A training curriculum outline for AI adoption: modules by role, session objectives, practical exercises and the assessment that proves capability. It is the practical companion to Training Your Team to Use AI Well and AI Skills by Role.
The curriculum is designed to be delivered alongside a workflow rollout, not before it. Training without a workflow gives people something to practise in the abstract, which is why so much of it is forgotten.
The unit of teaching is the workflow, not the tool.
The structure
Six modules over six weeks, roughly six hours per person.
| Module | Audience | Duration | Objective |
|---|---|---|---|
| 1. Context and limits | All staff | 1 hour | Know what AI does well and badly, and what the business has decided |
| 2. The rules | All staff | 1 hour | Know the data rule and the disclosure position |
| 3. The workflow | By team | 90 minutes | Be able to run the six steps on the team’s own task |
| 4. The playbook | By role | 1 hour | Know the prompts, the inputs and the output format |
| 5. Practice | Individual | Self-paced | Run the workflow on live work with review |
| 6. Verification drill | By team | 1 hour | Catch a seeded error, and know what to record |
Modules 1 and 2 are universal; 3 to 6 are role-specific and repeat for each workflow.
Module 1 — Context and limits
Objective: Each learner can state three tasks AI does well and three it does badly in their role.
Content: What AI does well (drafting, transforming, summarizing); what it does badly (arithmetic, citations, judgement with consequences, anything requiring internal context); what the business has decided to do first.
Exercise: Each learner lists two tasks from their own week that fall on each side of the line, and the group compares.
Why it matters: Most AI disappointment comes from using the tool on a task it is not suited to. Twenty minutes on the boundary prevents a quarter of wasted effort.
Module 2 — The rules
Objective: Each learner can state what may and may not be entered into a tool, and who to ask.
Content: The approved-tool list; the prohibited data categories with examples relevant to the audience; the default when uncertain; the person to ask; the disclosure position.
Exercise: Ten short scenarios — a client contract, a personnel file, a pricing sheet, a public article, a supplier email — sorted into permitted and prohibited, with the reasoning discussed.
Assessment: Pass requires correct handling of the scenarios, including the “unsure — ask” default.
Module 3 — The workflow
Objective: Each learner can run the six steps on their own task.
Content: Input, draft, check, correct, approve, record — demonstrated end to end on one real deliverable from the team.
Exercise: The group runs one cycle together, with the trainer narrating the check and the record step.
Why it matters: The six steps are the transferable skill. A learner who knows the pattern can apply it to the next workflow with far less support.
Module 4 — The playbook
Objective: Each learner can use the playbook entries for their role unaided.
Content: The five parts of an entry; how to supply the inputs; what the output format means; what must be checked.
Exercise: Each learner runs one playbook entry on their own work and reviews the output against the checks.
Module 5 — Practice
Objective: The workflow runs on live work, unaided, to the expected standard.
Content: No new teaching. Each learner runs two live cycles with review and feedback.
Design note: This module is self-paced and is where most of the skill is built. It is also the module most often cut, and cutting it is why training fails.
Module 6 — Verification drill
Objective: Each learner can catch a fabricated citation, a plausible wrong figure and an embedded commitment.
Content: Three prepared outputs, each containing one seeded error.
Exercise: The learner identifies the error in each, and states how they found it.
Assessment: Passing the drill is the standard for the verification skill, and it is reported as part of the adoption measures.
The assessment
A practical assessment, not a test.
- Run it. Complete one workflow cycle on live work, unaided.
- Explain it. State what was checked and why, in your own words.
- Catch it. Identify a seeded error in an output.
The third part is what proves the skill, and it is the part most often omitted from training programs. A person who can run a workflow but cannot catch an error does not yet have the capability the business needs.
Frequently asked questions
How long does the curriculum take?
Around six hours per person for one workflow, delivered over six weeks rather than in a single session.
Who should attend which modules?
Modules 1 and 2 are for everyone. Modules 3 to 6 are by team and role, and repeat for each new workflow.
Why is training spread over weeks?
Because the skill is built through repetition on real work. A single workshop transfers knowledge and not capability.
What if we cannot release people for six hours?
Then the workflow is likely adding work rather than removing it. Fix the workflow first, and the time argument usually resolves itself.
Do leaders need a different curriculum?
Leaders need modules 1 and 2, plus a short session on choosing use cases, owning the policy and reading the measures. What they specifically need is to use the workflow visibly.
How do we deliver this without a trainer?
An internal owner can deliver it with this outline and the Prompt Playbook. Where no internal owner is available, deliver the first cohort with an external partner and the internal owner alongside, so the capability stays.
Next step
Adapt the six modules to your first workflow, deliver the first two to everyone, and add the verification drill before the first cohort finishes. See Training Your Team to Use AI Well and the Prompt Playbook, or book a team training session and we will deliver it with your team.
Sources
- Curriculum design reflects standard skills-training practice applied to AI adoption: universal rules, workflow-based teaching, live practice and a practical assessment.
No statistic in this asset is invented.