When to Ignore This Toolkit¶

Permission Slip

The Hard Truth¶

This toolkit contains thousands of words across dozens of documents. If you actually did everything in here for every project, you'd never ship anything.

The toolkit is a menu, not a mandate. This document helps you figure out what to skip.

Situations Where You Should Ignore Most of This¶

1. You're Running a Quick Experiment¶

Skip: Full business case, comprehensive stakeholder analysis, detailed risk register, governance framework

Keep: Basic ethics check, data privacy sanity check, clear success/failure criteria

Why: Experiments are supposed to be cheap and fast. If the governance costs more than the experiment, you've defeated the purpose.

The test: Can you kill this in a week with zero political fallout? If yes, keep it light.

2. You're Using Off-the-Shelf AI for Low-Stakes Tasks¶

Skip: Custom model documentation, bias testing frameworks, MLOps architecture

Keep: Vendor security check, data handling review, user training

Why: You're not building AI, you're buying software. Treat it like software procurement, not a moon mission.

Examples: - Meeting transcription - Document search - Internal chatbot for IT help - Email categorization

3. You Have a 2-Week Window and Executive Air Cover¶

Skip: Extended stakeholder consultation, phased rollout plans, comprehensive change management

Keep: Whatever prevents immediate disaster, whatever you can actually do

Why: Sometimes speed is the feature. If you have genuine executive support and a narrow window, move fast and document later.

The catch: This only works if you actually have the air cover you think you have. Get it in writing.

4. The Problem Is Obvious and the Solution Is Simple¶

Skip: Extensive options analysis, weighted decision matrices, scenario planning

Keep: Sanity check with one trusted colleague, basic documentation of what and why

Why: Analysis paralysis is real. Sometimes the first reasonable option is the right option.

Warning sign that you've over-complicated: You're on your third workshop and still "exploring options."

5. Everyone Already Agrees¶

Skip: Elaborate stakeholder engagement plans, change resistance frameworks, coalition building

Keep: Brief confirmation that agreement is real (not just silence), simple communication of what's happening

Why: Consensus is rare and valuable. Don't manufacture obstacles by treating it as suspicious.

Test: Ask the quietest person in the room what they really think. If they agree too, you're probably fine.

6. You're the Expert and You Know It¶

Skip: Extensive research into "how others have done it," beginner-level guidance, prescriptive templates

Keep: Peer review from someone at your level, documentation for others who'll inherit this

Why: Templates are training wheels. If you've done this twenty times, you don't need the checklist.

Humility check: When did you last learn something that changed your mind? If it's been a while, you might be less expert than you think.

Situations Where You Should NOT Ignore This¶

Never Skip These¶

The "Would I Be Comfortable" Tests¶

Before skipping anything, ask:

1. "Would I be comfortable if my skip decision was audited?"

2. If no → do the thing

3. "Would I be comfortable explaining this to a journalist?"

4. If no → do the thing

5. "Would I be comfortable if I was the person affected by this AI?"

6. If no → do the thing

7. "Would I be comfortable if my skip decision was in an ANAO report?"

8. If no → you definitely need to do the thing

The Real Framework for What to Skip¶

The 2x2 That Actually Matters¶

Questions to Place Your Project¶

Stakes: - Who gets hurt if this goes wrong? - What's the blast radius of failure? - Is this in the minister's visibility? - Does this affect vulnerable people?

Reversibility: - Can we turn it off? - Can we fix mistakes after the fact? - Will people know if we pivot? - Is there a point of no return?

How to Skip Things Professionally¶

If you're going to skip parts of the recommended process, do it explicitly:

The "Conscious Skip" Template¶

## Process Element Skipped
[What you're not doing]

## Rationale
[Why it's appropriate to skip in this context]

## Risk Accepted
[What could go wrong as a result]

## Mitigation
[What you're doing instead, or why no mitigation is needed]

## Decision Maker
[Who made this call - name, not just role]

## Review Trigger
[What would cause you to revisit this decision]

Example¶

## Process Element Skipped
Comprehensive stakeholder analysis and engagement plan

## Rationale
Internal tool affecting only 12 users in one team, all of whom
were involved in requesting this solution. No external impact.

## Risk Accepted
May miss concerns from adjacent teams; could affect future scaling

## Mitigation
Will do lightweight check-in with IT and security. Will revisit
stakeholder analysis if we decide to scale beyond pilot team.

## Decision Maker
Jane Smith, Project Lead

## Review Trigger
Any request to expand beyond pilot team; any complaint from
users; any integration with other systems

The Meta-Point¶

The goal of this toolkit is better AI outcomes, not more documentation.

If following the toolkit is preventing good outcomes, the toolkit is failing at its job.

Use your judgment. Document why. Be prepared to explain.

That's what professional practice actually looks like.

Signs You're Using This Document Wrong¶

• You're skipping ethics review because "it's obvious"
• You're skipping stakeholder engagement because "they'll just slow us down"
• You're skipping risk assessment because "we're agile"
• You're citing this document to justify something you know is sketchy
• You're skipping things because the template was annoying, not because it was unnecessary

If any of these apply, go back and do the thing you were trying to skip.

"Rules are for the guidance of wise people and the obedience of fools." - Generally attributed to Douglas Bader

Use this document wisely.