

  • Chronology per delay event with sources
  • Events that survive challenge and events that do not
  • Late or missing notices with the clause relied on
  • Evidence gaps to close before submission

What to measure

  • Challenges the claims consultant accepts as real weaknesses
  • Adverse documents first found by the swarm
  • Hours per delay event reviewed
  • Cost per event

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 Real estate and construction

Find conflicts between specifications, schedules and drawing notes

For: Design manager or lead consultant coordinating disciplines before tender or construction issue

Architectural, structural and services documents are written by different teams.

Pattern: Hierarchical decompositionNeeds a connector6 decisionsDesigned for 30 to 500 agents

Level subcontractor bids against the scope, by specialist brief

For: Commercial manager, quantity surveyor or estimator comparing trade package bids

Bids for the same package arrive in different formats with different exclusions, qualifications and programme assumptions.

Pattern: Specialist panelRuns today6 decisionsDesigned for 4 to 80 agents