Frameworks & Templates4 min read

AI Skills for Employees: Build a Prerequisite Ladder

Build AI skills for employees in the right order: frame the task, set boundaries, check the result and decide what happens next.

Bokili EditorialPublished October 14, 2026· Verified October 14, 2026

A four-rung AI skills ladder builds from task framing and boundaries to output checking and a final decision

AI skills for employees are often taught as complete workflows: analyse this document, draft that response, automate this process. When a learner fails, the course repeats the whole task. That response is expensive and imprecise. A better approach is to identify the lowest missing prerequisite, practise it alone, then rebuild the workflow from that rung.

Skills England’s 2026 foundation benchmark separates clear instructions, routine task support, tool adjustment, automation, risk awareness and information analysis. European Commission guidance likewise says AI-literacy measures should reflect people’s knowledge, experience and the context and risks of the systems they use. NIST describes human–AI tasks through goals and activities, while OPM training guidance recommends diagnosing the gap between required and current performance. Together, those ideas support a prerequisite ladder: observable behaviours arranged in the order a person needs them.

Build an AI skills prerequisite ladder for employees

Four rungs before a complete workflow

1. Frame the task

State the goal, audience, source material and expected output. If the task is unclear, more prompting technique will not fix it.

2. Set the boundaries

Choose an approved tool and define permitted data, constraints, prohibited actions and the point that requires human approval.

3. Check the result

Test material facts, calculations, sources, constraints and missing information against an explicit review standard.

4. Decide the next action

Use, revise, reject or escalate the output. The employee must know who owns the decision and what evidence the decision needs.

The ladder is not a maturity score. An employee may be strong at checking a summary but weak at setting the data boundary for a new tool. Diagnose the rung for the specific task. That keeps support narrow and prevents a broad label such as “beginner” from hiding the actual learning need.

Worked example: diagnose the lowest missing rung

A logistics coordinator receives fictional shipment notes and must draft a supplier-delay brief. The output is clear, but it includes a customer name in a tool that is not approved for personal data and presents an estimated arrival date as confirmed. Repeating the entire exercise would mix two different gaps. The first missing prerequisite is Bound: choose the approved environment and remove or replace restricted data. Once that is fixed, Check becomes visible: distinguish confirmed dates from estimates.

Targeted practice for the same task

1
Frame
Given a short source pack, write the goal, audience and required three-part brief without drafting the brief.
2
Bound
Select the approved tool, identify prohibited data and replace the customer name with a safe placeholder.
3
Check
Mark every date as confirmed, estimated or unknown and trace it to the relevant shipment note.
4
Decide
Choose whether to send, revise or escalate the brief and name the person who must confirm the arrival date.

Reading is a start. Practice makes it stick.

Start learning

	Repeat the whole workflow	Train the missing prerequisite
Diagnosis	Failure appears as one low score.	The first broken rung is named.
Practice	Learner repeats every step.	Learner practises one behaviour with fast feedback.
Evidence	A polished final output can hide the gap.	A small artefact shows whether the behaviour changed.
Return to work	Support ends after resubmission.	The fixed rung is tested again inside the full task.

This sequence makes coaching more useful. The learner does not spend time redoing strengths, and the coach can give one precise correction. It also exposes gaps that training cannot solve. If nobody knows which tool is approved or who owns escalation, the missing element is an organisational rule, not an employee skill.

Turn the ladder into a learning path

Design each rung so it can be observed

Use one representative workplace task, not a generic feature tour.
Write the required behaviour as an action a reviewer can see.
Create a small artefact: a task frame, boundary note, check record or decision receipt.
Define a pass condition before practice begins.
Use fictional or approved sample data.
Give feedback on the lowest failed rung first.
Return to the full workflow only after the rung passes.
Record policy or tooling gaps separately from learning gaps.

This article complements, rather than replaces, Bokili’s role-based skills matrix at https://bokili.com/en/learn/ai-skills-for-employees-matrix. The matrix decides which behaviours a role needs; the prerequisite ladder decides the order in which to build them. Use the three-task baseline at https://bokili.com/en/learn/ai-training-employees-three-task-baseline to choose representative work, and pair creation with verification through https://bokili.com/en/learn/ai-learning-path-creation-verification-pairs. For company-wide practice and progress tracking, see https://bokili.com/en/for-hr.

Diagnose one failed task in ten minutes

Choose one recent practice task that did not pass.
Write one observable pass condition for Frame, Bound, Check and Decide.
Find the lowest rung that lacks evidence.
Create a three-minute practice that isolates only that behaviour.
Review the small artefact and give one concrete correction.
Repeat the full task only after the isolated rung passes.

Teach the next necessary behaviour

A strong learning path is not a pile of increasingly complex workflows. It is a sequence of prerequisites that lets each employee practise the next necessary behaviour. Start with the lowest missing rung, keep the evidence small and observable, and reconnect the skill to real work. That is how AI skills for employees become usable capability instead of course completion.