Six stages, one useful change at a time

Learn the method. Put it to work.

Start with the basics. Build a real workflow, inspect what it produces, and leave with instructions another person can repeat.

24 practical lessonsNo account required
Find my starting pointChoose a workplace path

Your saved progress loads in this browser.

Completion records your own checks. It is not a test score or a certification. You can open any lesson.

Start the first lesson

Follow the order or open the lesson you need. Each lesson combines an explanation, a fictional worked example, an exercise, evidence checks and an export. Completion is your own assessment of the work.

Module 1

Understand the tools

Learn the vocabulary through work you recognize, then define a useful result and a check.

You leave with: A clear task, useful context and the right division between AI, rules and people.

  1. Understand models, assistants, agents and automations

    Name the parts of an AI workflow without needing a technical background.

  2. Define the result before choosing the tool

    Turn a broad AI ambition into an artifact, a recipient and an observable completion check.

  3. Give useful context and trace the claims

    Build a small evidence pack and keep facts, assumptions and missing information distinct.

  4. Decide what AI, software and people should do

    Separate interpretation from exact calculations and decisions that require authority.

Module 2

Map the work

Make the current process, its constraints and its hidden effort visible before changing it.

You leave with: One suitable task with an owner, a boundary and a baseline.

  1. Map the work people actually do

    Expose inputs, handoffs, waiting and rework before deciding where AI belongs.

  2. Choose a first task you can judge

    Select useful work with available inputs, inspectable outputs and manageable consequences.

  3. Find the constraints that change the plan

    Separate data access, model processing, external actions and business authority.

  4. Record how the work performs today

    Observe the complete task so a future pilot has a fair comparison.

Module 3

Redesign the workflow

Choose the platform, context, responsibilities and recovery that the actual job requires.

You leave with: A reviewable plan with explicit ownership and dependencies.

  1. Choose a platform from the work and constraints

    Compare relevant options against requirements without relying on a universal vendor ranking.

  2. Write instructions that survive the handoff

    Turn your task and constraints into a concise brief an assistant or colleague can use.

  3. Give every agent a job and an owner

    Coordinate independent work while preserving dependencies and one writer per output.

  4. Design the check and the way back

    Place review where it prevents a bad result and define recovery before external changes.

Module 4

Run a small pilot

Use representative cases and real artifacts to find out whether the new workflow helps.

You leave with: A bounded trial, inspectable results and an honest exception log.

  1. Write a pilot you can actually evaluate

    Define the trial's cases, boundaries, owner and decision before running it.

  2. Test ordinary work, messy work and when to stop

    Create a small evaluation set whose expected behavior you can explain.

  3. Turn an approved source into useful marketing drafts

    Practice audience choice, distinct messages and claim checking without inventing results.

  4. Calculate first, then write the decision brief

    Reconcile a small dataset and separate observed results from explanations to investigate.

Module 5

Measure what changed

Check evidence, compare the whole job and make a decision that others can inspect.

You leave with: An adopt, adapt or stop decision supported by observed quality, effort and cost.

  1. Review the output against evidence, not confidence

    Inspect claims, calculations and behavior using the checks defined before the run.

  2. Measure the whole job, including review

    Compare equivalent accepted outputs and keep quality, effort, elapsed time and cost visible.

  3. Decide whether to adopt, adapt or stop

    Use the pilot evidence to make a bounded decision instead of treating every trial as a success.

  4. Leave a decision someone else can understand

    Record the choice, evidence, alternatives and conditions without preserving an entire chat transcript.

Module 6

Make it repeatable

Turn a useful trial into an owned workflow, then maintain and expand it deliberately.

You leave with: A working playbook another person can run and improve.

  1. Turn the accepted workflow into a playbook

    Package the conditions, inputs, instructions and checks so someone else can repeat the work.

  2. Make the operating decisions explicit

    Document approved inputs, tools, actions and escalation without pretending a template is a compliance verdict.

  3. Keep the workflow useful after launch

    Assign ownership for changes, failures, costs and the checks that should be rerun.

  4. Expand the scope only when the evidence transfers

    Choose the next useful change while recognizing new data, users and actions as new conditions.

Apply the foundations to your work

Marketing, business intelligence and building or automating a useful tool each have a practical path and a capstone exercise.

Explore the three paths