Learning Hub / Knowledge library / Build team AI competence
FIELD GUIDE 23 · Practitioner

Build competence around the decisions people make.

Design role-based AI learning with observable tasks, feedback, reassessment, and evidence of transfer to work.

4 min read · Plus practice · Editorial update: 23 October 2026
KNOW FIRSTDesign meaningful human review →
YOUR DOMAINAI Strategy & Implementation
GO NEXTEvaluate AI vendors →
What you will be able to do

Create a role-to-task learning plan and an assessment that checks behavior rather than attendance.

Define the capability you need

A learning objective names something a person can demonstrate. “Understands AI” is difficult to assess; “identifies an unsupported claim and checks the original source” is observable.

A work boundary states where the capability applies. Reviewing a routine internal summary does not establish readiness to approve every consequential AI-assisted decision.

A starting check uses a short fictional task before teaching. Look at the reasoning and actions, not only the final answer, to identify which skill needs practice.

Map roles to decisions

Example competence map for a fictional support team
RoleDecision to practiceEvidence to retain
Drafting colleagueUse the current policy and flag missing evidenceAnnotated draft with source references
ReviewerCatch an exception and withhold approvalReview decision and correction rationale
Workflow ownerRestrict or pause a failing workflowResponse record with owner and retest
Team leadAllocate practice for a demonstrated gapLearning plan tied to task evidence

One person may hold several roles. Assign responsibilities explicitly rather than assuming a job title determines what someone is authorized to approve.

Teach one decision at a time

A useful learning cycle starts with a worked example, then a similar task with guidance, then an unseen task. Ask the learner to explain which evidence changed the decision.

A deliberate variation changes the fact that matters: an outdated policy, a missing source, or an exception hidden in otherwise correct text. Rephrasing the same answer may test recall more than transfer.

Accessible practice provides readable source material and enough time to inspect it. Offer an equivalent format when the exercise format would otherwise obscure the skill being assessed.

Work through an approval mistake

The fictional exercise asks a reviewer to inspect a refund draft. The policy permits refunds within thirty days, except opened custom orders. The draft approves an opened custom order placed ten days ago.

The first learner response checks the date and approves. It correctly applies the general rule but misses the relevant exception. Calling the whole task correct would conceal the capability gap.

The feedback points to the exception and asks the learner to explain its effect before revising the draft. The next attempt uses a different scenario with another exception, not the identical answer.

Use a rubric that exposes the gap

Three separate checks ask whether the learner selected the current source, applied the relevant exception, and took the appropriate review action. Record a brief observation for each.

Our example learner meets the source check but misses the exception and approval checks. A high average from unrelated questions must not substitute for the missing review skill.

Reviewer calibration has two reviewers independently assess a small shared set, compare disagreements, and clarify the rubric. Decide in advance which criteria are required for the particular responsibility.

Check transfer and maintain the plan

Completion evidence shows that an activity was finished. It does not by itself show that the person can apply the skill later in a changed task.

A follow-up check uses an approved or fictional work-like case after a period chosen for the team. Record support needed and remaining gaps; do not turn uncertain observations into a claim of universal competence.

A review trigger includes changes to the workflow, source policy, tool permissions, or role. Keep only necessary learning records under the organization’s access and retention rules. This exercise is not a certification or legal compliance assessment.

Practice: design the next learning cycle

Choose two roles from the table. For each, write an observable objective, a fictional task, three rubric checks, targeted feedback, and an unseen follow-up case. Explain what evidence would make you assign more practice before expanding responsibility.

Compare with the worked answer

Worked answer: For a reviewer, use “detect the exception and withhold an unsupported approval.” Provide the policy and a plausible but incorrect draft. Check source selection, exception handling, and approval action separately. After feedback, use a new product and policy exception. For the workflow owner, provide repeated exception failures and ask for a bounded response with an owner and recovery check. Attendance, confidence ratings, and course completion can inform the plan, but none replaces those task observations.

Use the method on your own task

Editable worksheet: Download the Markdown template and open it in a text editor. Use fictional or approved information; no account or paid service is required.

Download the worksheet (.md) ↓

What to carry forward

Create a role-to-task learning plan and an assessment that checks behavior rather than attendance.

Show your reasoning with the exercise above. Keep the evidence, the decision, and the remaining uncertainty together.

Useful terms for this lesson

BaselineWorkflow

Sources & further reading

NIST AI RMF Core: GOVERN 2 and GOVERN 6 ↗

AI Competence: Learning routes ↗

The practical examples and exercises on this page are illustrative. Source links provide background; they do not certify the exercises or endorse this hub.