Rehearse a regulatory inspection against your own evidence

For: Chief compliance officer or quality head preparing for a regulator or certification audit

Pattern: Adversarial reviewNeeds scaleDesigned for 20 to 600 agents

The pain today

Readiness is a self-assessment where owners tick 'compliant'. The first person to test whether the evidence really shows it is the inspector, and findings then come with deadlines and sometimes fines.

The ask

I attached the requirements we will be inspected against, our self-assessment and the evidence folders the owners pointed to. For each requirement, make our best case from the evidence, then challenge it the way an inspector would, and tell me which claims would not survive and why.

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

  • Inspection standard or regulatory requirements
  • Self-assessment with claimed status per requirement
  • Evidence folders: procedures, records, logs, minutes
  • Findings from earlier inspections and audits

The unit of work

One worker task per one requirement with its claimed evidence.

Why a swarm fits

Each requirement is argued from its own evidence folder, and a full standard has very many of them. Defending and attacking the same claim with different model families exposes evidence that only looks sufficient.

Not for

A narrow inspection of one process with a handful of requirements: a walk-through with the process owner is better.

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

    Does the folder the owner pointed to contain any dated record from the inspection period for this requirement?

    Sees only: The folder index with dates and the requirement

    Why: An empty folder is a finding by itself and needs no debate.

    • Yes: 0.60 or higherthenAccept
    • Unsure: 0.30 up to 0.60thenAccept
    • No: below 0.30thenMark unresolved
  2. After workers, the judge checks

    Evidence checkYes or no, with a probability

    After a worker answers

    Do the quoted records show the required activity being performed, with dates and names, rather than a procedure saying it should be?

    Sees only: The requirement and the defence's quoted records

    Why: Inspectors test performance, not paperwork about performance.

    • 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

    Has the challenger named a specific record an inspector would ask for that is absent from the folder index?

    Sees only: The challenge, the folder index and the requirement

    Why: A challenge must say what evidence would have satisfied it.

    • Yes: 0.80 or higherthenMark unresolved
    • Unsure: 0.50 up to 0.80thenEscalate to a strong model
    • No: below 0.50thenSkip this unit
  4. Reconciler, while merging

    Conflict checkYes or no, with a probability

    While reconciling

    Does an earlier inspection finding on this requirement lack a documented closure in the folder?

    Sees only: The earlier finding and the closure records retrieved

    Why: Repeat findings draw the heaviest response from a regulator.

    • Yes: 0.60 or higherthenMark unresolved
    • Unsure: 0.30 up to 0.60thenMark unresolved
    • No: below 0.30thenAccept
  5. Run control, between rounds

    Another round?Yes or no, with a probability

    Between rounds

    Did the last round surface evidence or a gap not already cited for this requirement?

    Sees only: The citations of the last round

    Why: Ends the rehearsal on a requirement once nothing new appears.

    • Yes: 0.60 or higherthenContinue
    • Unsure: 0.30 up to 0.60thenStop
    • No: below 0.30thenStop
  6. Accountable person, before anything is settled

    Person decidesYes or no, with a probability

    Before anything is reported as settled

    Would the status of this requirement be downgraded from the owner's self-assessment?

    Sees only: The self-assessed status and the rehearsal outcome

    Why: The compliance officer decides what to remediate or disclose.

    Accountable: The compliance officer decides what to remediate, what to disclose proactively and who speaks to the inspector.

    • 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 standard into single requirements and names what an inspector would ask to see for each.

    Decisions here:1. Scope check

  2. Workers

    Small fast workers build the compliance case; workers from another family play the inspector and probe for gaps.

    Designed for 20 to 600 agents, one worker task per one requirement with its claimed evidence. Each worker receives only its own unit.

  3. Judge, from a different model family

    A decision model from a third family rules whether the claim survives, on the quoted evidence only.

    Decisions here:2. Evidence check3. Evidence check

  4. Reconciler

    A strong reasoning model groups failing claims by root cause and owner and orders them by inspection exposure.

    Decisions here:4. Conflict check5. Another round?

  5. Accountable person

    The compliance officer decides what to remediate, what to disclose proactively and who speaks to the inspector.

    Decisions here:6. Person decides

Checked before anything is accepted

  • A claim survives only with dated records showing the requirement was performed, not just a procedure
  • Evidence dates must fall inside the inspection period
  • Earlier findings are checked for documented closure
  • The challenger must state what evidence would have satisfied it

What comes back

  • Requirements whose evidence holds, with the evidence indexed
  • Claims that would not survive and what is missing
  • Repeat findings from earlier inspections still open
  • Remediation list per owner

What to measure

  • Inspector findings the rehearsal had predicted
  • Claims downgraded from the self-assessment
  • Preparation hours per requirement

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

Map a new regulation to policies and controls, rule by rule

For: Head of compliance or regulatory change manager implementing a new rule

A new regulation arrives with obligations buried in articles, annexes and guidance.

Pattern: Hierarchical decompositionNeeds live models6 decisionsDesigned for 8 to 400 agents

Test control evidence samples against the control description

For: Internal control manager or second-line tester running the annual control testing cycle

Every key control needs sampled evidence checked: was the approval there, by the right person, before the event, for the right amount.

Pattern: Map, verify, reduceNeeds scale6 decisionsDesigned for 20 to 800 agents