CLEAREDFORAI
Module 1
Ground School: What AI Is and Isn't
Cleared for AI · about 86 minutes · 7 lessons
Lesson 1 · ~12 min
The Automation Paradox
Lesson 1
The Automation Paradox
The more capable the tool, the more the supervising human matters.
Lesson 1
Complacency is predictable
Usually-right output makes people stop checking. Plan for it.
Lesson 1
Brief · Monitor · Take over
The three crew habits with automation, applied to AI.
Lesson 1
Responsibility doesn't transfer
The AI doesn't sign the logbook. You do.
Lesson 1 · Key points
Key points
- Capable automation shifts the human job from doing to briefing, monitoring and intervening.
- Complacency is a predictable response to a usually-right system. Plan for it rather than blaming people.
- Brief it, monitor it, be ready to take over.
- Accountability stays with the person and company that send the work.
Lesson 2 · ~10 min
What AI Actually Is (and Isn't)
Lesson 2
What AI actually is
A prediction engine trained on text, not a database of facts.
Lesson 2
Strong at
Drafting · summarizing · brainstorming · explaining · reformatting
Lesson 2
Isn't
An oracle · a person · automatically private · a fixed thing
Lesson 2
Chat → Workflow → Agent
Hand flying → flight director → autopilot
Lesson 2 · Key points
Key points
- Language models predict likely text. They're fluent drafters, not sources of truth.
- Fluency isn't accuracy. Confident wording proves nothing.
- Give the AI your real information and tell it to use only that.
- Chat, then workflow, then agent: more autonomy means more guardrails.
Lesson 3 · ~11 min
The Jagged Frontier
Lesson 3
The Jagged Frontier
AI's operating envelope has irregular, invisible edges.
Lesson 3
Fails convincingly
The danger isn't obvious errors. It's persuasive wrong answers.
Lesson 3
Map it like an envelope
Try · grade green / amber / red · share · revisit
Lesson 3
Stakes set the check
Brainstorm = glance. Price, contract, safety = line-by-line.
Lesson 3 · Key points
Key points
- AI's capability edge is jagged: excellent and poor tasks can look identical.
- Outside the frontier, AI fails convincingly, and people tend to accept persuasive wrong answers.
- Map your team's tasks green, amber or red with real examples, and revisit the map often.
- The higher the stakes, the stronger the check.
Lesson 4 · ~11 min
Confident and Wrong: Hallucinations
Lesson 4
Confident and wrong
Plausible text is sometimes false text.
Lesson 4
Watch for
Invented sources · wrong specifics · filled gaps · omissions · maths slips
Lesson 4
The cross-check
Check specifics · open links · compare to source · maths in a calculator
Lesson 4
Same sensor problem
AI checking AI isn't independent verification.
Lesson 4 · Key points
Key points
- Hallucinations are plausible but false outputs, including invented sources, wrong specifics and quiet omissions.
- Check every specific, open every link, and compare important summaries with the source.
- Brief to reduce errors: give source material, allow 'I don't know,' and break tasks into checkable steps.
- A second AI isn't independent verification. Check against a trusted source.
Lesson 5 · ~14 min
Learning Your Aircraft: Hands-On Hours
Lesson 5
Learning your aircraft
Confidence is earned in hours, not declared.
Lesson 5
Knowledge ≠ Ability
ADKAR: practice turns understanding into habit.
Lesson 5
Type-rating plan
Ground · Simulator · Supervised line · Debrief
Lesson 5
Keep a logbook
Task · prompt · result · G/A/R · better prompt
Lesson 5 · Key points
Key points
- Competence comes from hands-on hours on real tasks, not from reading.
- Knowledge isn't ability (Prosci ADKAR). Practice closes the gap.
- Use a ground → simulator → supervised line → debrief structure over about a month.
- Converse with the AI rather than asking one-shot questions, and keep a team logbook.
Lesson 6 · ~10 min
Where Your Data Goes
Lesson 6
Where your data goes
Information is cargo. Know what you're loading.
Lesson 6
Read the terms
Consumer and business plans differ. Know what's connected.
Lesson 6
Traffic lights
Green: fine · Amber: approved tools · Red: don't paste
Lesson 6
Prompt injection
Outside content can carry instructions. Keep a human on the switch.
Lesson 6 · Key points
Key points
- Data shared with an outside tool can't be unshared. Treat it as cargo you must account for.
- Read the terms for your specific tool and plan. Business plans often differ from consumer ones.
- Information traffic lights: green is fine, amber is approved tools only, red is don't paste. When in doubt, strip it out.
- Prompt injection: keep humans approving any AI action that sends, pays, deletes or shares.
Lesson 7 · ~10 min
Train Judgment, Not Buttons
Lesson 7
Train judgment, not buttons
Tools change monthly. Judgment transfers.
Lesson 7
The half-life problem
Prosci: tool skills fade in about 3–4 months.
Lesson 7
Durable skills
Sort · Brief · Verify · Protect data · Escalate
Lesson 7
Recurrent training
Monthly near misses · quarterly map review · differences training
Lesson 7 · Key points
Key points
- Assume today's AI is the weakest you'll use. Plan for constant change.
- Tool-specific skills fade in months (Prosci). Invest in durable judgment.
- Durable skills: task sorting, briefing, verification, data discipline, escalation.
- Schedule recurrent training and share near misses without blame. Leaders reinforce the habits.
Wrap-up
Next: Module 2, the AI Preflight Checklist
Brief it · Monitor it · Verify it · Protect the data · Keep training
Prepared with AI assistance and reviewed by Nils Roed. No ROI promises.