AIR-F · General Foundation

Get ready before AI gets ahead of you.

AIR-F is the foundational level of the AI Readiness (AIR) Framework — for everyone in an enterprise, not just the technical team. It builds the cognitive, operational, and ethical preparedness to adopt AI safely, so you stay independent and empowered instead of afraid.

E8 · Public Trust & ConfidenceAI Nation 2030 · Human CapitalAIGE-aligned
Cognitive delegation — knowing what to hand to smart systems. Operational governance — knowing how to supervise them so nothing goes off the rails.

Three goals, three outcomes

1
Eliminate AI Fear
Stop thinking AI will take over your job. See it as a helpful assistant, not a threat — and learn the basics so you feel in control.
2
Establish Safe Guardrails
Set clear rules for what AI can and cannot do. Protect private data and company secrets. Check AI work for mistakes before you use it.
3
Achieve Basic Productivity Gains
Use AI for small, boring tasks — emails, lists, summaries — and free up time for genuinely human work. Proven, and done safely.
How each module maps to a goalModule 01 → Goal 1 (Eliminate fear) · Module 02 → Goal 2 (Guardrails) · Module 03 → Goal 3 (Productivity) · Module 04 → Goal 2 (the "check before you use" half). Every module ends with a hands-on exercise you do yourself — not just read.
Interactive · Your readiness check

Where do you stand today?

Five quick questions. Your score maps to the AIR maturity path — Aware → Capable → Autonomous — so you know where you start and what to strengthen.
1. When you think about AI in your work, you mostly feel…
It's a threat to my jobUnsure, I don't fully get itIn control, or I want to be
2. Do you know what data is safe to paste into an AI tool?
No ideaSome rules, some guessesYes — clear rules I follow
3. Do you check AI output for mistakes before using it?
Never / rarelySometimesAlways, it's a habit
4. Can you write a structured prompt to get a useful result?
Not reallyBasic prompts onlyYes — context, role, constraints
5. Do you already use AI for small daily tasks?
NoOccasionallyYes, most days
0 / 100
Your readiness profile
Not started
Answer the five questions to see where you start.
Scoring is a prototype self-assessment, computed locally from your answers — not a validated instrument. Replace with a psychometrically-tested rubric before production.
Module 01 · AI Mental Models

It's a reasoning engine, not a search engine.

Before you can supervise AI, you have to know what you're actually supervising. This module replaces fear with a working mental model.

Goal 1 · Eliminate AI FearE8 · Public Trust

What an LLM actually does

Large language models generate text by predicting the next token, given everything before it. That's a probabilistic reasoning engine — not a lookup table, not a search engine, and not a mind.

Mental model shift"Why is it lying?" → "I need to verify this." Lying implies intent. A model that predicts probable tokens has no intent to deceive — it has a tendency to be plausibly wrong.

The three trigger types

  • Direct — you ask; it answers.
  • Event-driven — a system prompt triggers it when a condition is met.
  • Ambient — it watches context in the background and acts on its own.
Hands-on · Goal 1

Which mode is the AI in?

Read each scenario and pick what the AI is really doing. This is the reframe that kills the fear: once you can name it, you can judge it.

"It wrote a confident answer about my industry that was mostly right but had one wrong number."
"I asked it the same question twice and got two slightly different answers."
"It gave me a link to a page and quoted it exactly."
Score: 0 / 3
Your result is computed live from your clicks — real data about your understanding, not a demo metric.
Module 02 · Enterprise Safety & Data Hygiene

Know what not to paste.

The compliance floor of AIR-F. This is the module that protects you, your data, and the people whose data you hold.

Goal 2 · Safe GuardrailsE8 · Privacy & Safety

The single rule that covers most of it

Treat anything you paste into an AI tool as published to an audience you don't control — unless your tool is explicitly approved and the data is cleared to leave your environment.

Never pastePasswords, API keys, IDs, bank details, customer PII, internal financials, unpublished strategy, proprietary source code, and anything subject to confidentiality.
Usually fineYour own public content, generic questions, de-identified examples, and material you'd happily post in public.
Ask firstAnything from a client or employer, mixed datasets, and "cleaned" data that could still be re-identified.
Submodule S1

Catastrophic Risk × Red-Teaming (entry level)

When a paste isn't just your data — it's a lever. Open the standalone submodule for the fused lesson + red-team exercise.

Hands-on · Goal 2

Paste / Don't Paste — quick classifier

Tap each item's classification. Instant feedback, and your score tracks how well you'd protect real data.

A customer list with names, IC numbers and addresses
A generic question: "how do I write a better subject line?"
An internal, unpublished sales forecast
A client's draft report you've been asked to tidy up
Score: 0 / 4
Live score from your clicks. In production this same classifier would be graded and logged against a leakage-risk rubric.
1:1 Coaching

Not sure what counts as "safe to paste" in your role?

Book a short coaching session and we'll audit your actual workflow — the exact files, tools, and copy-paste habits you touch daily — and give you a personal paste/don't-paste map.

Book a session →
Module 03 · Fundamental Prompting

Delegate and refine, don't just copy-paste.

Prompting isn't magic words. It's giving the model enough context and constraints to produce something you can actually judge — and that's how you win real time back.

Goal 3 · Productivity GainsE8 · Responsible use

CREATOR method

  • Context — who you are, what you need it for
  • Role — what perspective the model should take
  • Examples — one or two, when useful
  • Action — the specific deliverable
  • Tone & format — style, length, structure
  • Outcome — what "good" looks like
  • Restrictions — boundaries and "don't do X"

Chain-of-Thought & boundaries

Ask the model to reason step by step before answering, and state what you don't want explicitly ("Do not invent citations"). Zero-shot and few-shot let you dial how much steering it needs.

Hands-on · Goal 3

CREATOR prompt builder — live

Fill any fields that apply and watch a structured prompt assemble in real time. The meter shows how "specified" your prompt is.

Your prompt will appear here…
Specificity: 0 of 4 fields
The specificity meter is computed from how many CREATOR fields you complete — a real, live measure of prompt structure.
Module 04 · Verification & Critical Thinking

You are the human in the loop.

The final and non-negotiable layer: nothing AI produces goes out the door unchecked. Verification is a habit, not an afterthought.

Goal 2 · Check before you useE8 · Trust

Spotting hallucinations

  • Confident, specific claims with no traceable source
  • Invented citations, papers, or people
  • Numbers that are suspiciously precise
  • Answers that shift when you re-ask slightly differently

Three-step verification checklist

  1. Trace — can I find the claim in a real source?
  2. Cross-check — does a second, independent source agree?
  3. Sanity-check — does it contradict what I already know to be true?
The obligationVerification isn't optional for work you sign off on. The model drafts; you author.
Submodule S2

Catastrophic Risk × Evals (entry level)

One wrong answer is a bug. A wrong answer at scale is a failure mode. Open the standalone submodule for the fused lesson + eval exercise.

Hands-on · Goal 2

Spot the hallucination

One of these claims has a fabricated detail. Find it — this is exactly what slips into real work.

Claim A: "AI-generated output should always be reviewed by a human before use."
Claim B: "According to the 2024 Global AI Audit (Mercer et al.), 98.4% of all AI outputs contain at least one factual error."
Claim C: "Models generate the next token based on the tokens before it."
Score: 0 / 3
The fabricated claim (B) is a classic tell: a fake citation + an absurdly precise statistic. Your score is computed live from your picks.
1:1 Coaching

Want a second pair of eyes on your verification habit?

Send us three real AI outputs you rely on, and in one session we'll pressure-test them together — catching what slips past you and hardening your checklist.

Book a session →
Lab F1 · Safety

Paste / Don't Paste.

Classify five work-ish snippets, justify each, and see the leakage consequence before you'd have made it real.

Participatory Lab

Classify and justify

  1. Take five snippets — a client email, a customer ID list, an internal sales forecast, a public blog draft, an API key.
  2. Mark each Paste / Don't Paste / Ask first.
  3. Write one sentence of justification per item.
  4. For each "Don't Paste," name the concrete consequence of leaking it.
  5. Compare against the Module 02 callouts. Did you catch every red flag?
Lab F2 · Prompting

Rewrite a vague ask into a CREATOR prompt.

One vague request, one structured rewrite, one visible quality jump.

Participatory Lab

Vague → structured

  1. Start with a vague ask — e.g. "help me with my email".
  2. Rewrite it using every CREATOR element that applies.
  3. Run both. Compare output quality, length, and usefulness.
  4. Note which single element made the biggest difference.
Lab F3 · Verify

Catch the hallucination.

Receive a confidently-wrong AI answer, run the checklist, and flag the error you'd otherwise have published.

Participatory Lab

Run the three-step check

  1. Take a plausible-sounding AI answer with one fabricated citation or number buried in it.
  2. Run Trace → Cross-check → Sanity-check on the suspicious claim.
  3. Flag exactly what fails and where.
  4. Rewrite the answer with the error corrected and the source noted.
1:1 Coaching

Caught one? Bring it to coaching.

Send us the hallucination you found and your corrected version — we'll show you how to build a reusable eval so it never slips through again.

Book a session →
AIR-F → E8 · Foundation Edition (USD 300) · AI Readiness Institute · Source: yasmin-syllabus.md · All benchmarks & scores are illustrative prototype values computed locally from your input — not sourced facts. Not for public distribution until CLO compliance + Sha review clear it for production.