Free AI Basics v2 Student Version: AI-at-work Mini Simulator

0. How to use this free starter

What this is: An introductory AI-at-work experience. Scenario choices help you practice safety, tool fit, prompting, and verification.

Why it matters: Strong external courses do more than explain concepts: they place AI inside office tools, cases, exercises, and badges. This lesson borrows those formats while keeping the AIAGES evidence boundary.

What you produce: 12 proof cards plus a modest starter certificate.

Do not: Do not treat it as a professional certification, job-readiness proof, or L5 mastery.

1. Mini simulator modules

Module 1/12 · Learn → Try → Prove

AI is an assistant, not an authority

Inspired by: Elements of AI / Google AI Essentials

Learn: AI can generate useful drafts and patterns, but it does not own truth, approval, or responsibility.

Scenario: A learner asks AI: “Does our refund policy allow this exception?” AI replies confidently: “Yes, policy allows it.”

Try: What is the safest next action?

A. Check the official policy/source or ask the owner before using the answer. B. Use the AI answer because it sounds specific. C. Ask AI to confirm again and treat the second answer as proof.

Module 2/12 · Learn → Try → Prove

Match AI to the right work task

Inspired by: Google AI Essentials / Microsoft Work Smarter with AI

Learn: AI is strongest for draft, summarize, classify, compare, brainstorm, and check; weaker for final approval or accountable decisions.

Scenario: Your manager gives you messy notes from a workshop and asks for next steps by tomorrow.

Try: Which AI help is appropriate?

A. Ask AI to extract action items, owners, and unknowns, then manually verify. B. Ask AI to decide the project priority and send it to the team. C. Ask AI to invent missing owner names to complete the list.

Module 3/12 · Learn → Try → Prove

Decisions AI must not make for you

Inspired by: UNESCO / Law Society / HMG Framework

Learn: Legal, medical, financial, HR, customer promises, budget approvals, and safety-sensitive decisions require human authority and source review.

Scenario: A customer complaint contains personal data and asks whether the company will compensate them.

Try: What should the learner do with AI?

A. Remove sensitive details, draft a bounded internal summary, and escalate compensation decision. B. Paste the full complaint and ask AI to decide compensation. C. Ask AI to write a final promise to the customer.

Module 4/12 · Learn → Try → Prove

Choose the right tool for the work context

Inspired by: Microsoft Copilot / Gemini Workspace / Claude for Work

Learn: Tool choice should follow where the work lives: Office stack, Google Workspace, or long-document analysis.

Scenario: Pick the best tool carrier for each task: A) Teams meeting recap, B) Gmail reply draft, C) compare a 25-page policy with a checklist.

Try: Which mapping is best?

A. A Copilot, B Gemini, C Claude. B. A Claude, B Copilot, C Gemini because all tools are identical. C. Use Zapier for all three as the first AI starter tool.

Module 5/12 · Learn → Try → Prove

Build a structured prompt

Inspired by: Microsoft goal/context/source/expectation + OpenAI structured outputs

Learn: A reliable prompt includes goal, context, source/material, output format, constraints, and what not to do.

Scenario: Bad prompt: “Make this better.” You need a meeting follow-up list from notes.

Try: Which prompt is strongest?

A. Using the notes below, extract action items as a table: owner/task/due date/unknowns. Do not add facts not in notes. B. Make this more professional and complete. C. Write the final team plan and fill any missing details.

Module 6/12 · Learn → Try → Prove

Repair weak or unsafe AI output

Inspired by: Anthropic Interactive Prompt Tutorial / OpenAI iteration guidance

Learn: Good AI use is iterative: inspect output, identify unsupported parts, correct the prompt, and rerun or manually edit.

Scenario: AI summary says “All attendees approved the launch,” but the notes only say “Maya will ask Finance.”

Try: What repair instruction is best?

A. Remove unsupported approval claims; mark Finance approval as pending/unknown unless source says otherwise. B. Make the approval sentence sound less risky but keep it. C. Ask AI to justify why approval probably happened.

Module 7/12 · Learn → Try → Prove

Check sources and evidence

Inspired by: Anthropic/OpenAI structured outputs + AIAGES evidence chain

Learn: Every claim that matters needs a source, a caveat, or an unknown marker.

Scenario: Source says: “3 of 12 students missed onboarding because the calendar invite was unclear.” AI writes: “Most students failed onboarding due to calendar problems.”

Try: Which rewrite is evidence-safe?

A. 3/12 missed onboarding; unclear calendar invite is one observed issue, not proof of the main cause. B. Most students failed because calendar systems are confusing. C. Calendar invite clarity is the root cause of onboarding failure.

Module 8/12 · Learn → Try → Prove

Everyday office mini simulation

Inspired by: Microsoft Learn Scenario Library / Google Workspace guides

Learn: A useful AI workflow has a work situation, input, prompt, output, and human check.

Scenario: You have meeting notes and need an email follow-up. Notes: “Alex drafts slides. Priya checks data. Deadline Friday. Budget approval pending.”

Try: Which output is safest?

A. Draft email listing Alex/Priya tasks, Friday deadline, and budget approval as pending. B. Draft email saying budget is approved and launch is confirmed. C. Skip source notes and ask AI to create a full project plan.

Module 9/12 · Learn → Try → Prove

Responsible AI checkpoints

Inspired by: Salesforce Responsible AI / UNESCO / HMG

Learn: Before using AI, check data sensitivity, bias, permission, source reliability, and whether review/escalation is needed.

Scenario: You want AI to summarize student support cases that include names, grades, and personal issues.

Try: What is the right first move?

A. Anonymize/minimize data and confirm permission before using AI; keep sensitive details out. B. Paste everything because the goal is educational. C. Ask AI to remove bias after seeing all personal data.

Module 10/12 · Learn → Try → Prove

Create a starter proof card

Inspired by: IBM SkillsBuild / Salesforce Trailhead / Microsoft badges

Learn: A useful certificate should show what you practiced without claiming professional mastery.

Scenario: You completed the starter. What should your proof say?

Try: Choose the honest certificate claim.

A. I completed an introductory AI-at-work starter: tool choice, structured prompt, source check, and safety boundary. B. I am certified as an AI workplace professional. C. I am job-ready for AI operations roles.

Module 11/12 · Learn → Try → Prove

Spreadsheet and data patterns: do not invent numbers

Inspired by: Microsoft Excel Copilot / Google Sheets / Ben Collins spreadsheet AI

Learn: AI can help classify rows, spot patterns, and propose formulas, but it must not invent missing numbers or treat weak signals as statistical proof.

Scenario: A small table shows 8 support tickets: 3 login issues, 2 billing questions, 2 unclear onboarding notes, 1 feature request. AI writes: “Billing is the main problem affecting most users.”

Try: Which response is safest?

A. Classify visible ticket types and say login is largest in this small sample; do not generalize to most users. B. Accept AI’s billing conclusion because it sounds analytical. C. Ask AI to estimate the missing company-wide ticket totals.

Module 12/12 · Learn → Try → Prove

Responsible-use gate before the certificate

Inspired by: Salesforce Trailhead / HMG approval gates / UNESCO human agency

Learn: Before claiming completion, learners should show they know when to proceed, ask permission, or stop.

Scenario: Sort three cases: public course notes, private customer record, unsupported AI claim with fake source.

Try: Which sorting is correct?

A. Public notes: okay to use; private record: ask permission/anonymize; fake source: stop and repair. B. Use all three because this is only practice. C. Let AI decide which records are sensitive and continue.

2. Starter proof card + certificate

I understand: this certificate only shows completion of an introductory AI-at-work experience. It does not mean job readiness, L5 mastery, or professional certification.

Complete 12/12 proof cards and confirm the boundary statement to generate the certificate.

Render Report

renderer_mode: free_ai_basics_v2_mini_simulator
modules: 12
interaction_pattern: Learn -> Try -> Prove
radio_inputs: 36
textareas: 0
tools: Copilot / Gemini / Claude
external_course_strengths: scenario_cases / tool_situated_tasks / prompt_lab / output_repair / source_verification / responsible_use / progress_proof
certificate_boundary: starter_completion_only

AI-at-work Mini Simulator: a 30–45 minute starter