Appendix C: Implementation Checklists

Print these. Pin them to your wall. Use them until the steps become automatic. For visual decision flowcharts that guide you through each framework step by step, see Appendix D.

---

Checklist 1: Your First AI Workflow (Chapter 7)

Complete this checklist to build your first AI-assisted workflow in under an hour.

Before You Start

• Identified a task you do at least 3 times per week
• Scored it with the FFCC Scorecard (Appendix A)—confirm it’s above 100
• Confirmed the task has clear inputs and measurable outputs
• Identified who reviews the output before it’s used

Design the Workflow

• Defined the trigger—what event starts this workflow?
• Defined the input—what specific information does AI receive?
• Defined the AI processing step—what does AI do with the input?
• Defined the human review step—what does the human check?
• Defined the action—what happens after human approval?

Build and Test

• Written the prompt (use Appendix B templates as a starting point)
• Tested with 3 real examples from the past week
• Noted where AI output needed correction
• Refined the prompt based on those corrections
• Tested 3 more examples with the refined prompt

Go Live

• Set review level to Deep Review for the first two weeks
• Documented the workflow (trigger, input, AI processing prompt, human review criteria, action)
• Scheduled a two-week check-in to evaluate quality and adjust

---

Checklist 2: Permission Assessment (Chapter 11)

Use this before expanding AI into any new area of your work.

Benefit Assessment

• Named the specific, measurable benefit (not “efficiency” or “saves time”)
• Estimated the time saved per occurrence: _____ minutes
• Estimated the frequency: _____ times per week
• Calculated total weekly time saved: _____ hours
• Confirmed this benefit justifies the setup and review overhead

Exposure Assessment

• Listed all data the AI will see
• Identified any sensitive or confidential information in the input
• Identified who could be affected by errors (internal team, customers, public)
• Determined the blast radius of a mistake (one person, one team, all customers)
• Confirmed data handling meets your organization’s policies

Failure Mode Assessment

• Named the 3 most likely failure modes
• For each: determined if the error is recoverable or permanent
• For each: documented the recovery plan
• Identified who would detect the error and how quickly
• Confirmed you have a kill switch (can stop the workflow immediately)

Decision

• All three assessments complete
• Proceed / Proceed with safeguards / Pilot first / Decline (circle one)
• If safeguards: named the specific safeguards (extra review, monitoring, kill switch)
• If pilot: defined the limited scope and what success looks like before broader rollout
• If decline: documented why and who needs to approve an exception

---

Checklist 3: Review Calibration (Chapter 14)

Use this monthly to ensure your review levels match actual risk.

For Each Active AI Workflow:

• Workflow name: _____________________________
• Current review level: Scan / Spot-Check / Deep Review / Rewrite
• Error rate in the past 30 days: _____ errors out of _____ outputs
• Most serious error: _____________________________
• Was the error caught before it reached its audience? Yes / No

Calibration Decision

• Error rate below 5% with no serious errors → Consider moving to lighter review
• Error rate 5–15% → Maintain current level, refine the prompt
• Error rate above 15% → Move to heavier review or redesign the workflow
• Any uncaught error that reached audience → Move to Deep Review immediately

Review Level Transitions

Moving lighter (e.g., Deep Review → Spot-Check):

• At least 30 days at current level
• Error rate consistently below 5%
• No uncaught errors reaching audience
• Documented what you’ll spot-check (the 2–3 things most likely to go wrong)

Moving heavier (e.g., Spot-Check → Deep Review):

• Trigger identified (uncaught error, new use case, model update)
• Updated review criteria documented
• Team notified of the change and the reason

---

Checklist 4: Team Rollout (Chapter 18)

Use this when introducing AI workflows to your team for the first time.

Preparation (Week -2 to -1)

• Selected 1–2 workflows that scored highest on FFCC for the team
• Tested both workflows yourself for at least 2 weeks
• Documented clear instructions with examples of good and bad AI output
• Prepared talking points addressing the top 3 concerns your team will have
• Identified 1–2 early adopters who can be peer champions

Launch (Week 1)

• Communicated the change (use the Team Rollout Communication template in Appendix B)
• Led a live demo with a real example—not a slide deck
• Gave everyone a chance to try the workflow during the meeting
• Set the review level to Deep Review for all team members for week 1
• Established a channel for questions and feedback (Slack, email, whatever your team uses)

First Month (Weeks 2–4)

• Checked in individually with each team member after their first 5 uses
• Collected and shared examples of good catches during review
• Addressed any workflow adjustments based on team feedback
• Documented common mistakes and added them to the review criteria
• Celebrated a specific win that the workflow enabled (make it visible)

Steady State (Month 2+)

• Ran the Review Calibration checklist (above) for each team member
• Calibrated review levels individually (not everyone moves at the same pace)
• Identified the next workflow to introduce based on team readiness
• Updated the team documentation with lessons learned
• Scheduled quarterly review of all active workflows

---

Checklist 5: Monthly AI System Maintenance (Chapter 28)

Keep your personal AI system sharp. Takes 30 minutes once a month.

Master Prompt Review

• Glanced through your Priorities section—does it still reflect what matters right now? (The full-document pass happens quarterly)
• Removed any context that’s no longer current (finished projects, old priorities)
• Retired any instruction you’ve given twice into the prompt (Evolution Protocol)
• Noted any task setup you’ve assembled 3+ times—that’s a Task Library variant asking to exist

Workflow Audit

• List all active AI workflows
• For each: Is it still providing value? (If not, retire it)
• For each: Has the error rate changed? (If so, recalibrate review)
• For each: Can it be improved based on what you’ve learned?

Skills Check

• Tried one new AI capability this month (new model, new feature, new use case)
• Read or watched one resource about AI developments in your industry
• Identified one task you’re still doing manually that AI could assist with

Documentation

• Updated your workflow documentation with any changes
• Noted any lessons learned worth sharing with your team
• Archived any retired workflows for future reference