Teaching library
Concept lessons
The teaching library behind the quizzes. Every concept the course builds on is explained the way you would explain it to a board. Read them to learn; the quizzes will test whether it stuck.
AGENT as a decision path
A decision path for moving an AI idea through diagnosis, design, review, and operation.
- ▸Audit establishes the conditions around the decision.
- ▸Gauge names the constraints that affect the work.
- ▸Engineer shapes and tests the proposed workflow.
- ▸Navigate assigns accountability and support.
- ▸Track creates a review habit around evidence.
The executive decision
An AI decision affects strategy, operating model, and workforce. Its owner needs authority to connect those areas.
- ▸A strong case names the business outcome and the people who receive it.
- ▸An effective mandate includes a named owner, decision rights, and a review point.
- ▸An executive brief gives the board a decision, evidence, and a next action.
Capability and work design
AI capability changes over time. Leaders need a grounded view of current limits and the tasks that may change first.
- ▸Periods of disappointment follow claims that outrun reliable capability.
- ▸A task map gives a clearer view of work than a job title.
- ▸The leader decides where judgment stays with people and where a system can assist.
Use-case prioritization
Score candidate AI use cases on value and feasibility to build an opportunity map and a prioritized shortlist.
- ▸Opportunity map = a structured view of use cases across functions.
- ▸High value + high feasibility = do first (quick wins).
- ▸High value + low feasibility = invest in enablers (data, talent).
- ▸Low value = deprioritize regardless of feasibility.
Work redesign and trust
AI changes can affect identity, status, and daily work. Leaders need a practical response that earns participation.
- ▸Name the concerns people bring to the change.
- ▸Give affected teams a visible path toward new capability.
- ▸Share what is known, what remains open, and when decisions will be reviewed.
Model judgment for leaders
A working vocabulary for model choice, generative systems, agents, and the limits that shape executive decisions.
- ▸Supervised learning uses labeled examples to predict an outcome.
- ▸Language models produce plausible text from learned patterns. Their confidence requires review.
- ▸An agent combines a model with tools, memory, and a task loop.
Data, context, and access
Data quality and access are a leading cause of stalled AI work. Readiness is defined for a specific use case.
- ▸Silos trap data in separate systems and block access.
- ▸Data readiness means the data exists, is accessible, is good enough, and is safe to use.
- ▸Infrastructure in plain terms: warehouse for clean data, lake for raw data, pipelines for movement, APIs for access.
Prototype the working moment
A prototype is evidence that answers one question. Non-technical leaders build it with no-code tools to win support.
- ▸A prototype answers one decision with evidence. Everything else is deferred.
- ▸Match the tool to the workflow and to what the organization allows.
- ▸Scope a one-week build around one job, one decision, and a real data slice.
Choose the operating shape
A plain-language view of the AI stack and the build, buy, or partner decision that follows the workflow, data, and capacity.
- ▸Retrieval grounds an answer in the organization's own data and reduces unsupported claims.
- ▸Rent the frontier model, build the integration, and keep the data.
- ▸Score every option against the organization's data reality on a weighted card.
Accountability by design
Governance is decision rights. The core is a living risk register with owners, mitigations, and a review cadence.
- ▸A risk register entry has a category, a score, a tier, an owner, and a mitigation.
- ▸Use a recognized framework as a skeleton: govern, map, measure, manage.
- ▸Every system has one accountable person and one escalation path.
- ▸Consent has limits when the risk lives in what is inferred from a person.
Regulatory navigation
Regulatory risk is often tiered by the nature of the system. The tier drives the obligations, and they arrive on a calendar.
- ▸Some uses are prohibited. Some carry the strictest obligations. Some carry only disclosure duties.
- ▸Map each use case to its tier, because the obligations follow the classification.
- ▸Obligations arrive on a fixed calendar that should be re-verified at implementation time.
Judgment under uncertainty
Ethical review is a decision process. Bias can enter at several points, and a fairness standard must be chosen and owned.
- ▸Bias can enter the data, the labels, the model, the deployment, and the feedback loop.
- ▸Choose one fairness standard and own the trade-off.
- ▸Check outputs before and after deployment.
- ▸Consent is the wrong tool when the risk lives in an inference.
Value and evidence
Value comes in four forms. The hardest part is attribution, which requires a baseline and a credible range.
- ▸Name one value type per use case and prove it.
- ▸Set the baseline before launch or argue about attribution forever.
- ▸Report a credible range with stated confounders, not a single hero figure.
Adoption as an operating capability
Pilots stall before production. Scaling needs a phased rollout, adoption measures, and a reskilling plan.
- ▸Design for production, not pilot, conditions. Scale in waves.
- ▸Adoption needs capability, confidence, and capacity. Find which one is missing.
- ▸Reskill in three tiers: fluent leaders, augmented professionals, and specialists.
The AI Transformation Playbook
The consolidated executive case built around one use case. It earns the mandate to execute.
- ▸Specific to your organization.
- ▸Defensible, because it is built on evidence.
- ▸Board-ready, because the summary comes first and the detail follows.
- ▸Actionable, because someone could fund and run it.