Beginner · About 39 hours · 3 projects · Free

AI Roadmap for Non-technical

In about 39 hours you use your AI tool well on real work, check what it tells you, keep data safe, and can explain the words in the news to your team.

0of 39 hours · 0%

Saved in this browser. No account, no email.

What you'll build

Go at your own pace. Stop at any milestone: each one leaves you with something to show.

  1. 1
    My weekly AI workflow

    Take one task you do every week and build a prompt that does it well, with a check you run every time.

    hour 15
  2. 2
    A research brief you checked

    Draft a brief with your AI tool, then check every fact yourself and keep what it got wrong.

    hour 24
  3. 3
    How I use AI: a one-page note for my team

    Write the rules you actually follow with AI, in one page your team can copy.

    hour 39

Skills you'll learn

  • Prompting
  • Giving context
  • Checking output
  • Safe data use
  • Fact-checking
  • Source tracing
  • Spotting made-up claims
  • Research with AI
  • Data safety
  • Disclosure
  • Working with agents
  • Judgment

Prerequisites

  • A ChatGPT, Claude, Gemini or Copilot account you are allowed to use at work.
  • One real task you do every week.
  • No coding and no math.

Program outline

3 milestones · 8 sections · 40 lessons · 3 projects · Every course and source is free

Learn as a feed

Milestone 1 · Hours 1 to 15 · 3 sections · 15 lessons

Use the tools well

Learn what these tools do and how to get good results on your real work.

Hours 1 to 50/17

Explain in plain words what ChatGPT, Claude, Gemini or Copilot does when it answers, and name three things it cannot do.

In the Encyclopedia: What is AI?, Artificial Intelligence (AI), Generative AI

  1. What AI is, in one hour60 min

  2. The four words: AI, machine learning, deep learning, generative AI60 min

  3. Elements of AI, chapter 1: what is AI60 min

  4. Narrow AI, AGI, and the headlines55 min

  5. Where AI helps and where it hurts at work: the jagged frontier55 min

Hours 6 to 100/17

Say what a language model does with your words, why the same question gets different answers, and what the context window means for your long chats.

In the Encyclopedia: Large Language Models (LLMs), Text Generation & Sampling, Context Window & Long Context

  1. Large language models, explained briefly60 min

  2. Elements of AI, chapter 4: machine learning without the math60 min

  3. Why the same prompt gives different answers55 min

  4. The context window: what the model can see right now60 min

  5. Explaining ChatGPT to anyone, and this site's 21-stage teardown60 min

Hours 11 to 150/16

Have a reusable prompt for one real weekly task, with role, context, an example and a checking step, and run it twice.

In the Encyclopedia: Prompt Engineering & Using LLMs Effectively, Prompt Engineering, Conversational AI & Chatbots

  1. The prompt formula60 min

  2. The official guides: what OpenAI and Anthropic tell you to do60 min

  3. AI Fluency: delegate, describe, discern, be diligent60 min

    • Read
      Claude Academy: free courses · Anthropic
    • Do
      Start the course 'AI Fluency: Framework and foundations' on this hub. Map your weekly task to the four Ds: what you delegate, how you describe it, how you judge the result, and what you check before using it.

  4. Examples, reasoning, and letting the tool ask you questions60 min

  5. Chatbots and assistants, then the project: run your workflow60 min

Milestone 1 project · about 4 hours

My weekly AI workflow

Take one task you do every week and build a prompt that does it well, with a check you run every time.

  • Prompting
  • Giving context
  • Checking output
  • Safe data use

Courses that teach this

  • Claude 101

    Anthropic · 2.5 hr

    Write clear prompts, refine them, and use Projects and Skills for everyday work.

  • Build foundational generative AI skills

    Microsoft Learn · 2 hr 20 min

    Learn generative AI basics, write effective Copilot prompts, and review output responsibly.

  • OpenAI Academy

    OpenAI

    Short sessions on using ChatGPT in everyday work.

  • Discovering Generative AI

    Udacity · 45 minutes

    Learn core generative AI concepts and how to use it at work.

Build it

  1. Pick one task you do every week, such as a status update or a meeting summary.

  2. Time how long it takes you without AI.

  3. Write a prompt with your role, the context, an example of good output and a checking step.

  4. Run it twice on real work and fix what went wrong.

  5. Write down what you may and may not paste in, and the time it took before and after.

You finish with

A one-page workflow: the task, the prompt, what you may and may not paste in, the checks you run, and the time it took before and after.

Check your work

  • It runs on real work, twice, with a better second result.

  • It includes a check you do every time.

  • You measured the time before and after.

  • You know what you may not paste in.

See how others did it

Milestone 2 · Hours 16 to 24 · 2 sections · 10 lessons

Check the work

Learn why AI makes things up and how to check every answer.

Hours 16 to 200/17

Spot a made-up fact, quote or citation in an AI answer, explain in one sentence why it happens, and run a three-step check before you use an answer.

In the Encyclopedia: AI Hallucinations, Human in the Loop

  1. Why language models hallucinate60 min

  2. What OpenAI and Anthropic say about it55 min

    • Read
      Why language models hallucinate · OpenAI
    • Read
      Reduce hallucinations · Anthropic
    • Do
      Write your answer-checking rule in three steps: allow 'I do not know', ask for exact quotes and where they came from, verify every name, number and date. Apply it to one answer you used recently.

  3. The lawyer who cited fake cases, and the airline that paid for a chatbot's promise60 min

  4. Human in the loop: where you stay in charge55 min

  5. Draft the brief, then mark every claim to check55 min

Hours 21 to 240/18

Finish a two-page research brief where every claim traces to a named source, and tell when an image, voice or statistic came from AI.

In the Encyclopedia: AI Bias & Fairness, Image Generation, Deepfakes & Face Synthesis

  1. Bias: when the answer tilts60 min

  2. Check the brief, claim by claim60 min

  3. Images: what the tool really does when it draws55 min

  4. Deepfakes and voice clones: how to tell, and what to do55 min

  5. Watermarks, provenance, and the finished brief60 min

Milestone 2 project · about 4 hours

A research brief you checked

Draft a brief with your AI tool, then check every fact yourself and keep what it got wrong.

  • Fact-checking
  • Source tracing
  • Spotting made-up claims
  • Research with AI

Courses that teach this

  • AI Fluency: Framework and foundations

    Anthropic · 4 hr

    Apply Delegation, Description, Discernment and Diligence to collaborate with AI and check outputs.

  • Elements of AI

    University of Helsinki · about 30 hours

    What AI can and cannot do, with no math or code, and a free certificate.

Build it

  1. Pick a question your team needs decided soon.

  2. Draft a two-page brief with your AI tool.

  3. Highlight every name, number, date and quote in the draft.

  4. Check each one against a source you open yourself, and correct or remove it.

  5. Keep a list of what the AI got wrong, and add a short note on how you checked.

You finish with

A two-page brief with a source table (claim, source, verified or corrected or removed), a list of what the AI got wrong, and a short method note.

Check your work

  • Every name, number, date and quote traces to a source you opened.

  • You kept a list of what the AI got wrong.

  • A colleague could repeat your checks.

See how others did it

Milestone 3 · Hours 25 to 39 · 3 sections · 15 lessons

Safe, informed, and able to explain it

Learn what is safe to share, what agents are, and how to explain it to others.

Hours 25 to 290/18

Know what your tool does with what you type, which data never goes into a chat, and have a draft of your personal rules.

In the Encyclopedia: AI Safety, AI Policy and Acceptable Use, Data Protection Law and AI (GDPR and beyond)

  1. AI safety, the umbrella word55 min

    • Watch
      AI Safety · Computerphile
    • Read
      What Is AI Safety? | IBM · IBM
    • Do
      Collect ten recent AI headlines. Sort each into accident, misuse, bias, privacy, or hype. Write one line on which of the five your job actually faces.

  2. Shadow AI and the acceptable use policy60 min

  3. Personal data and the law, the plain parts60 min

  4. The model's own rulebook: what it is told to refuse and whose instructions win60 min

    • Read
      OpenAI Model Spec (2026/10/29) · model-spec.openai.com
    • Read
      Anthropic: Claude's character · Anthropic
    • Do
      Open your own tool's settings. Find chat history, memory, and the training opt-out. Write down what you changed and why, and one sentence from the Model Spec or Claude's character that explains a refusal you have seen.

  5. Prompt injection, in plain words: why pasted content can hijack your assistant60 min

Hours 30 to 340/18

Explain agents, vibe coding, voice agents and AI-made media in one sentence each, and say which of them touch your job this year.

In the Encyclopedia: AI Agents, Vibe Coding, Voice Agents

  1. What an AI agent is (and is not)60 min

  2. Directing work with an agent: OpenAI Academy, Apply AI at Work60 min

    • Course
      OpenAI Academy · free official learning hub · OpenAI
    • Do
      Do one lesson from the Apply AI at Work track on repeatable workflows or directing agents. Rewrite one step of your weekly workflow as instructions you would hand to an agent, with the result you expect and the check you will run.

  3. How much should an agent decide on its own?55 min

  4. Vibe coding, as a tourist60 min

  5. Voice agents, voice clones, and AI-made media at work60 min

Hours 35 to 390/14

Hand your team a one-page 'how I use AI' note and hold a calm conversation about AI and jobs using evidence rather than headlines.

In the Encyclopedia: AI and the Future of Work, AI Ethics, Children, Teens and AI

  1. AI and jobs: the evidence, not the headlines60 min

  2. Elements of AI, chapter 6: implications60 min

  3. Responsible AI: six principles for the top of your note60 min

    • Course
      Embrace Responsible AI Principles and Practices · Microsoft Learn
    • Do
      Do units 1 to 7 (the assessment is optional). Next to each of Microsoft's six principles, write one concrete rule from your own practice that enforces it. These become the first lines of your team note.

  4. Learning with AI without skipping the thinking: schools, teens and companions60 min

  5. Deliver the note55 min

Milestone 3 project · about 3 hours

How I use AI: a one-page note for my team

Write the rules you actually follow with AI, in one page your team can copy.

  • Data safety
  • Disclosure
  • Working with agents
  • Judgment

Courses that teach this

Build it

  1. Write what goes into a chat and what never does.

  2. Write how you check an answer, and when you say AI helped.

  3. Write which steps an agent may do alone and which need your approval.

  4. Write where you stop and ask a person.

  5. Share the page with your team and take two edits.

You finish with

A one-page note your team can copy, with principles at the top and the never-paste list, the checking rule and the labeling rule below.

Check your work

  • A teammate can follow it without asking you.

  • It says what never goes into a chat.

  • It says when you stop and ask a person.

  • Two people edited it.

See how others did it

Check yourself at the end

  • You can explain what ChatGPT, Claude, Gemini or Copilot does with your words, and why it can be fluent and wrong at the same time.

  • You run one weekly task through a reusable prompt with a checking step, and you know when to start a new chat.

  • You catch made-up facts, quotes and citations before they leave your hands, and you can trace a brief to its sources.

  • You know which data never goes into a chat, what your tool's settings do with your history, and how pasted content can hijack an assistant.

  • You can say in one sentence what an agent, vibe coding, a voice agent, a deepfake and a context window are, and which of them touch your job.

  • You have given your team a one-page note with the rules you follow, and you can discuss AI and jobs with evidence.