Ld Compliance Training

---
name: ld-compliance-training
description: Plans a compliance programme that changes choices on the job, producing the few decisions that matter per rule, scenario checks with feedback, a refresh cadence and a record-use statement, with legal points marked "check with a qualified adviser". Use for "run ld-compliance-training", "compliance training plan", "mandatory training", "annual compliance refresh", "compliance quiz", "make compliance training less boring", "harassment prevention training", "compliance records", part of the AI for L&D and Corporate Training Pack by Polar Bear.
---

# Compliance Training

## When To Use
Compliance training is a box to tick: everyone clicks through the annual module, passes a quiz an AI browser could pass alone, and the quiz record is kept as evidence against people if something goes wrong. Use this to plan a mandatory programme, or a refresh of one. It answers one question: on the job, which few decisions does each rule turn on, and how do people practise them?

## When Not To Use
If you only need questions for one course, use Knowledge Check. If the question is what the rule requires, what must be recorded or for how long, that is not a training question; check with a qualified adviser before planning anything.

## Inputs
- The policy or rule text, as the compliance or legal owner gave it to you.
- Real cases where people got it wrong or asked for help, anonymised: incidents, helpline themes, audit findings.
- The current module, quiz and how completion is recorded today.
If you have none of this, I start from the policy text and mark the output as a first draft, with every decision "to confirm with the policy owner".

## Approach
Two methods carry it. Merrill's First Principles of Instruction (2002, ETR&D) says learning is promoted by real problems and application, so each rule becomes decisions practised in scenarios, not clauses read on screen. Retrieval practice (Roediger and Karpicke 2006, Psychological Science) found that recalling beat restudying on a delayed test, so the refresh asks people to decide again rather than rewatch. The judgment is to cut every rule down to the moments people get it wrong. The failure it prevents: a rule nobody breaks gets forty slides while the one that causes incidents gets one, and the record proves only that the slides were clicked.

## Workflow
1. Ask at most three questions: which rules are in scope and who owns each, where have people actually got it wrong, and what is the completion record used for today?
2. Per rule, list the few on-the-job decisions where people get it wrong, in the words of the work ("a supplier offers tickets to the final"). Everything else in the rule goes to a look-up reference, not the course. Each decision is confirmed by the policy owner.
3. Write one scenario check per decision: a realistic situation, options built from the real mistakes, and feedback for every option saying why. Retakes are allowed; the check is practice, not a gate on the person.
4. Set the refresh cadence as retrieval at spaced intervals the user sets: short scenario prompts that ask people to decide again, not a rewatch of the full module. Where a rule sets a required frequency, mark it "check with a qualified adviser".
5. Write the record-use statement: the completion record shows that training was offered and completed. It is never used to grade, rank or build a case against a person, and quiz answers are reported by item and in aggregate only, with groups smaller than a size the user sets suppressed. Refuse any request to design quiz records as disciplinary evidence.
6. Mark every legal, regulatory, retention or privacy point "check with a qualified adviser" and list them together for the policy owner.

## Output Format
```markdown
# Compliance Training Plan
Programme: [name] | Rules in scope: [list] | Policy owner: [role] | L&D owner: [role]
## Decisions that matter
| Rule | On-the-job decision | Common mistake | Source of the mistake | Confirmed by |
|---|---|---|---|---|
| [rule] | [decision] | [what people do wrong] | [incident theme, helpline, audit] | [policy owner role] |
## Scenario checks
| Decision | Scenario | Options | Feedback per option |
|---|---|---|---|
| [decision] | [situation] | [options from real mistakes] | [why each is right or wrong] |
## Refresh cadence
| Interval | Format | Decisions covered |
|---|---|---|
| [interval the user sets] | [scenario prompt, micro-unit] | [decisions] |
## Record use
May be used for: [showing training was offered and completed; fixing the course from item results]
Never used for: grading, ranking or building a case against a person.
## For a qualified adviser
- [legal, regulatory, retention or privacy point] (check with a qualified adviser)
## Decision
[Policy owner] confirms the decisions and [L&D lead] and [legal or compliance adviser] sign the record-use statement by [date].
```

## Done When
- Every rule has at least one on-the-job decision confirmed by its owner.
- Every decision has one scenario check with feedback for every option.
- The record-use statement is written and says what the record is never used for.
- Every legal or regulatory point sits in the adviser list, not stated as fact.

## Quality Bar
- Decisions come from real mistakes, never from the order of clauses in the policy.
- No legal requirement, frequency or retention period is stated as advice; each says "check with a qualified adviser".
- Quiz results report by item and in aggregate only; no individual answer goes to a manager.
- The record shows training happened; it is never used to grade, rank or build a case against a person.

## Next
Run ld-training-request-intake (Training Request Intake) to handle the next request from the front door.

## About the makers

This pack is made by Polar Bear, a consultancy built by ex-McKinsey founders with a dream to make AI work for People, not instead of them. We help our clients build people systems and AI-first ways of working, and we run our own company on Claude. If your team has outgrown the self-serve version, message Pauline (linkedin.com/in/paulinebertry).