Beginner · About 41 hours · 3 projects · Free

AI Roadmap for Leadership

In about 41 hours you can set an AI strategy on one page, govern its use with a policy you wrote, and decide one use case on its value and its risks, asking the questions that vendors and your own teams must answer.

0of 41 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
    A one-page AI strategy for your organization

    Write the one page your board would read: where AI will be used, what each use should gain, who owns it, and the first 90 days.

    hour 15
  2. 2
    An AI acceptable-use policy

    Write the policy your staff will actually follow: allowed tools, what data goes in, when a person must review, and who decides.

    hour 26
  3. 3
    A decision memo on one use case

    Take one use case through value, cost, risk and vendor, and write a memo the board can say yes or no to in one meeting.

    hour 41

Skills you'll learn

  • Use case selection
  • Measuring AI value
  • Build vs buy
  • AI maturity
  • Planning
  • AI policy
  • Data rules
  • Decision rights
  • Human review
  • Disclosure
  • Risk
  • Business cases
  • Total cost
  • Risk controls
  • Vendor scorecards
  • Stop rules

Prerequisites

  • A chatbot account you can use with non-confidential text.
  • One real use case in your company to carry through the plan.
  • Nothing technical.

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

See it clearly, then set the direction

Understand what AI is and does, then decide where your company should use it.

Hours 1 to 50/15

By the end you can explain to your board, in plain words, the difference between AI, machine learning and generative AI, and point to five places AI already runs in your business.

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

  1. AI in one hour: the map before the detail55 min

  2. Four words vendors mix up: AI, machine learning, deep learning, generative AI65 min

  3. The state of AI in numbers: the Stanford AI Index60 min

    • Read
      Stanford HAI AI Index Report 2026 (free) · hai.stanford.edu
    • Do
      Read the top takeaways and the economy chapter only. Pick five numbers you would quote to your board. For each, write the number, the chapter it came from, and the question it raises for your company.

  4. A university course in an hour: Elements of AI, chapter 165 min

  5. The board paragraph: what AI is for us55 min

Hours 6 to 100/16

By the end you can say how a language model produces an answer, why it can be confidently wrong, what an agent adds, and which approval step you would require before an agent acts for your company.

In the Encyclopedia: Large Language Models (LLMs), AI Hallucinations, AI Agents

  1. What a language model actually does55 min

  2. Karpathy's hour: the one lecture worth your time75 min

    • Watch
      Andrej Karpathy · Intro to Large Language Models (1hr) · Andrej Karpathy
    • Do
      Write five questions for your CTO from the talk: which model we use, who trained it on what data, what it costs us per month, what it is not allowed to do, and how we would know if it failed. Send them.

  3. Hallucination: why confident is not correct55 min

  4. Agents: software that takes steps on its own60 min

    • Watch
      What are AI Agents? · IBM Technology
    • Read
      Anthropic's Building Effective Agents guide · Anthropic
    • Do
      List three multi-step tasks in your company an agent could run (refund handling, supplier follow-up, report assembly). For each, write the one approval step you would require and the one thing the agent must never do.

  5. How much autonomy, and who signs60 min

    • Read
      What is Agentic AI? | IBM · IBM
    • Read
      AI Agents Are Here. What Now? · Hugging Face
    • Do
      Draft a four-rung autonomy ladder for your company: suggest, draft, act with approval, act alone. Give one example of what you would allow on each rung today and who can move a task up a rung.

Hours 11 to 150/17

By the end you have a draft one-page AI strategy: three use cases with the measure each moves, what you will build, buy and stop, and your maturity stage with the evidence for it.

In the Encyclopedia: AI Strategy, AI Maturity Models, Build, Buy or Partner

  1. Where the opportunities are60 min

  2. What AI already pays for in your industry65 min

  3. The transformation playbook, and the business value module65 min

  4. Where you stand: the maturity stage55 min

  5. Build, buy or partner, then write the page65 min

Milestone 1 project · about 4 hours

A one-page AI strategy for your organization

Write the one page your board would read: where AI will be used, what each use should gain, who owns it, and the first 90 days.

  • Use case selection
  • Measuring AI value
  • Build vs buy
  • AI maturity
  • Planning

Courses that teach this

  • Elements of AI

    University of Helsinki · about 30 hours

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

  • Transform your business with AI

    Microsoft Learn · about 2.5 hours

    Four modules on where AI fits in a business and how to adopt it.

  • Discovering Generative AI

    Udacity · 45 minutes

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

Build it

  1. List where AI could help, with the task, who does it now, and how often.

  2. Pick three use cases and name the number each one should move, with today's baseline.

  3. Choose build, buy or boost for each, and say what you will stop funding.

  4. Score your maturity stage with evidence, and name one owner.

  5. Write it on one page with the first 90 days, and have one peer challenge it.

You finish with

One page: three use cases with the number each moves, a buy, boost or build choice for each, the maturity stage with evidence, the owner, and the first 90 days.

Check your work

  • Each use case names one number, with today's baseline.

  • Each use case has a build, buy or boost choice.

  • One person owns it, by name.

  • It says what you will stop funding.

See how others did it

Milestone 2 · Hours 16 to 26 · 2 sections · 10 lessons

Govern it: risk, law and the policy

Learn who decides, what the law says, and how to write rules people follow.

Hours 16 to 200/17

By the end you can say who in your company decides, owns and answers for each AI use, ask your teams the NIST four-function questions, and point to three real incidents in your industry with the control that would have caught each.

In the Encyclopedia: AI Governance, AI Risk Management, AI Incidents and Incident Databases

  1. Governance: who decides, who answers55 min

  2. NIST's four verbs: govern, map, measure, manage50 min

  3. Safety and risk: the words, and the catalog65 min

  4. Incidents: learn from other people's failures65 min

  5. Responsible AI in practice: the Microsoft Learn module70 min

    • Course
      Embrace Responsible AI Principles and Practices · Microsoft Learn
    • Do
      Write the responsible-AI paragraph for your board: six commitments in your own words (fairness, reliability, privacy, inclusion, transparency, accountability) and the five controls that make them real in your company.

Hours 21 to 260/20

By the end you have a draft AI acceptable-use policy, and you can say which laws touch your AI use (the EU AI Act's risk tiers, GDPR Article 22, copyright) and what each one asks of you.

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

  1. The EU AI Act, and why it reaches US companies60 min

  2. US rules: executive orders, states and agencies60 min

  3. Data protection: decisions about people65 min

  4. Copyright, training data, and what you own65 min

  5. Shadow AI, and the policy60 min

Milestone 2 project · about 3 hours

An AI acceptable-use policy

Write the policy your staff will actually follow: allowed tools, what data goes in, when a person must review, and who decides.

  • AI policy
  • Data rules
  • Decision rights
  • Human review
  • Disclosure
  • Risk

Courses that teach this

Build it

  1. List the AI tools already in use, including the ones nobody approved.

  2. Decide what data may go in, what never does, and when a person must review the output.

  3. Write the decision rights: who approves, who owns, who answers, with a name in every box.

  4. Add disclosure, exceptions, and what happens on a breach.

  5. Cut it to one page and send it to your counsel for review.

You finish with

A one-page acceptable-use policy plus a half-page decision-rights chart with a named owner in every box, ready for your counsel's review.

Check your work

  • A new employee can tell what is allowed in one read.

  • Every decision has a named owner.

  • It covers data, review, disclosure, exceptions and breach.

  • It is ready for your counsel to review.

See how others did it

Milestone 3 · Hours 27 to 41 · 3 sections · 15 lessons

Decide one use case on value, cost and risk

Learn to judge one use case on value, cost, vendor and people.

Hours 27 to 310/17

By the end you can state one use case's value in money or hours against its full cost (model, people, data, governance), read a productivity study without being fooled, and say which data it needs and whether you may use it.

In the Encyclopedia: Measuring AI Value and ROI, Inference Cost & Token Economics, Data Strategy for AI

  1. ROI and total cost of ownership55 min

    • Read
      How to maximize AI ROI in 2026 · IBM Think
    • Read
      What Is Total Cost of Ownership (TCO)? · IBM Think
    • Do
      Build the value sheet for your first use case: today's baseline, the expected change, and every cost line (licenses or tokens, people, data work, governance, training). Leave a blank where you do not know; the blanks are your next questions.

  2. The evidence: what measured gains look like60 min

  3. The J-curve and the pilot trap60 min

  4. What it costs to run: tokens and energy60 min

  5. Data: the ceiling on every AI project65 min

Hours 32 to 360/18

By the end you have a ten-question vendor scorecard, can tell a benchmark score from a result on your own data, and know the one security question every AI vendor must answer.

In the Encyclopedia: Evaluating AI Vendors and Procurement, LLM Evaluation & Benchmarks, Prompt Injection & Jailbreaking

  1. Procurement basics: evidence, not demos55 min

  2. Transparency and standards you can demand60 min

  3. Benchmarks are not your results65 min

  4. Security: the question every vendor must answer60 min

  5. Build, buy, open or closed, and lock-in65 min

Hours 37 to 410/16

By the end you have the decision memo on one use case (value, cost, risks, owner, measure, stop rule) and a people plan: who is trained, what changes in the job, and what you will tell staff about roles.

In the Encyclopedia: AI Adoption and Change Management, AI and the Future of Work, AI Center of Excellence and Operating Model

  1. Why AI projects fail, and the change work65 min

  2. The jagged frontier: where the tool helps and where it hurts60 min

  3. Jobs: what the data says, and what you will tell staff65 min

  4. Operating model: central, federated or hybrid65 min

    • Course
      Transform your business with AI · Microsoft Learn
    • Read
      What Is an AI Center of Excellence? · IBM Think
    • Do
      From the Microsoft path, do only module 4, Scale AI in your organization (23 minutes). Decide your operating model for the next year and name the owner of standards, skills and governance. Write the three things the central team does and the three things the business units do.

  5. The decision memo60 min

Milestone 3 project · about 5 hours

A decision memo on one use case

Take one use case through value, cost, risk and vendor, and write a memo the board can say yes or no to in one meeting.

  • Business cases
  • Total cost
  • Risk controls
  • Vendor scorecards
  • Stop rules

Courses that teach this

Build it

  1. Pick one use case from your strategy page and write today's baseline.

  2. Build the value sheet: expected gain against the full cost, including people and governance.

  3. List the risks and the control for each.

  4. Score the vendors with your ten-question scorecard.

  5. Name the owner, the measure, the review date and the rule that stops the project.

  6. Write the two-page memo so the board can say yes or no in one meeting.

You finish with

A two-page decision memo with a value sheet, a scored vendor scorecard and a stop rule, written so the board can say yes or no in one meeting.

Check your work

  • Value is compared with the full cost, including people.

  • Every risk has a control.

  • The vendor score comes from your own questions.

  • A written rule says when the project stops.

See how others did it

Check yourself at the end

  • You can explain to your board what AI, machine learning and generative AI are, why a language model can be confidently wrong, and what an agent adds, without a slide from a vendor.

  • You have a one-page AI strategy with three use cases, the measure each moves, a build, buy or boost choice, and a maturity stage backed by evidence.

  • You have an acceptable-use policy and a decision-rights chart with a named owner for every AI use, and you can ask your risk team the NIST govern, map, measure and manage questions.

  • You can read a vendor's claim and separate a benchmark score from a result on your own data, and you know the security, data, liability and lock-in questions to ask before signing.

  • You can state a use case's value against its full cost, including people, data and governance, and you know why the gains lag the spend and when to check.

  • You have a two-page decision memo on one use case with its owner, measure, review date and stop rule, and a paragraph you can say to staff about what changes in their jobs.